





          Illuminator User's Manual   Release 0.10
                              
                              
                              
                  Illuminator User's Manual
                              
                              
                              
                      Release 1.0alpha0
                              
                              
                              
                              
                              
                              
                              
                              
                              
                              
                              
                              
                              
                              
                              
                          11/11/95
                              
                              
                    RAF Technology, Inc.
               16650 NE 79th Street, Suite 200
                     Redmond, WA  98052
                     Tel: (206) 867-0700
                    FAX:  (206) 882-7370
                   E-mail: illcomm@raf.com
                              
                              
                              
                              
                     _Table of Contents
                              

Table of Contents                                        i

Overview                                                 1
     Status of Illuminator                               2

The Illuminator Vision                                   3
     DAFS: Document Attribute Format Specification       3
          Entities, Bounding Boxes, and Contents         3
          Definition of Terms: Hierarchy and Traversal   4
     GUI                                                 5
     International Text                                  5
          Japanese                                       6
     Text and Image Storage Formats                      6
     Image Mode                                          8
     Out-of-Context Mode                                 9
     Flagged Mode                                        10
     Hierarchy Mode                                      10
     Text Mode                                           12
     Future Enhancements                                 14

Using Illuminator                                        15
     Other Documents                                     15
     Starting Up                                         15
     Command Line Options                                15
     Main Menu and Action Bar Options                    16
          Main Menu                                      17
          Action Bar                                     17

The Main Menu                                            18
     The File Menu                                       18
          New                                            18
          Open ...                                       19
          Save                                           20
          Save As ...                                    20
          Revert to Saved                                21
          Next File                                      21
          Previous File                                  21
          Page Up                                        21
          Page Down                                      21
          Page Right                                     22
          Page Left                                      22
          Quit                                           22
     The Edit Menu                                       22
          Delete                                         23
          Kill Contents                                  23
          Previous                                       24
          Next                                           24
          Up                                             25
          Down                                           25
          Find Next Flag                                 26
          New Entity                                     26
          Enter Unicode ...                              26
          Japanese Input ...                             27
     The Display Menu                                    31
          Keyboard...                                    32
          View Selection...                              32
     The Preferences Menu                                35
          Display in Image Mode...                       36
          Type Properties...                             38
          Text on Image                                  40
     The Page Menu                                       40
          Select ...                                     40
          Forward                                        41
          Back                                           41
     The Entity Menu                                     42
          Center                                         42
          Rotate Image                                   42
          Bounding Box Adjustment Options                42

The Action Bar                                           44
     File Status Label                                   44
     The Mag Menu: Image Magnification                   44
     The Modes Menu: Changing Views                      44
          A Word About Moving Between Modes              45
     The Keyboard Menu: Foreign Scripts                  45
     Special Features                                    46
          Image Mode                                     46
          Out-of-Context Mode                            47

Working with Illuminator's Display Modes                 48
     Working in Image Mode                               48
          Moving on the Page                             48
          Displaying Entities                            48
          Selecting and Deleting Entities                48
          Editing Contents of Entities                   49
          Creating a New Entity                          49
          Moving a Bounding Box                          49
          Resizing a Bounding Box                        49
     Working in OOC Mode                                 49
          Moving in OOC Mode                             50
          Selecting and Relabeling Glyphs                50
          Relabeling a Glyph with a String               51
     Working in Flagged Mode                             51
          Setting Flags for Flagged Mode                 51
          Moving in Flagged Mode                         51
          Editing Flagged Entities                       51
     Working in Hierarchy Mode                           52
          Entity Selection                               52
          Rearranging the Hierarchy                      52
          Newly Created Entities                         52
          Collapsing and Uncollapsing                    53
          Viewing and Editing Entities                   53
     Working in Text Mode                                53
          Displaying Text                                53
          Moving in Text Mode                            54
          Selecting and Editing Text                     54
     Creating a Glyph Training Set And Ground Truth      54

Appendix A.                                              57
     Illuminator Command Accelerators                    57

Appendix B. Trademarks, Copyrights and Acknowledgments   59

Index                                                    60


                          _Overview
                              
Illuminator is an editor created for building document
understanding test and training sets, for easy correction of
OCR (Optical Character Recognition) errors, and for reverse-
encoding the essential information and attributes of a
document. It is being developed by RAF Technology, Inc., a
provider of custom OCR and pattern recognition applications,
under the auspices of ARPA (Advanced Research Projects
Agency). Illuminator edits documents that have a combination
of text and images, displaying either or both. It is
configured to handle text in many major European languages
and Japanese. The resulting error-free documents can be used
as training data for document recognizers, as a reference
set (ground truth) for testing such recognizers, and for
entry into databases. Illuminator is intended to be used by
those in the OCR field to create training and reference
sets, and by those in the conversion business for providing
accurate on-line data to the customer.

Illuminator uses a new document format called DAFS (Document
Attribute Format Specification), which is being defined by
RAF Technology and ARPA. This new format is specifically
designed for applications involving OCR and document
decomposition. DAFS provides fundamental structures called
entities for representing image, bounding boxes defining
regions of image, and associated text, and provides means
for other developers to store their own proprietary data.
Through DAFS, developers will be able to exchange training
and test documents and work together on shared problems.
Illuminator is a GUI (Graphical User Interface) for viewing
and altering data in DAFS files. It can read and write
Unicode and ASCII/UTF text files, and can read TIFF and PDA
images.

Illuminator's GUI offers five modes of operation. Image Mode
is for viewing and working with the actual image of a
document and its entities, which are specific regions of
image and associated textual content. Text Mode, a simple
text editor, supports many major scripts and languages. Out-
of-Context Mode collects entities of the same type into a
single file (e.g., all the `A's from a page), where errors
may easily be corrected. Flagged Mode provides easy means of
correcting entities flagged as questionable by OCR.
Hierarchy Mode let you view the entity structure at a
glance.

This manual provides a summary of the current status of
Illuminator, a description of its features, and how to use
them. The Illuminator distribution includes documentation,
header files, binaries, and sample programs for the DAFS
API. Refer to the DAFS Library Programmer's Guide and
Reference for more information on DAFS and the DAFS Library.
Comments, suggestions, and reports of bugs may be sent to:
                      Comments:   illcomm@raf.com
                      Bugs:     illbugs@raf.com
                      Fax:      (206) 882-7370
                      Voice:    (206) 867-0700
                    Status of Illuminator
                              
This section discusses those features that have been added
or modified since the last release of Illuminator.

The New Preferences Menu combines several items formerly
found in the Display Menu, some of which have expanded
capabilities. The new menu contains features permitting the
user to customize the way documents are displayed. These are
Display in Image Mode (formerly Choose), Type Properties (an
expanded version of Color Table), and Text on Image. The
Display Menu now contains only Keyboard Map Display feature
and View Selection.

The Type Properties feature of the Preferences Menu is
capable of much more than its predecessor, Color Table. It
still allows you to select the colors in which entity boxes
and content will be drawn for different entity types in
Image Mode, and now Hierarchy Mode. In addition, entity
types can be customized in Text Mode via the toggle features
"Place on its own page in Text Mode" and "Edit as a separate
region in Text Mode".

The utility and appearance of Hierarchy Mode have been
improved. Now you can rearrange the Hierarchy by selecting
one or a block of entities and dragging them to a new
position. The Hierarchy is now represented by a more tree-
like structure. Entities are denoted by lozenge-shaped
markers which are drawn in the color specified by the new
Type Properties feature.

Hand-in-glove with these changes to Hierarchy Mode, the Edit
Menu now has a New Entity feature, allowing new children to
be created for existing entities. The child has the same
bounding box as its parent and type "Type". Both box and
entity type can now be edited from View Selection. New
entities appear in Hierarchy Mode below their parents.

With this release, Illuminator adds Arabic to its language
repertoire. There are Korean fonts, but as yet no Korean
input method.

In Out-of-Context Mode, it is now possible to enter a string
of characters as the content of a selected Glyph entity by
using the new text entry bar at the bottom of the screen.
Select a character, then press escape to access the bar.
Glyphs labeled with strings are moved to a new page which
contains all Glyphs labeled with the same string.

Rotate Image in the Entity Menu rotates the entire image
clockwise 90.

                              
                              
                   _The Illuminator Vision
                              
        DAFS: Document Attribute Format Specification
                              
While many formats exist for composing a document from
electronic storage onto paper, no satisfactory standard
exists for the reverse process. DAFS has been designed from
its inception to be a standard for document decomposition,
with applications such as OCR and data-base building in
mind. Illuminator is designed to work with DAFS files. Its
flexibility and unique approach to handling document images
and text can be largely attributed to DAFS. DAFS and a
library of C routines for manipulating DAFS files are
describe in the DAFS Library Programmer's Guide and
Reference.

DAFS provides a format for breaking down documents into
standardized entities, defining entity boundaries and
attributes, and labeling their textual contents. Entities
may also have other attributes called Properties (font, for
example). The user may define new Properties as well,
tailoring them to suit the individual application.

DAFS is specifically designed to handle a large variety of
languages and types of scripts. Illuminator's handling of
these scripts is discussed in the "International Text"
section.

Entities, Bounding Boxes, and Contents

A document may be composed of many different kinds of
entities defined under DAFS. Some example types are:
Document, Paragraph, Line, Word, Glyph, Zone, Region, and
Space. Users are free to invent their own entity types as
well. Each entity is a region of image, usually delimited by
a rectangular "bounding box". DAFS stores the bounding box
coordinates in this format: upper left x and y, width and
height. For example, a Word entity is represented by a
bounding box surrounding the image of a word on screen.
Entities are arranged in a hierarchy. Within a typical
document, the image of a line of text will be associated
with a Line entity which may have Word child entities; each
Word may have Glyph children; each Glyph may be labeled with
its textual content (generally a character). All of these
entity types are merely conventional, and users may create
their own entities and their own schemes for organizing the
data hierarchically.

Entities can have image content (the region of image inside
the bounding box) and textual content. In general the terms
"content" and "contents" are used in this manual to refer to
the textual content of an entity. Content can be a single
character, a string, an integer, or a type of data named by
the user. When both image and text are on display, an
entity's content is shown as character(s) appearing below
the bounding box.

        Definition of Terms: Hierarchy and Traversal
                              
 We say that a DAFS structure is a hierarchy or a tree. To
 explain the concept of a hierarchy, we can use the
 example of the organization of the United States Armed
 Forces. The President, who is commander-in-chief, is at
 the top of the hierarchy; we call the top of a hierarchy
 the root. Underneath the President, is the Secretary of
 Defense and the Joint Chiefs of Staff, and underneath the
 Secretary of Defense are the Secretaries of the Navy,
 Army, and Air Force. For the purposes of this example, we
 will ignore the rest of the hierarchy. We would say that
 the President is the parent of the Secretary of Defense
 and the Joint Chiefs of Staff. We say that the
 Secretaries of the Navy, the Army, and the Air Force are
 children of the Secretary of Defense and siblings of each
 other.
 
 Another important concept is the idea of traversing a
 hierarchy. This means listing the elements of the
 hierarchy in an ordered and predictable fashion.
 Illuminator and DAFS tend to use preorder traversal. Here
 are the (recursive) rules for preorder traversal.
 
 1. List the root of the tree.
 
 2. For each child of the root in order, traverse it as
 though it were the root.
 
 In our example, a preorder list would be:
 
 President, Secretary of Defense, Secretary of the Navy,
 Secretary of the Army, Secretary of the Air Force, Joint
 Chiefs of Staff.
 
Some of the sample files you receive with Illuminator may
show unusually shaped or multiple boxes surrounding the
entities. Illuminator can be used to draw and display
entities defined by bounding "polygons", non-rectangular
entity boundaries shaped to accommodate the entity.
Polygonal bounds are useful in certain situations, such as
overhanging characters in variable pitch print. Their shapes
cannot currently be edited.

Entity image quality, boundaries, appropriateness of text,
etc. are only as accurate as the producer of the DAFS file.
For example, a Word recognized by OCR may contain
misrecognized Glyphs. Illuminator has been designed to make
quick and easy corrections.

A Glyph may be labeled with a string, a long, or another
data type if desired. Even in an ordinary OCR application,
it may be necessary to label Glyphs such as ligatures (`ff',
`ffl') with strings. Under DAFS, a Glyph entity's designated
content will usually be a single character, or a set of
possible characters from which just one member will
ultimately represent the Glyph. These are called multi-rec
Glyphs (since they are often generated by OCR); their
contents are termed possibility sets. For example, the Glyph
`n' might be recognized as one of the following set: {n, ri,
ii}. If confidence values have been assigned, the list will
be in order of decreasing confidence. The possibility set
allows DAFS to represent uncertainty in text content,
preserving information for use by multiple recognition
techniques.

DAFS provides a second way to represent alternative text
contents: the OR child-type. Entities typically have AND-
type child entities, whose contents are intended to be
displayed in sequence, such as a Word and its component AND-
type Glyph children. In contrast, an entity with OR-type
children can be represented by only one of its children at a
time. The OR child-type might be used, for example, where
there are alternative spellings for a Word, such as `honor'
and `honour'. The Word would have two OR children, each with
one of the spellings as its content. Only one of the
spellings would be on display at any given time. The OR
child-type is available for use with all entity types.

In general with Illuminator, an object cannot be directly
affected if it does not appear on the screen. For example,
bounding boxes do not appear in Text Mode, and Text Mode
provides no mechanism for modifying them. A bounding box can
be indirectly changed, though, as when deleting a block of
text deletes the entities it contained.

                             GUI
                              
Illuminator's GUI runs in the OSF/Motif environment. It is
designed to display the input document as a scanned image,
as a combination of image and text, or as text alone.
Documents may be viewed and altered in any of its Modes,
which are described in detail below. Each Mode provides a
different presentation of document data, and different ways
to edit entities and their bounding boxes and content. The
underlying data structure is shared between Modes. When
switching Modes, changes made in one mode show up in the new
one. There is not yet an undo command, but there is a Revert
to Saved option.

                     International Text
                              
Illuminator makes it easy to work with documents in foreign
languages via Unicode, the emerging universal character
encoding standard. Many major foreign language scripts are
currently available. Keyboard Map Displays may be accessed
as mnemonic and typing aids. Unicode characters may also be
entered individually with the Enter Unicode feature.

Illuminator maintains Level-1 conformance to Unicode, which
implies that all non-composed characters can be displayed.
Unicode distinguishes base characters, which always occupy
separate spaces, from non-spacing diacritical marks, which
may share the space of a base character. Some characters are
composed of a base character and one or more non-spacing
elements. Separate characters are defined for many commonly
occurring composites (such as ), and Illuminator displays
these. It can also display some actual composed characters,
mostly in the Arabic script. Others are depicted as a base
character followed by the "non-spacing" marks displayed in
their own spaces. Fonts for many Unicode values, which
Illuminator would otherwise be capable of displaying, may
not be included in this release. Such Unicode values are
represented on screen by a small black triangle.

Japanese

Typing Japanese script is accomplished by a Japanese Input
option rather than the Keyboard feature used for other
languages. Japanese is treated as a special case because it
is extremely difficult to map kanji, an important Japanese
character set, to a Western-style keyboard.

Modern Japanese employs four different character sets. They
are kanji, hiragana, katakana, and romaji. Kanji are
ideographic characters borrowed from Chinese. There are tens
of thousands of kanji, and a literate Japanese must know
many thousands. Each may have several pronunciations,
usually one or two syllables, which are dependent on
context. There are many homonyms. Hiragana and katakana
(together referred to as "kana") are closely related
syllabaries used to represent about 50 sounds. Hiragana is
used to sound out kanji and to decline Japanese; it has a
cursive appearance. Katakana is used to sound out foreign
borrowed words, and is sometimes used for emphasis. Its
appearance is blockier than that of hiragana. Romaji is
simply the Roman character set, used for Western words under
some circumstances.

A Japanese document typically uses all of these diverse
character sets, so any Japanese language input method must
be capable of handling them all. Romaji input is
straightforward with an English language keyboard. Kana
could be mapped to a keyboard as well, though Illuminator
has not yet done so. The most common scheme requires
multiple keystrokes to represent some characters, of which
Illuminator is not yet capable of doing. Kanji poses a very
difficult problem. A number of different schemes have been
used to classify kanji (number of strokes, pronunciation,
position in the standard character set, etc.), each
inspiring an input method. Illuminator uses a third-party
Japanese input system called "Canna", which employs a sound-
classification scheme. (Canna is a contribution to the X-
Windows distribution; it is also sometimes referred to as
Iroha.)

               Text and Image Storage Formats
                              
Illuminator reads and writes Unicode and ASCII/UTF.
ASCII/UTF is a multibyte encoding method for Unicode which
encodes actual ASCII characters as ASCII, and non-ASCII
characters as two- or three-character sequences. It allows
any Unicode file to be saved as ASCII without losing non-
ASCII characters. When reading an ASCII file, Illuminator
interprets all non-ASCII sequences as ASCII-UTF.

Illuminator also reads and writes DAFS-BINARY files, and can
read TIFF and PDA image files, saving them in DAFS-BINARY
format.

Figure 1 to Figure 6 depict a simple example document that
has an image and two lines, which will be used in describing
Illuminator's Modes below.

                              
                              
            Figure 1. Example Document Structure
                              
Figure 1 diagrams the example document, which has a single
Document entity and two Line entities, both with bounding
boxes. In the first Line, each Word and Glyph is defined by
a bounding box. The text, which reads "The cat ran." in the
original, was misrecognized as "Me cat raa" by some OCR
program. The final character in the line was tentatively
recognized as an `a' but the OCR program also considered
`n', the correct value, a possibility. The second Line
contains only the image of a sentence, "The dog chased the
cat.", and has no associated textual content.

                         Image Mode
                              
In Image Mode, the image of a document is displayed with the
bounding boxes and contents of the entities overlying it.
Different entity types can be represented in separate user-
specified colors. All entity types can be displayed, or just
the types selected by the user. Entities may be created and
deleted. A given entity's content can easily be edited, and
its bounding box moved or resized. Those which do not have
bounding boxes cannot be directly represented, but they may
be indirectly affected if a related entity is modified. For
multi-rec Glyphs (see DAFS section), Image Mode displays the
entire content possibility set. Features such as Keyboard,
Enter Unicode, and Japanese Input make it easy to add text
labels in foreign languages. Image Mode is capable of
reading and displaying TIFF and PDA images as well as DAFS
documents.

                              
                              
           Figure 2. Sample Document in Image Mode
                              
                              
Figure 2 shows how the example document might appear in
Image Mode. The image quality is badly degraded, causing OCR
segmentation and recognition errors. The Glyph labeled `M'
needs to be split into two, one for the `T' and one for the
`h'. This could be done by deleting the `Th' entity and
creating two new Glyph entities, drawing bounding boxes
which separate the two characters and relabeling them.

                     Out-of-Context Mode
                              
All the Glyphs of one text value can be displayed together
in this Mode. The user is able to bring up a window of just
the `a's, for example. They will be displayed packed from
left to right and top to bottom, without regard to the
original location on the page. In Figure 3, an `n' has shown
up on the `a' page in Out-of-Context Mode. It is immediately
obvious that it is out of place. After it is relabeled, the
image of the `n' will no longer appear with the `a's. It
will instead be present on the `n's page. A multi-rec Glyph
is classified as the first character in its possibility set.

                              
                              
            Figure 3. Sample Document in OOC Mode
                              
                              
This Mode is a completely different way of verifying a
document's accuracy, and provides a powerful tool for
creating error-free documents and training sets. Simply
proof-reading a document does not produce an error-free
document. The very abilities which make people good at
reading degraded documents, and capable of filling in a
pattern from a few clues, also cause them to miss many kinds
of typographical errors. This is especially true for
documents which were recognized by OCR, where errors tend to
look like the correct result. If, for example, the word
`bold' were misrecognized as `hold', a proof-reader might
well read `hold' as the similar-looking `bold' just because
the surrounding context would suggest it. A spell-check
program would also miss this error. Because Out-of-Context
Mode brings all entities of the same type into one window,
all the contextual information is stripped out of the
document. The user cannot read the document and therefore
cannot read right over errors. Errors can usually be spotted
with a quick glance at the page.

                        Flagged Mode
                              
The look of Flagged Mode is somewhat similar to Out-of-
Context Mode. Flagged Mode is for correcting entities which
have been flagged by OCR as questionable or unknown. Some
OCR programs may flag single glyphs; some may only flag
entire words. The flagged entities are grouped by entity
type rather than by content, so that the user may view a
page of flagged Glyphs, a page of flagged Words, flagged
Lines, etc. The OCR-derived content of each entity is
displayed along with its image, and is easily edited. A
document in Flagged Mode appears in Figure 4.



          Figure 4. Sample Document in Flagged Mode
                              
                              
                              
                              
                              
                       Hierarchy Mode
                              
Hierarchy Mode displays information about the entity
hierarchy in compact form, listing the entities in preorder.
The example document's entity hierarchy would be listed as
shown below in Figure 5. This Mode offers a quick, concise
overview of the document's structure as a whole. Entities
are drawn in user-specified colors as part of a tree-like
structure. The user can rearrange the hierarchy by selecting
and dragging entities to new locations.

                              
                              
         Figure 5. Sample Document in Hierarchy Mode
                              
The triangle preceding some of the entities is used to
indicate those which have child entities below them. The
entity structure can be collapsed, if desired, at these
points (see Working in Hierarchy Mode).

                          Text Mode
                              
Text Mode is Illuminator's best way to enter original (new)
text, and the only way to edit ASCII or Unicode text files.
It displays the text contents of all entities which have
text, displaying the combined contents of contiguous
entities as readable words and lines of text. If an entity
structure or other formatting information is present, the
text can be laid out in lines, paragraphs, fields, columns,
and multiple pages. If the document does not contain
formatting information, the text is run together on a single
line.

Figure 6 shows how the example document would appear in Text
Mode. The second sentence does not appear here, because the
Line entity has not been labeled with any text.

Any entity, including a Glyph, can have an ASCII string or a
Unicode string for text value. Multi-rec Glyphs and other
entities with a possibility set for content are represented
by the character with highest confidence. Selected text is
shown in reverse video, and text from flagged entities is
shaded. The images of selected text entities can be called
up in Text Mode by using the View Selection feature, which
will show the entity's image and surrounding context.

                              
                              
           Figure 6. Sample Document in Text Mode
                              
The basic operations of Text Mode are to select an insertion
point or a run of text, and to insert or delete text. Such
text is sometimes inserted into an existing entity, as when
an OCR error is corrected, and sometimes inserted into a
newly created entity, as when a new line of text is typed in
the middle of a document. This entity assignment function is
only approximate, so that Text Mode is not the tool of
choice for creating ground truth documents or training data.
It is appropriate for making corrections to OCR output when
the desired result is the text alone and not correct
bounding boxes.

Text Mode is a little different than the other Modes in how
it displays information. The order in which text is
displayed here is precisely the preorder traversal of the
entity hierarchy. The entity images play no role, so that
entities without text content do not appear at all in Text
Mode. If both an entity and its descendent entities have
text content, the contents of the descendants are ignored.
Otherwise, the contents of descendants appear together
sequentially by hierarchy position. While Text Mode cannot
display entities without content, it can display text which
is not associated with any image. Therefore, this Mode uses
a different model of selection than the other Modes.

                Table 1.  Mode Summary Table
                              
  Mode      Organized      Information       Operations
                            Displayed
 Image      by image     Entity images    Move and resize
            position     and bounding     bounding boxes;
                         boxes; entity    create new
                         content          entities; edit
                         including Glyph  entity content.
                         possibility
                         sets
  OOC       by Glyph     Glyph images     Correct
              label      and bounding     mislabeled
            (highest     boxes            Glyphs. Remove
           confidence                     degraded Glyphs
             value)                       from a training
                                          set.
  Text    by read order  Entity content   Insert/delete
           (preorder)    (highest         text.
                         confidence)
Flagged     by entity    Entity images,   Correct OCR
           type (only    bounding boxes   flags.
             flagged     and content.
            entities)
Hierarch   by preorder   Entity           Collapse,
   y                     hierarchy,       uncollapse and
                         entity types,    rearrange
                         bounding boxes,  portions of the
                         child types,     hierarchy.
                         text type and
                         content.
                              
                              
                              
                     Future Enhancements
                              
General

One or more language scripts from Chinese, Hebrew, and
Korean are planned for a future release. Illuminator will be
capable of SGML output, based on a DTD that captures the
DAFS data structure.

Image Mode

In Image Mode, an image de-skewing tool may be added. A
limited ability to edit image pixels may be developed, and
it may be possible to perform measurements on areas of
image.

Text Mode

Cut, Copy and Paste features will be added. Text Mode will
be capable of rendering text from different languages
according to their customary layout methods (for example,
left-to-right, and top-to-bottom).

Out-of-Context Mode

It will be possible to save pages of correctly labeled
Glyphs of one type into a separate Glyph file, useful in
training OCR systems.

View Selection

Editing names and values of properties from this dialog will
be possible in future.

                              
                              
                     _Using Illuminator
                              
Illuminator is an X-windows program running under OSF/Motif.
It is currently available on SPARC Workstations and Linux
PC's in binary format. Source code will be released with
version 1.0. Installation instructions are found in the
release notes accompanying the executable Illuminator.
Consult the OSF/Motif User's Guide general information on
using Motif and X-windows.

Other Documents

The use of Illuminator is described by this User's Manual
(called illum.ps and illum.txt in the online versions).
Other accompanying documents are The DAFS Library
Programmer's Guide and Reference (dafslib.doc, dafslib.ps,
and dafslib.txt), which describes a library of C++ routines
for manipulating DAFS files.

                         Starting Up
                              
To begin using Illuminator, type "illum" at the shell
prompt. Illuminator's opening screen will come up. You will
see one window, a Main Menu of options and an Action Bar for
changing Modes and adjusting window magnification. By
default, Illuminator opens in Text Mode with an empty file
named "Untitled." You can also supply a file name on the
command line (see Command Line Options). Illuminator
determines the type of file you are loading (DAFS file,
image file, or text file) and displays it appropriately.

                    Command Line Options
                              
Illuminator's command line options are described below.
Illuminator will come up in a particular Mode as directed,
even if it makes no sense for the current file (for example,
-o will bring up a file with no Glyphs in Out-of-Context
Mode). Square brackets ([]) indicate optional arguments. Any
of the optional arguments represented by lower-case letters
(except -h) may be listed together on the command line,
along with just one optional argument represented by upper-
case letters. If "-h" is present, the command line options
list will be printed, and other arguments will be ignored.
If none of the optional arguments are included, Illuminator
comes up in Text Mode, displaying an empty new file.

                Table 2. Command Line Options
                              
illum  [-c <config file>]      Reads and displays
                               Illuminator's configuration
                               file (usually .illum)
       [-l <canna_lib_path>]   Instructs Illuminator where
                               to find the Japanese (Canna)
                               library. This library must
                               be available to use the
                               Japanese Input feature.
       [-j <canna_host>]       Instructs Illuminator where
                               to find the server for the
                               Japanese Input feature. Must
                               be available to use the
                               feature.
       [-h]                    Prints this list of current
                               command line options and
                               exits. Does not start up
                               Illuminator.
       [-H]                    Comes up in Hierarchy Mode.
       [-M]                    Comes up in Flagged Mode
       [-O]                    Comes up in Out of Context
                               Mode
       [-T]                    Comes up in Text Mode.
       [filename]              If a filename is supplied,
                               Illuminator will read and
                               display the file. Otherwise,
                               it displays a new file
                               called "untitled".
                               
                              
                              
                              
                              
              Main Menu and Action Bar Options
                              
Selecting a Main Menu or Action Bar option will pull down a
menu. Selections may be made with the mouse pointer by
pressing the left button or by using the accelerator keys
which are displayed on the menu. In the following sections,
we have listed the accelerator keys to the right of each
menu function, in a manner similar to the way it appears in
the actual menus of the running program. Options are grayed
out and inactive when they are inappropriate.

Some items are listed in the menus with a trailing ellipsis
(...). These options bring up dialog boxes. Some of these
are specific to Illuminator, and their use is described in
this manual. Some of these options bring up OSF/Motif dialog
boxes, such as the FileBox dialog. See your OSF/Motif User's
Guide for information on these.

The Main Menu has the following items:

Main Menu

* File
* Edit
* Display
* Preferences
* Page
* Entity

The Action Bar has the following components:

Action Bar

* File Status Label
* Mag Option Menu
* Modes Option Menu
* Keyboard Option Menu
                        The Main Menu
                              
                        The File Menu
                              
File
* New
* Open ...          ctrl-o
* Save              ctrl-s
* Save As ...       ctrl-a
* Revert to Saved
* Next File         ctrl-n
* Previous File     ctrl-p
* Page Up           PgUp
* Page Down         PgDn
* Page Right        ctrl-PgUp
* Page Left ctrl-PgDn
* Quit  ctrl-q




New

New opens a new file in Text Mode for typing and editing.

Open ...

Selecting Open ... brings up a dialog entitled Load Document
for loading a file (see Figure 7). This is a standard Motif
File Dialog. After selecting a file, you will be prompted to
save any current file if there are unsaved changes.
Illuminator is designed to open DAFS-BINARY files, ASCII/UTF
text files, Unicode text files written by Illuminator, and
TIFF and PDA image files.

                              
                              
               Figure 7. Load Document Dialog
                              
Save

Save saves the changes to the document. Saves DAFS-BINARY
files as DAFS-BINARY, ASCII/UTF text files as ASCII/UTF, and
Unicode text files as Unicode. Provides warning if the file
could not be saved. Cannot save TIFF and PDA files in TIFF
or PDA format. These can be saved as DAFS files by using
Save As.

                              
                              
       Figure 8. Save Document Dialog from Save As...
                              
Save As ...

This brings up the Save Document dialog box, which allows
you to enter a new filename for the document, and to specify
whether the file will be saved as DAFSB (DAFS-BINARY),
ASCII, or Unicode. It is a standard Motif File Dialog with
an added option menu to select the output format. See Figure
8. If a file already exists under the new name, you will be
asked for confirmation. The document will be saved to the
new filename and file type as directed, and the later use of
Save will save to the new filename and file type. Provides
warning if the file could not be saved.

If a DAFS file is saved as Unicode or ASCII/UTF, the image
and other non-text information associated with the document
will not be saved. The new text file will contain only the
information visible in Text Mode.

A DAFS file is recognized by its content and not by its
suffix, and ASCII and Unicode text files are also recognized
automatically. Any suffix may be used when saving DAFS or
text files. It is customary (though not necessary) to begin
the suffix of DAFS files with a `d' (`.d?').

Revert to Saved

Reloads the current file, restoring the file to its state at
the last save, or when it was loaded if it hasn't been
saved. You will be asked to confirm that you wish to abandon
any changes you have made.

Next File

Next File closes the current file, then brings up the file
with the same extension which alphanumerically follows it.
You will be prompted to save the current file if there are
unsaved changes. If the current file is the last one in the
directory, it remains open, and a warning dialog will inform
you that this is the last file.

Previous File

Previous File closes the current file, then brings up the
file with the same extension which alphanumerically precedes
it. You will be prompted to save the current file if there
are unsaved changes. If the current file is the first one in
the directory, it remains open, and a warning dialog will
inform you that this is the first file.

Page Up

 Whenever scroll bars are present, moves up one screen in
the current file. If this is the top of the page, nothing
happens. Works in all Modes except Text Mode.

Page Down

Whenever scroll bars are present, moves down one screen in
the current file. If this is the bottom of the page, nothing
happens. Works in all Modes except Text Mode.

Page Right

Whenever scroll bars are present, moves right one screen in
the current file. If this is the right edge of the page,
nothing happens. Works in all Modes except Text Mode.

Page Left

Whenever scroll bars are present, moves left one screen in
the current file. If this is the left edge of the page,
nothing happens. Works in all Modes except Text Mode.

Quit

Quits Illuminator. Prompts to save first if there are any
unsaved changes.

                              
                              
                        The Edit Menu
                              
Edit
* Delete            Delete
* Kill Contents     ctrl-u
* Previous          Left
* Next              Right
* Up                Up
* Down              Down
* Find Next Flag    ctrl-g
* New Entity        ctrl-e
* Enter Unicode ... ctrl-F9
* Japanese Input ...ctrl-F10

These options all pertain to viewing or editing entities.
The selection of entities and text in each of the Modes is
discussed in detail in later sections entitled "Working in
Image Mode", "Working in Text Mode", etc. In brief, entities
are selected using the left mouse button; multiple entities
may be selected by depressing the shift key during mouse
use. Note that the Text on Image feature must be turned on
in order to edit entity contents in Image Mode.

Delete

Enabled when one or more entities are selected. In Image
Mode, Out-of-Context Mode, and Flagged Mode, deletes the
selected entities and any child entities associated with
them. Deleting an entity in no way affects the area of the
image it referenced.

If you have begun editing the content of an entity in Image
Mode or Flagged Mode, Delete works as a backspace rather
than deleting entities. In this case, it deletes the
characters you typed, most recent character first. You can
tell whether you are editing an entity because the text will
appear with a black border around it. If the last character
of an entity's content has been deleted during editing,
delete does nothing.

In Text Mode, deletes selected text, or the character to the
left of the cursor. Entities whose entire contents have been
deleted this way (such as Words or Lines) are also deleted.

Kill Contents

This feature is not enabled in Text Mode. Deletes the
textual content of the selected flagged entity. Does not
delete the entity itself. It is equivalent of using setting
the contents of an entity to "No Contents" using the Entity
Contents Editor of the View Selection.

Note: Inside the Input Japanese dialog, the Kill Contents'
accelerator key (ctrl-u) has an alternate use: it clears the
entry buffer.

Previous

Previous refers to the left arrow key. In Image Mode, Out-of-
Context Mode and Flagged Mode, selects the entity preceding
the currently selected entity. If this is the first entity,
nothing happens. If nothing is selected, moves selection to
the last entity. If multiple entities are currently
selected, moves to the entity preceding the one which was
selected first chronologically.

For large files, pages of entities in Out-of-Context Mode
and Flagged Mode may be organized into subpages. In these
cases, if the current entity is the first one on this
subpage, Previous moves selection to the bottom of the
preceding subpage.

In Text Mode, moves the cursor back one character. Works
only after the insertion point has been chosen. If any text
has been selected, moves the cursor in front of the selected
region.

In Hierarchy Mode, moves selection up to the previous
sibling, if there is one.

Next

Next refers to the right arrow key. In Image Mode, Out-of-
Context Mode and Flagged Mode, selects the entity following
the currently selected entity. If this is the first entity,
nothing happens. If nothing is selected, moves selection to
the last entity. If multiple entities are currently
selected, moves to the entity following the one which was
selected last chronologically.

For large files, pages of entities in Out-of-Context Mode
and Flagged Mode may be organized into subpages. In these
cases, if the current entity is the last one on this
subpage, Next moves selection to the top of the following
subpage.

In Text Mode, moves the cursor ahead one character. Works
only after the insertion point has been chosen. If any text
has been selected, moves the cursor after the selected
region.

In Hierarchy Mode, moves selection down to the next
sibliing, if there is one.

Up

Up refers to the up arrow key. In Flagged Mode and Out-of-
Context Mode, moves selection to the entity above with the
same horizontal position, or as near as possible. If this is
the first row of entities, nothing happens. If multiple
entities are currently selected, moves selection to the
entity above the one which was selected first
chronologically.

For large files, pages of entities in Out-of-Context Mode
and Flagged Mode may be organized into subpages. In these
cases, if the current entity is at the top of one of the
columns on this subpage, Up moves selection to the bottom of
the same column on the preceding subpage.

In Text Mode, moves the cursor up one line to the same
horizontal position, or as near as possible. Works only
after the insertion point has been chosen. If any text has
been selected, moves the cursor up one line to the
horizontal position above the left edge of the first line of
the selected region.

In Hierarchy Mode, moves selection up one line. Skips
children of collapsed entities.

Has no effect in Image Mode.

Down

Down refers to the down arrow key. In Flagged Mode and Out-
of-Context Mode, moves selection to the entity below with
the same horizontal position, or as near as possible. If
this is the last row of entities, nothing happens. If
multiple entities are currently selected, moves selection to
the entity below the one which was selected last
chronologically.

For large files, pages of entities in Out-of-Context Mode
and Flagged Mode may be organized into subpages. In these
cases, if the current entity is at the bottom of one of the
columns on this subpage, Down moves selection to the top of
the same column on the following subpage.

In Text Mode, moves the cursor down one line to the same
horizontal position, or as near as possible. Works only
after the insertion point has been chosen. If any text has
been selected, moves the cursor down one line to the
horizontal position below the right edge of the last line of
the selected region.

In Hierarchy Mode, moves selection down one line. Skips
children of collapsed entities.

Has no effect in Image Mode.

Find Next Flag

In Flagged Mode, clears the flag (if any) of the currently
selected entity, and selects the next flagged entity,
wrapping if necessary, until there are none left. If
multiple entities are currently selected, moves to the
flagged entity following the entity which was selected last
chronologically.

Has no effect in any other Mode.

New Entity

New Entity creates a child entity for the currently selected
entity. The new entity has the same bounding box as its
parent, and the new entity's type is called "Type". Both the
box and the type may be edited from the View Selection
dialog.

This feature is most useful from Hierarchy Mode with an open
View Selection dialog, where it may be used to build or
rearrange the entity hierarchy.

Enter Unicode ...

Brings up the Enter Unicode dialog (Figure 9). Unicode
characters are designated by `U+nnnn', where n represents a
four-digit hexadecimal number. For example, `U+005a' encodes
the character `Z'. To type a Unicode character, you must
know the Unicode value representing the character. Enter the
four digit "nnnn" portion in the Unicode Value field,
following "U+". You may edit your entry with Delete and
Backspace, and with the mouse, which chooses the cursor
insertion point and selects strings of characters for
typeover. The Unicode character you have entered will appear
beneath the word "Glyph" in the dialog. OK or Apply will
type the character on the document (you must first have
selected a cursor insertion point in Text Mode, or one or
more entities in the other Modes). Unicode characters which
Illuminator is not capable of displaying will appear as a
small filled-in triangle.

                              
                              
               Figure 9. Enter Unicode Dialog
                              
OK                  Types the selected Unicode character
                    on the document and closes the dialog.
Apply               Types the selected Unicode character
                    on the document, leaving the dialog
                    open.
Close               Closes the dialog. Does not type
                    anything.
                    
Local Accelerator          Effect
Key
Enter               Types selected character (same as
                    Apply).
                              
                              
                     Japanese Input ...
                              
     Brings up the Japanese Input dialog (Figure 10). To
 accomplish this, Illuminator employs an input system called
 Canna, which uses Japanese pronunciation to help you select
  the character you want to type. To use this feature, you
 must be able to pronounce the character. The pronunciation
    is entered phonetically in Roman characters (romaji).
Illuminator supplies a list of possible characters which are
  homonyms for the sounds which have been entered, allowing
  you to select and type the desired character. The list of
 choices will include kanji, katakana, hiragana, and romaji
                         characters.
                              
 The Japanese Input dialog provides an Input field for entry
  of phonetic Japanese, a Choices field displaying possible
  Japanese characters, and a Mode indicator. It offers four
Entry Modes and two character Choice Modes. The current Mode
  is indicated by the characters under the word Mode in the
                         dialog box.
                              
             Typing Japanese in the Input Field
                              
 The Japanese words you type phonetically can be represented
  many ways. Each has katakana and hiragana forms, and most
 are associated with more than one kanji form because of the
frequent occurrence of kanji homonyms. Roman characters also
 have their place, since some words are customarily spelled
 out in romaji. The four Entry Modes allow you to select the
 character set used in "translating" your phonetic entries.
      Note that as long as you remain in an Entry Mode,
 hiragana/kanji entries will be displayed in hiragana. Kanji
             is accessible only in Choice Mode.
                              
    In the kana modes, the Choices field will display the
  corresponding hiragana or katakana on the fly as you type
    the syllables of the word. Spurious or untranslatable
  sequences of Roman characters will appear unchanged among
  the Japanese characters. In the ASCII modes, the Choices
  field contains only the Roman characters you have typed.
                              
             Table 3.  Entry Modes for Japanese
                              
Entry      Symbo  Activat  Description
Mode       l      ion
hiragana/         Tab      Phonetics from the entry field
kanji                      are rendered in hiragana in the
                           Choices field, translated into
                           kanji later in Choice Mode (see
                           Table 4).
katakana          Tab      Phonetics are rendered in
                           katakana.
full-       [ a   Tab      Phonetics are rendered in ASCII.
width      ]               Each ASCII character is as wide
ASCII                      as a Japanese character.
ASCII        [a]  Tab      Phonetics are rendered in ASCII.
                           Each ASCII character is half as
                           wide as a Japanese character.
                              
Begin by using the Tab key to select the Entry Mode you want
   (see Table 3). You may cycle through the Entry Modes by
repeatedly pressing Tab. Now type one or more Japanese words
 in the Input field. Use a phonetic form of the word, using
Roman characters. For example, the word for `cherry blossom'
might be typed as `sakura'; the word for `Japanese language'
                        as `nihongo'.
                              
  The Input Field can only be edited by clearing the buffer
   and starting again. Use ctrl-u to erase and start over.
                              
     You may change the Entry Mode (via Tab) as you go.
  Characters present in the Choices field before the change
 will not be affected, but subsequent phonetic entries will
be rendered in accordance with the new Mode. In this way you
  can enter a mix of hiragana, katakana and ASCII into the
                       Choices buffer.
                              
 To type your entry onto the document, tap the Space Bar to
   leave the Entry Modes and move on to the Choice Modes,
   discussed in "Choosing a Character". Once you have left
 Entry Mode, no additional entries can be made. Clearing the
      buffers with ctrl-u puts you back in Entry Mode.
                              
                              
                              
              Figure 10. Japanese Input Dialog
                              
Choosing a Character

Typing Space moves you from Entry Mode to kanji mode, the
first Choice Mode. The Entry field is no longer active. Any
hiragana in the Choices buffer will be changed into the
kanji possibility which Illuminator considers most likely.
If there is no hiragana present in the Choices field, Choice
Mode may not have anything new to offer. It is primarily
used for selecting the appropriate kanji.

Obtaining just the kanji you want can take a couple of
iterations. If the "most likely" kanji offering is not what
you want, you may view additional kanji by continuing to
type Space. You are now in list mode, the second of the
Choice Modes. The Space Bar cycles through all current kanji
possibilities, continuing with appropriate kana
representations.

If the desired kanji does not appear among the Choices, it
is sometimes helpful to try phonetic spellings of other
pronunciations for that kanji, or to include more words in
the Input field for context.

             Table 4.  Choice Modes for Japanese
                              
Choice     Symbo  Activation   Description
Mode       l
kanji             First use    Hiragana entries in the
mode              of the       Choices field are rendered
                  Space Bar,   in the most likely kanji.
                  from an      Non-hiragana characters may
                  Entry Mode.  or may not have any
                               alternate form.
list mode         Subsequent   Space bar cycles through
                  uses of the  alternate Japanese
                  Space Bar.   representations, including a
                               list of possible kanji,
                               hiragana, and katakana.
                              
            Entering a Character on the Document
                              
  Once the character you want appears in the Choices field,
       you are ready to make an entry on the document.
                              
  If you are in Image Mode or Flagged Mode, you may enter a
string of any length as the content of an entity. Select the
 entity you wish to edit, and use Enter, OK or Apply to type
your Japanese entry. As long as the entity is selected, each
 new character or string will be appended to what was typed
previously. Text on Image must be turned on in order to edit
                   entities in Image Mode.
                              
    In Text Mode, position the cursor where you want the
 Japanese text to appear, or select the text you want it to
   replace. Enter each character or by using Enter, OK or
                           Apply.
                              
 In Out-of-Context Mode you may label a single Glyph with a
    single Japanese character. If there is more than one
 character in the buffer, the Glyph will be labeled with the
   last character. Select the Glyph to be labeled with the
             mouse, then use Enter, OK or Apply.
                              
  The local accelerator keys listed below work as described
             only in the Japanese Input dialog.
                              
OK                  Types the Japanese character in the
                    Choices field onto the document and
                    closes the dialog. The entity or
                    cursor insertion point must have been
                    chosen beforehand.
Apply               Types the Japanese character in the
                    Choices field onto the document,
                    leaving the dialog open. The entity or
                    cursor insertion point must have been
                    chosen beforehand.
Close               Closes the dialog. Does not type
                    anything.
                    
Local Accelerator          Effect
       Key
Tab                 Selects the Entry Mode.
Space               Displays the next choice (primarily
                    useful in kanji and list modes).
ctrl-u              Clears the Input and Choices buffers.
                    This function replaces Delete and
                    Backspace, which do nothing in this
                    dialog.
Enter               Types the Japanese character(s) in the
                    Choices field onto the document,
                    leaving the dialog open. The entity or
                    cursor insertion point must have been
                    chosen beforehand. (Same as Apply.)
ctrl-c              Closes the dialog.
                              
                              
                              
                              
                      The Display Menu
                              
Display

* Keyboard...
* View Selection... ctrl-shift-v

Keyboard...

Brings up the current Keyboard Map Display. (The keyboard
layout is set via the Keyboard Menu, described below.
American, Arabic, British, French, German, Greek, Italian,
MICR, Norwegian, Russian, Spanish, and Swedish layouts are
available at this time. MICR (Magnetic Ink Character
Recognition) symbols are used on bank checks and in similar
applications. Figure 11 shows one page of the Keyboard Map
Display for Russian.

           Figure 11. Russian Keyboard Map Display
                              
The Keyboard Map Display may be used as a mnemonic aid in
typing foreign scripts from the keyboard. Alternatively, the
display itself may be used to type a given character by
placing the mouse pointer on the character and clicking the
left button. The Shift key toggles whether shifted or non-
shifted characters are displayed. The Meta key toggles the
display of additional characters for larger character sets,
such as Russian. Shift-Meta toggles the display of shifted
and nonshifted characters in this second set. (The Meta key
may be called some other symbol or name, such as Alt, on
many keyboards. The Keyboard Map Displays are modeled after
the keyboard of a Sun SPARC work station, which has both Alt
and Meta keys.)

Modifier Key        Keyboard Map Display
none                Lower case characters, set 1
Shift               Upper case characters, set 1
Meta                Lower case characters, set 2 (if any)
Shift-Meta          Upper case characters, set 2 (if any)


View Selection...

Brings up an informational dialog about a single selected
entity (Figure 12). Displays a pop-up image of the entity
with surrounding context. The dialog describes the type,
contents, confidence values, properties, bounding box
location, and image size in pixels of the entity. In every
Mode except Text Mode, the user may change the entity type,
data type, bounding box dimension, and in most cases edit
text content. In Text Mode, most of this information may be
viewed, but cannot be changed.

View Selection also lets you enter multi-line ASCII or
Unicode text for an entity, something which can't be done
from any of the Modes. To do so, set the entity content to
String, and you can use a standard Motif Text widget to
enter text.

The Image Field

The image appears across the top portion of the dialog. If a
single entity is selected, the entity's image and
surrounding context will be shown. If multiple entities are
selected, the View Selection dialog becomes inactive, except
in Text Mode. The selected entity is shown enclosed by a
bounding box except in Text Mode, where it will only be
surrounded by a bounding box if the selected area
corresponds precisely to the content of a single entity. If
the selected region in Text Mode overlaps entities, the
smallest entity which contains the selected area (which may
be the Document entity itself) will be displayed, or as much
of it as will fit in the image field. If nothing is selected
but the insertion point, the image of the entity containing
the insertion point will appear, and will be demarked by the
left and bottom sides of the bounding box, which is drawn in
red on a color screen.

                              
                              
                  Figure 12. View Selection
                              
The Type Button

The type button shows the selected entity's type, followed
by the ellipsis symbol indication the presence of a dialog.
To change the entity type, click on the type button to bring
up the Choose Type dialog ; this is a standard Motif List
Dialog. Select one of the entity type names listed, or
specify your own new type name by entering it in the
selection field. To complete the change, click on OK. The
new entity type will now appear on the type button.

        OK          Changes the entity type to the
                    one specified in the dialog
                    selection field and closes the
                    dialog.
      Cancel        Closes the dialog without
                    changing the entity type.

The Property Field

This lists the names and values of properties associated
with the selected entity. At this time, it is not possible
to make changes to properties from View Selection.

The Contents Field

Just as there are different types of entities, the textual
contents associated with them have different data types.
Some examples of content data types are Long (integer),
String, Unicode String, and Choices (possibility set). For
example, Glyphs typically have possibility sets with
confidences for each member of the set; Spaces have integer
contents, the integer value indicating the number of spaces
present. The View Selection dialog can be used to edit
entity contents and to change their data types.

      Table 5.  Table of Data Types for Entity Contents
                              
Data Type                    Operations
No Content                Change type; Edit content
Posset                    Change type; Edit content
String                    Change type; Edit content
User Data                 Cannot change the type to
                          User Data, or edit this type
                          of content at this time.
Confidence                This feature available in the
                          future.
Long                      Change type; Edit content
Unicode String            Change type; edit content.
                          You cannont change the
                          keyboard or use Japanese
                          input from this dialog.
                              
The entity's contents, if any, are displayed in the contents
field. The contents field appears below the contents button,
  which will be labeled with the data type of the entity's
       contents. See Table 5 for standard data types.
                              
 To change the contents or data type of the entity, click on
  the contents button to bring up the Edit Contents dialog.
 The Edit Contents dialog will appear differently, depending
  on the data type to be edited. In each case, contents are
 edited by selecting a run of text or an insertion point in
              the dialog and typing in changes.
                              
   The data type is listed on the button next to the word
"Type" at the bottom of the dialog. To change the data type,
  click on the button and select a new data type (see Data
                     Type table above).
                              
           To complete your changes, click on OK.
                              
        OK          Changes the entity contents or
                    data type as specified in the
                    dialog selection field and closes
                    the dialog.
      Cancel        Closes the dialog without making
                    changes.
                              
                   The Bounding Box Field
                              
   Bounding box (labeled Box) and pitch bounding box (PBB)
 measurements, if any, are in pixels; x and y determine the
    upper left corner, and w and h are height and width.
                              
                       The Box Button
                              
Click on the Box button to bring up a dialog for editing the
  bounding box dimensions. Type in a new value, then select
   OK. The edited box can now be seen in the Image Field.
                              
                              
                              
                              
                              
                    The Preferences Menu
                              
Preferences
* Display in Image Mode...
* Type Properties...
* Text On Image     ctrl-k
  


Display in Image Mode...

Brings up the Display in Image Mode dialog (see Figure 13).
The purpose of this feature is to reduce view screen clutter
in Image Mode by displaying only those overlays of interest
to the user.

                              
                              
           Figure 13. Display in Image Mode Dialog
                              
The Display in Image Mode dialog offers three sets of
options. The first set of options lets the user select which
types of entities will be displayed in Image Mode. The
second set allows selected display of the contents of Glyph
entities, according to their certainty. The third allows
flagged entities to be displayed or hidden. A row of buttons
along the bottom of the dialog box activate the user's
selections and close the dialog.

         Table 6.  Features of Display in Image Mode
                              
Document            Document bounding box to be displayed.
Paragraph           Paragraph bounding boxes to be
                    displayed.
Line                Line bounding boxes to be displayed.
Word                Word bounding boxes to be displayed.
Glyph               Glyph bounding boxes to be displayed.
other               All other types of displayable
                    entities or properties to be displayed
                    (e.g., Line pitch and phase
                    properties), if any.
     low rec        Glyphs with low confidence values to
                    be displayed.
     non rec        Unrecognized Glyphs to be displayed.
     multi rec      Glyphs whose content is a set of
                    possible characters to be displayed.
     O0             Glyphs whose content is either O or 0
                    to be displayed.
other               All Glyphs not in the preceding
                    categories to be displayed.
flagged             Flagged entities to be displayed.
not flagged         Entities not flagged to be displayed.
                              
An entity category may be selected or deselected by clicking
   the mouse on the button in front of it. The buttons of
 selected items have an altered appearance. Set each item as
   desired, then activate the selection with OK or Apply.
                              
OK                  Applies your selections and closes the
                    dialog.
Apply               Applies your selections; dialog
                    remains open.
Revert              Restores buttons to their state at the
                    last OK or Apply.
Default             Restores buttons to the default state.
                    Default setting is all buttons on.
Close               Closes the dialog.
                              
                     Type Properties...
                              
  This feature allows you to choose how different types of
entities will be displayed. You can select the colors entity
  types are drawn in, put selected entity types on separate
  pages in text mode, or treat them as a block for editing
purposes. You may modify one type, or each in turn. The Done
      button enters your changes and closes the dialog.
                              
 Changing Type Properties is a two-step process. First, the
  Select DAFS Entity Type dialog allows you to select which
  type you want to modify. It offers a set of buttons with
 entity type names. The color of each button shows the color
currently assigned to that entity type. Click on the name of
the entity type whose display properties you wish to edit. A
   second dialog will appear with a color table and three
    toggles, discussed below (see Table 7). You may make
   changes, then click on OK to return to the Select DAFS
  Entity Type dialog. Select another entity type to modify.
                When finished, click on Done.
                              
   When you use Type Properties to change the entity type
  display parameters, your preferences are saved (in a file
named ~/.illum) and will be loaded the next time Illuminator
                       is brought up.
                              
           Table 7. The Type Properties... Dialog
                              
           Feature                     Description
Type Color Table              Select a display color for
                              this type of entity
Place on its own page in Text Creates a separate page in
Mode                          Text Mode for each entity of
                              this type. (Default off.)
Edit as separate region in    Allows you to edit each
Text Mode                     entity of this type as a
                              block in Text Mode. If this
                              feature is set for an entity
                              type, you can select text
                              from within the entity
                              currently being edited, but
                              cannot select text beyond its
                              border. (Default off.)
Display in Image Mode         Lets you turn on or off the
                              display of bounding boxes and
                              text content in Image Mode.
                              This can also be accomplished
                              from the Display in Image
                              Mode dialog, discussed on
                              page 36. (Default on.)
                              
                      Type Color Table
                              
   Select a color for the entity type and click OK. Entity
  bounding boxes and text content will now be drawn in this
  color in Image Mode, as will entity markers in Hierarchy
                            Mode.
                              
                   Type Properties Toggles
                              
  Toggle the button by clicking on it with the mouse, then
             click on OK to effect the changes.
                              
There are a couple of caveats in working with these toggles.
 First, "Place on its own page in Text Mode" works for only
  one type of entity at a time. If it is turned on for two
 entity types, only the type higher in the entity hierarchy
 will be affected. Hierarchy Mode can show you the order of
  precedence for your document. Second, "Edit as a separate
region in Text Mode" will only be effective if the region is
smaller than a page of text. For example, if each Line is on
   its own page of text, it makes no sense to try to edit
              Paragraphs as a separate region.
                              
                        Text on Image
                              
   Toggles whether entity textual content is displayed on
screen in Image Mode. The default setting is text display on
(the text is displayed). A toggle button to the left of Text
 on Image changes appearance with state (text display on or
  off). Text on Image affects all entity types, and affects
text only (not bounding boxes). Note that it is not possible
 to edit entity contents in Image Mode when Text on Image is
                            off.
                              
                              
                        The Page Menu
                              
Page
* Select ...        ctrl-x
* Forward           ctrl-f
* Back              ctrl-b

A page is a single image in Image Mode, a page of text in
Text Mode, a set of Glyphs with the same label in Out-of-
Context Mode, and a set of flagged entities of the same type
in Flagged Mode. The page number, Glyph or type is shown in
parentheses, following the filename at the top of the screen
(e.g. "dafsfile.dr(1)").

In Out-of-Context Mode and Flagged Mode, a large page is
organized into a series of subpages. For example, for a page
of `b's in Out-of-Context Mode with five subpages, the third
might be named "hugh.dr(b[3 of 5])".

Text Mode currently displays text on a single page, with
scroll bars if needed. A future release will include
separate pages in this Mode.

Hierarchy Mode shows information on a single scrollable
page.

Select ...

This item brings up the Select dialog, which is a standard
Motif List Dialog. Where there are multiple pages in Image
Mode and Text Mode, Select lets the user select and move
directly to the desired image or text page.

In Out-of-Context Mode and Flagged Mode, it allows the user
to select a particular Glyph or type of flagged entity, and
move to the page containing them. Subpages cannot be
directly accessed by Select, but they can be traversed using
Forward, Back, Previous, Next, Up and Down.

OK                  Close the dialog and brings up the
                    selected page of Glyphs.
Cancel              Closes the dialog without effect.

Forward

In Image Mode or Text Mode, displays the document image or
text page following the currently displayed page.

In Out-of-Context Mode, shows the next subpage, if there is
any. Otherwise, moves to the next page of Glyphs (i.e. the
alphanumerically next character for which examples exist.)

In Flagged Mode, shows the next subpage, if there is any.
Otherwise, moves to the page of flagged entities which are
one level down hierarchically. For example, you would move
from the page of flagged Words to the page of flagged
Glyphs.

In all Modes, if this is the last page, nothing happens.

Back

In Image Mode or Text Mode, displays the document image or
text page preceding the currently displayed page.

In Out-of-Context Mode, shows the preceding subpage, if
there is any. Otherwise, moves to the previous page of
Glyphs (i.e. the alphanumerically preceding character for
which examples exist, and are so labeled).

In Flagged Mode, shows the preceding subpage, if there is
any. Otherwise, moves to the page of flagged entities which
are one level up hierarchically. For example, you would move
from the page of flagged Glyphs to the page of flagged
Words.

In all Modes, if this is the first page, nothing happens.

                              
                       The Entity Menu
                              
Entity
* Center            ctrl-l
* Rotate Image
* Grow Right        ctrl-Right
* Shrink Right      ctrl-Left
* Move Right        shift-Right
* Move Left         shift-Left
* Grow Down         ctrl-Down
* Shrink Down       ctrl-Up
* Move Up           shift-Up
* Move Down         shift-Down

Center

Centers a selected entity in the window, and redraws the
image. Centers the entity as far as possible, without
scrolling beyond the edge of the image. If the selected
entity is already centered, the image is redrawn but nothing
else happens. In Out of Context Mode, changes to the page
and subpage containing the selected entity.

If multiple entities are selected, does nothing.

Rotate Image

Rotates the entire image clockwise 90.



Bounding Box Adjustment Options

The following eight related options are used to move and
resize entity bounding boxes. Each operates only when a
single bounding box has been selected. Each may also be
accomplished using the mouse alone, as described in "Entity
Creation and Adjustment", part of the Special Features
section.

Grow Right       Expands a single selected bounding box by
                 moving the right edge one image pixel to
                 the right.
Shrink Right     Shrinks a single selected bounding box by
                 moving the right edge one image pixel to
                 the left.
Move Right       Moves a single selected bounding box
                 (relative to any underlying image) one
                 image pixel to the right.
Move Left        Moves a single selected bounding box one
                 image pixel to the left.
Grow Down        Expands a single selected bounding box by
                 moving the bottom edge one image pixel
                 down.
Shrink Down      Shrinks a single selected bounding box by
                 moving the bottom edge one image pixel up.
Move Up          Moves a single selected bounding box one
                 image pixel up.
Move Down        Moves a single selected bounding box one
                 image pixel down.
                 
                       The Action Bar
                              
                      File Status Label
                              
This label is located at the far left of the Action Bar. It
shows the symbol `--' if the file is unchanged since it was
loaded or the last time it was saved, or the symbol `**' if
the file has been changed.

              The Mag Menu: Image Magnification
                              
Mag
* 1/8               ctrl-shift-8
* 1/4               ctrl-shift-4
* 1/2 ctrl-shift-2
* 1 ctrl-1
* 2 ctrl-2
* 4 ctrl-4
* 8 ctrl-8

These options change the magnification of the images in the
Image Mode, Out-of-Context Mode or Flagged Mode window. Only
images are affected by Mag; labels and other text do not
change. Illuminator starts with a magnification of 1.
Changes in magnification persist on changing Modes or files.

Reduce the magnification of displayed images with 1/2 and
1/4; increase it with 2 and 4. The window size is not
affected, but the number of subpages (if any) may change in
Out-of-Context Mode and Flagged Mode. This menu is not
functional in Text Mode or Hierarchy Mode.

               The Modes Menu: Changing Views
                              
Modes
* Image             ctrl-i
* OOC               ctrl-c
* Flagged ctrl-m
* Hierarchy ctrl-h
* Text  ctrl-t

This menu allows you to switch between Illuminator's Modes.

              A Word About Moving Between Modes
                              
   When switching Modes, the same entity will generally
   be selected in the new Mode as was in the previous
   Mode. If a single entity is selected in Image Mode,
   Flagged Mode, or Hierarchy Mode, the entity will also
   be selected in the other Modes and the display will be
   centered on the area containing it. This is also often
   true for Text Mode, though this Mode is different
   since selected text may contain just part of an
   entity's content, and entities with no content cannot
   be represented.
   
   Out-of-Context Mode is the exception. Upon switching
   from another Mode to Out-of-Context, however, Out-of-
   Context Mode will continue to show the same page it
   last displayed, in spite of selection changes made in
   the other Modes. If the new selected entity is a
   Glyph, the Center option (Entity Menu) can be used to
   move to the Out-of-Context Mode page and subpage on
   which it is found.
   
   
             The Keyboard Menu: Foreign Scripts
                              
Keyboard
* American
* Arabic
* British
* French
* German
* Greek
* Italian
* MICR
* Norwegian
* Russian
* Spanish
* Swedish

The above language or symbol scripts have keyboard mappings
in this release. Selecting a language specifies the keyboard
mapping you wish to use when typing on the keyboard. As an
aid in typing, the Keyboard Map Display may be called up
from the Display Menu.

                      Special Features
                              
Image Mode

Image Mode allows you to create new entities on screen by
drawing bounding boxes around regions of image. These new
entities can only be created using the mouse; there is no
corresponding menu function. The mouse may also be used in
place of the eight menu items for moving and resizing
bounding boxes (see the Entity Menu).

Creating a New Entity

A new entity may be created under these conditions:  it must
have a parent entity, and it may not contain previously
existing entities. The entity and its parent need not
overlap or adjoin on the image. Currently, the new entity's
type is usually assigned as one step below the parent in the
entity hierarchy (e.g., the new entity is a Glyph if its
parent is a Word). If Illuminator's heuristic can determine
the type, the new entity will be given that type name. If
not, the entity will be given the type name "blob". Type
names can be changed from the View Selection dialog. A newly
created entity is placed in the hierarchy as the last child
of the parent entity. It may be edited just like any other
entity.

To create a new entity, first use the mouse to select the
new entity's parent. Then click and drag the middle mouse
button to draw a rectangle around the area of image you wish
the new entity to contain. Releasing the button locks in the
shape of the new bounding box. To avoid unintended slips
with the mouse, entities below a minimum size cannot be
created.

Moving and Resizing Bounding Boxes

When you select an entity, its bounding box thickens, with a
moving bar across the top and resizing handles at the
corners. To move a bounding box, click and hold the first
mouse button on the moving bar (but not near a corner), and
drag it to its new location. To resize, click and hold a
corner handle and drag to make it the size you want. The
position of the entity may seem to shift slightly at the end
of a move or resize. This happens when the magnification is
greater than 1 and the box must adjust to lie along image
pixel boundaries.

The mouse is most appropriate for making large or rough
changes in bounding box size. To fine-tune, use the Entity
Menu options which offer pixel-by-pixel control.

Out-of-Context Mode

Labeling a Glyph with a String

You can label a Glyph with a string as well as a character.
To do so, select a Glyph (or more than one) and press
Escape. Whatever you type now will appear in the status bar
at the bottom of the window. The left segment of the status
bar shows the Glyph's current content. (It also shows
whether or not there are multiple subpages of Glyphs with
the same content.) The right segment is a text entry field.
Enter the string, the press Return. The Glyph will move to a
separate page for Glyphs labeled with that particular
string, and will no longer appear on the current page. To
abort the string edit without labeling the Glyph, press
Escape again.

          Working with Illuminator's Display Modes
                              
                    Working in Image Mode
                              
Once you have a file up on the screen, you are ready to
explore the capabilities of the Image Mode. See the
preceding section for a complete description of Menu options
and functions.

Moving on the Page

Scroll through the displayed image using the scroll bars.
Scroll bars will appear whenever the current image is too
large for the window. Try adjusting Magnification or
resizing the window to view more of the page or see an area
in greater detail. Magnifications of less than 1 show only a
limited set of the image pixels. Magnification 1 shows all
the pixels at the screen's resolution, and higher
magnifications create enlarged pixels.

You can center a selected entity on the screen with the
Center option of the Entity Menu. If the document has more
than one image, the Page functions move you from image to
image.

Displaying Entities

Since the structure of the entity hierarchy is often complex
and there may be a great many entities in a document, it can
become hard to read the document when all of them are
displayed. Several features from Display Menu are available
to help. You can select the types of entities you wish to
display using Display in Image Mode; though the bounding
boxes and contents of unselected entity types will disappear
from the screen, the image itself will remain unchanged.
Text on Image allows you to turn the display of text on and
off. The display colors of entity types can be customized
with the Type Preferences dialog.

Selecting and Deleting Entities

Select an entity by placing the mouse pointer inside its
bounding box and clicking the left button. The entity's
bounding box will become a bolder than it was, and any
content labels will be displayed in reverse video. To toggle
selection/deselection, press the Shift key and click. Select
several entities by depressing the Shift key while selecting
the bounding boxes. Selecting a new bounding box without
depressing the Shift key deselects any previously selected
ones. Clicking outside of any bounding boxes deselects
everything selected. Using the Edit Menu's Previous, Next,
Up or Down option, selection may be moved to the next entity
in the indicated direction.

Selected entities can be deleted using the Delete option
from the Edit Menu.

Editing Contents of Entities

The text contents of selected entities can be edited by
selecting an entity and typing in new text. Content can be
entered as a single line of arbitrary length in ASCII-UTF or
Unicode. Multiple selected entities all receive the same
content. An entity's type can be changed from View
Selection.

It is possible to give an entity muti-line ASCII text from
View Selection by changing the entity type to String and
typing the lines, ending each with a carriage return. Multi-
line Unicode text will be possible in the future.

Creating a New Entity

Select the entity you wish to designate as the new entity's
parent. Now click and hold the middle mouse button near a
spot you would like to be a corner of the new entity's
bounding box to be. Drag the mouse to draw a bounding box
around the area of image you wish to new entity to contain.
Release the button when the box is ready. The new entity may
be labeled as described in the preceding section.

Moving a Bounding Box

Select a bounding box to move. Click and hold the right
button near the middle of the entity, and drag the mouse to
move the box to its new location. Release the button when
finished moving. You can also click, hold and drag outside
but near the selected bounding box, as long as the mouse
pointer is not near any of its corners. Pixel by pixel
(image pixels) control can be achieved by using the options
Move Right, Move Left, Move Up and Move Down from the Entity
Menu.

Resizing a Bounding Box

Select a bounding box to resize. Click and hold the right
button near a corner of the bounding box, inside it or
outside, and drag the mouse to alter the size and shape of
the box. Release the button when the size and shape are as
desired. Pixel by pixel (image pixels) control can be
achieved by using the options Grow Right, Shrink Right, Grow
Down and Shrink Down from the Entity Menu.

                     Working in OOC Mode
                              
Select a single Glyph in the Image Mode. Now switch to Out-
of-Context Mode using Modes on the Action Bar. Select Center
from the Entity Menu to move the page and subpage of the
selected Glyph. You will see a collection of Glyphs like the
one you selected, and perhaps a few mislabeled interlopers.
These are actual excised pieces of image from the original
document, arranged in a rectangular matrix.

When you invoke Out-of-Context Mode, it comes up on the
first page on which Glyphs are present, in Unicode order.
Glyphs with no content are called nonrecs and appear on a
special page entitled "Nonrec". ("Nonrec" is a commonly used
OCR nickname for "unrecognized Glyph", and OOC Mode is most
useful in OCR-type applications.) Finally, if the document
contains no Glyphs with bounding boxes at all, the screen
will be blank in this Mode.

Moving in OOC Mode

Pull down the Page Menu and select Forward or Back to show
(in ASCII or Unicode order) the next or previous page or
subpage containing Glyphs. Use Select... to move to more
distant pages. If a single Glyph has been selected in one of
the other Modes, you can bring up the page of Glyphs
containing the selected Glyph by using the Center option.

To move on the page, Previous, Next, Up and Down in the Edit
Menu move to the next Glyph in the indicated direction.

Selecting and Relabeling Glyphs

Glyphs are selected in OOC Mode the same way they are in
Image Mode. Try relabeling an entity by selecting it and
typing some other character. It will disappear from the
page. Look for it on the page of Glyphs for the character
you typed (use Select in the Page Menu, or the Center option
from the Entity Menu). Multiple Glyphs requiring the same
label change can be changed simultaneously by selecting them
all and typing the new label. An entity labeled with a set
of alternate possible characters will be classified as the
first character. Relabeling a Glyph which has such a
possibility set replaces the entire set with the new label.
You may also enter a string as a Glyph's content. See
"Relabeling a Glyph with a String", below.

CAUTION:  Out of Context Mode is the only Mode in
Illuminator in which it is possible to edit an entity which
does not appear on the screen. Selected entities in OOC Mode
remain selected even after they've been relabeled and are no
longer on screen, so that keys struck afterwards will
continue to relabel the Glyph. The purpose of this behavior
is to allow the user to recall or correctly label Glyphs
after a keyboard slip.

To deselect all Glyphs in OOC Mode, click on an area of the
window outside the matrix of bounding boxes.

Relabeling a Glyph with a String

You can also give a Glyph a string as content by pressing
Escape, and typing in the status bar at the bottom of the
screen. See Labeling a Glyph with a String, page 47, for
more information.

                   Working in Flagged Mode
                              
Setting Flags for Flagged Mode

It is customary for an OCR program to flag characters or
words on which it thinks it may be in error. Flagged-Mode
recognizes entities as "flagged" if they have an integer-
type Property named Flag set (the value of the integer
doesn't matter). See the Programmer's Guide to the DAFS
Library for information on how to set Properties.

Moving in Flagged Mode

Flagged entities are grouped and displayed by entity type in
Flagged Mode. On entering Flagged Mode, the first entity on
the page in view will already be selected, and its content
ready for correction.

The Find Next Flag option from the Edit Menu will unflag the
current selection (so it disappears from view in Flagged
Mode), and moves selection to the next flagged entity.

The Page Menu moves between pages and/or subpages. Previous,
Next, Up and Down from the Edit Menu move selection to the
next entity in the indicated direction. They do not affect
the state of the flag. You may also use the mouse to select
entities. Center from the Entity Menu centers the selected
entity on the page.

Editing Flagged Entities

The editor in this Mode is similar to the overtype mode of
the Text Mode editor. To correct a selected entity, just
type in the new content. Pressing Return or invoking the
Find Next Flag option from the Edit Menu indicates that the
new contents are now correct, and clears the flag. The
corrected entity disappears from view, and selection moves
automatically to the next entity on the page. If the flag is
already correct, you can skip the editing and just invoke
the Find Next Flag command. Kill Contents from the Edit Menu
deletes the content of the selected entity. Delete from the
Edit Menu deletes the entity, automatically moving selection
to the next entity. In addition, any flagged entity whose
content has been completely erased will be deleted when the
user leaves Flagged Mode.

Flagged entities appear in Text Mode as highlighted (reverse
video) text. Flagged entities do not look any different than
other entities in Image, OOC, or Hierarchy Modes.

                  Working in Hierarchy Mode
                              
Hierarchy Mode provides a useful visual mode for displaying
the structure of the current document. It shows the entity
hierarchy as a tree, with parent/child/sibling relationships
clearly visible. Entity types and up to ten characters of
content are displayed. This minimal display gives the user a
quick overview of the hierarchy.

Hierarchy Mode lists entities in preorder by entity type
(see Hierarchy Mode in The Illuminator Vision section, page
3). It shows bounding box location and size, child type,
content data type, and the content itself. (Note that DAFS
does not use a true hierarchy in the entity structure. For
more information, see the DAFS section beginning on page 3.)

Entity Selection

Entities are selected by clicking on the entity marker
representing them. As in the other Modes, more than one
entity may be selected at a time by depressing the shift key
during selection. Selections made in other Modes before
switching to Hierarchy Mode are also apparent. If a selected
entity is located in a collapsed section(see below), it will
not be visible.

CAUTION: In this Mode, it is easy to select a higher-level
entity, such as the Document, and enter text contents for it
via View Selection. This can cause confusion later when
viewing the text from Text Mode. Because Text Mode gives
precedence to contents from higher-level entities, only the
contents of the Document entity would appear. It could look
as though the other data had been lost, when in fact it
would merely not be on display.

Rearranging the Hierarchy

Once you've selected one or more entities, you may drag them
to a new location (new parent) by clicking the left mouse
button, and holding it down while dragging the selection to
its new parent. As a guide, an outline of the selected
entities will move as the mouse moves.

Newly Created Entities

The Entity Creation feature of Image Mode insert new
entities into the entity structure, and Illuminator has
algorithms for determining the type of new entities. (See
"Entity Creation and Adjustment"). An entity created this
way is placed in the hierarchy as the last child of its
parent, and not necessarily in the position implied by its
location on image. It can easily be moved to a new place.
This feature is best used where a numerous entities are
being created on an image from scratch. When you work with
this tool, it is best to begin with the highest order
entity, then define all its children of the next highest
type in order, then the higher order children of each child,
etc. Following this patter will decrease the need to move
entities around afterward.

New Entity in the Edit Menu creates entities of type "Type"
(which may be edited via View Selection), and places them at
the beginning of parent's list of children. This feature is
most useful for modifying pre-existing entity structures. In
this case, new entities are placed at the beginning of the
list of children so that they will be immediately available
for editing. Otherwise, you'd be scrolling through the
hierarchy.

Collapsing and Uncollapsing

A node in the hierarchy are indicated by a downward pointing
triangle ()to the left of the entity type. These indicate
entities with one or more child entities below them.
Portions of the hierarchy may be collapsed at these points
by clicking on the triangle. The lines describing the
children disappear, and the triangle points right ().

To uncollapse, click again on the triangle. The hidden lines
will be restored, and the triangle returned to its down-
pointing orientation.


Viewing and Editing Entities

Selected entities can be accessed and edited in Hierarchy
Mode by invoking View Selection, which allows you to view
the image and edit content, content data type, entity type,
and bounding box dimensions and location.

                    Working in Text Mode
                              
Text Mode is capable of displaying, editing and saving ASCII
and Unicode files as well as DAFS files. It will display
only text, so text-less documents will show a blank screen
in this Mode.

Displaying Text

If positional information about the file is available, Text
Mode approximates the look of the original image; otherwise,
it presents a stream of text. For Glyphs with a possibility
set as content, the first possibility is displayed as the
text content.

The Mag option is not available for Text Mode.

Moving in Text Mode

Choose a point in the text and position the cursor there
using the mouse. The Previous, Next, Up and Down options
from the Edit Menu are now available. Previous and Next move
one character back or forward. Up and Down move one line up
or down. Scroll bars are available where necessary.

The Page Menu allows movement from page to page for multiple
page documents.

Selecting and Editing Text

Text Mode is capable of insertion, deletion, and overtype.
The mouse determines the insertion point. Delete from the
Edit Menu deletes the character to the left of the cursor,
if the selection is a single point, or all selected text if
a range is selected. Regions of text may be selected by
click-hold-dragging the mouse, and will appear in reverse
video. Text in selected regions will be replaced by whatever
is typed over it.

Entities of a given type can be treated as a discrete area
of text, or displayed on separate pages, via the Type
Properties dialog. This feature allows you to inject into
Text Mode a sense of where some of the entity boundaries
are.

The text of an entity which was previously selected in the
other Modes will also be selected in Text Mode.

       Creating a Glyph Training Set And Ground Truth
                              
An Illuminator user may wish to create a glyph training set
and a collection of perfect documents for error counting
(such perfect documents are called "ground truth" in OCR
circles). An off-the-shelf or custom OCR product could be
used to scan the test and training documents, segment the
images into glyphs (creating bounding boxes), and perform
the initial recognition. It is also possible, though very
time-consuming, to use Illuminator's entity creation tools
to segment and label example glyphs for the training set.
This would involve drawing each individual bounding box and
entering each label by hand.

To obtain perfectly labeled documents, the user would first
correct any flagged errors, which can best be done in Image
Mode or Flagged Mode. The next step would be to open Out-of-
Context Mode and page through all the Glyphs, correcting any
mislabeled Glyphs which appear there, and removing noise.
Bad bounding boxes can be spotted and, switching to Image
Mode, corrected. Once all errors are fixed, the document can
be saved, and an error counter could use it for ground
truth.

For the documents that are intended to be training sets, the
same files above could be used, but the user would perform
the additional steps of deleting undesirable Glyphs (for
example noisy, broken, or atypical ones).

                              
                              
                        _Appendix A.
                              
              Illuminator Command Accelerators
                              
Enter        ctrl-F9   Opens a dialog for Unicode character
Unicode                input.
Japanese     ctrl-F10  Opens a dialog for inputting
Input                  Japanese characters.
Save As      ctrl-a    Opens a dialog for file storage.
Back         ctrl-b    Displays the page preceding the
                       current page.
OOC Mode     ctrl-c    Switches to Out-of-Context Mode.
New Entity   ctrl-e    Creates a new child entity of type
                       "Type".
Forward      ctrl-f    Displays the page following the
                       current page.
Find Next    ctrl-g    Moves selection to the next flagged
Flag                   entity.
Hierarchy    ctrl-h    Switches to Hierarchy Mode
Mode
Image Mode   ctrl-i    Switches to Image Mode.
Go to        ctrl-l    If single entity selected, goes to
selection              and centers on selected entity.
                       Refreshes screen regardless of
                       selection.
Delete       Delete    Deletes a single selected text or
                       entity and children.
Down         Down      If a single entity is selected or
                       the text cursor is inserted, moves
                       selection to the entity or character
                       below.
Flagged      ctrl-m    Switches to Flagged Mode.
Mode
Grow Down    ctrl-     Expands a single selected bounding
             Down      box by moving the bottom edge down
                       one pixel.
Grow Right   ctrl-     Expands a single selected bounding
             Right     box by moving the right edge one
                       pixel to the right.
Kill         ctrl-u    Deletes selected entity's content in
Contents               Flagged Mode.
Mag 1        ctrl-1    Changes magnification to original
                       size.
Mag 1/2      ctrl-     Changes magnification to 1/2
             shift-2   original size.
Mag 1/4      ctrl-     Changes magnification to 1/4
             shift-4   original size.
Mag 1/8      ctrl-     Changes magnification to 1/8
             shift-8   original size.
Mag 2        ctrl-2    Changes magnification to twice
                       original size.
Mag 4        ctrl-4    Changes magnification to four times
                       original size.
Mag 8        ctrl-8    Changes magnification to eight times
                       original size.
Move Down    shift-    Moves a single selected bounding box
             Down      down one pixel.
Move Left    shift-    Moves a single selected bounding box
             Left      one pixel to the left.
Move Right   shift-    Moves a single selected bounding box
             Right     one pixel to the right.
Move Up      shift-Up  Moves a single selected bounding box
                       up one pixel.
Next         Right     If a single entity is selected or
                       the text cursor is inserted, moves
                       selection to the next entity or
                       character.
Next File    ctrl-n    Opens the alphanumerically next
                       file, same suffix.
Open         ctrl-o    Opens a dialog for file retrieval.
Page Down    PgDn      Moves down one screen.
Page Left    ctrl-     Moves left one screen.
             PgDn
Page Right   ctrl-     Moves right one screen.
             PgUp
Page Up      PgUp      Moves up one screen.
Previous     Left      If a single entity is selected or
                       the text cursor is inserted, moves
                       selection to the previous entity or
                       character.
Previous     ctrl-p    Opens alphanumerically previous
File                   file, same suffix.
Quit         ctrl-q    Quits Illuminator.
Save         ctrl-s    Saves the current file; does not
                       quit.
Select...    ctrl-x    Opens page selection dialog.
Shrink Down  ctrl-Up   Shrinks a single selected bounding
                       box by moving the bottom edge up one
                       pixel.
Shrink       ctrl-     Shrinks a single selected bounding
Right        Left      box by moving the right edge one
                       pixel to the left.
Sort         ctrl-v    Sorts entities.
Text         ctrl-t    Switches to Text Mode.
Text on      ctrl-k    Toggles whether entity contents
Image                  displayed (default on).
Up           Up        If a single is entity selected or
                       the text cursor is inserted, moves
                       selection to the entity or character
                       above.
View         ctrl-     Brings up View Selection dialog.
Selection    shift-v
                              
   Appendix B. Trademarks, Copyrights and Acknowledgments
                              
The use of the term Unicode Standard in this document refers
to The Unicode Standard Worldwide Character Encoding,
Version 1.0, The Unicode Consortium, Addison-Wesley, 1991.
It also incorporates the Unicode 1.0.1 Addendum.

OSF/Motif is a trademark of the Open Software Foundation,
and refers to the OSF/Motif application programming
interface. Illuminator currently uses Revision 1.2.

PDA (Processed Document Architecture) is a trademark of
Calera Recognition Systems, Inc. (Now a part of Caere
Corporation.)

The use of the term Linux PC in this document refers to IBM-
compatible PC's running the Linux Operating System created
by Linus Torvalds. It resembles the Unix Operating System of
Unix Systems Laboratories, but is completely free and non-
derivative.

The use of the term TIFF in this document refers to the
Tagged Image File Format, invented by Aldus.

                            Index
                              
                                16, 23, 27, 29, 31, 57
American, 32, 45                
Arabic, 2, 5, 32, 45            Keyboard, ii, 5, 6, 8, 32,
                                 45
Back, ii, 40, 41, 50, 57        Kill Contents, ii, 23, 24,
Bounding Box Field, The, 35      51, 57
Box Button, The, 35             
British, 32, 45                 low rec Glyphs, 37
                                low-rec Glyphs, 37
Center, ii, 42, 45, 48, 49,     
 50, 51                         Mag Option Menu, 17
Collapse, iii, 53               MICR, 32, 45
Color Table, 38, 48             Modes Option Menu, 17
Command Line Options, 15,       Move Down, 42, 43, 49, 58
 16                             Move Left, 42, 43, 49, 58
Contents Field, The, 34         Move Right, 42, 43, 49, 58
                                Move Up, 42, 43, 49, 58
Delete, ii, 23, 26, 31, 48,     multi rec Glyphs, 37
 51, 54, 57                     multi-rec Glyphs, 37
Display, ii, iii, 2, 17,        
 32, 36, 37, 39, 45, 48         New, i, ii, iii, 2, 18, 19,
Display in Image Mode, ii,       23, 26, 46, 49, 53, 57
 36, 37, 39, 48                 New Entity, ii, iii, 2, 23,
Down, ii, 18, 21, 23, 25,        26, 46, 49, 53, 57
 40, 42, 43, 48, 49, 50,        Next, ii, 18, 21, 23, 24,
 51, 54, 57, 58                  26, 40, 48, 50, 51, 54,
                                 57, 58
Edit, ii, 17, 23, 34, 35,       Next File, ii, 18, 21, 58
 39, 48, 50, 51, 53, 54         non rec Glyphs, 37
Enter Unicode, ii, 5, 8,        non-rec Glyphs, 37
 23, 26, 57                     Norwegian, 32, 45
Entity, ii, iii, 2, 4, 13,      
 17, 23, 24, 26, 34, 38,        O0 Glyphs, 37
 39, 42, 43, 45, 46, 48,        Open, i, 18, 19, 58, 59
 49, 50, 51, 52, 53, 57         
                                Page, ii, iii, 17, 18, 21,
File, i, ii, 17, 18, 19,         22, 40, 48, 50, 51, 54,
 20, 21, 44, 58, 59              58
File Status Label, ii, 17,      Page Down, ii, 18, 21, 58
 44                             Page Left, ii, 22, 58
Find Next Flag, ii, 23, 26,     Page Right, ii, 22, 58
 51, 57                         Page Up, ii, 18, 21, 58
Forward, ii, 40, 41, 50, 57     Preferences, ii, 17, 36, 48
French, 32, 45                  Previous, ii, 18, 21, 23,
                                 24, 40, 48, 50, 51, 54,
German, 32, 45                   58
Greek, 32, 45                   Previous File, ii, 18, 21,
Grow Down, 42, 43, 49, 57        58
Grow Right, 42, 43, 49, 57      Property Field, The, 34
                                
Image Field, The, 33            Quit, ii, 22, 58
Italian, 32, 45                 
                                Revert to Saved, i, 5, 18,
Japanese Input, ii, 6, 8,        21
 Rotate Image, ii, 42
Russian, 32, 45

Save, i, 18, 20, 57, 58
Save As, i, 18, 20, 57
Select, ii, 30, 33, 38, 39,
 40, 48, 49, 50, 58
Shrink Down, 42, 43, 49, 58
Shrink Right, 42, 43, 49,
 58
Spanish, 32, 45
Swedish, 32, 45

Text On Image, 36
Type Button, The, 33
Type Preferences, 48

Uncollapse, iii, 13, 53
Up, i, ii, 15, 18, 21, 23,
 25, 40, 42, 43, 48, 49,
 50, 51, 54, 58

View Selection, ii, 12, 14,
 24, 26, 32, 33, 34, 46,
 49, 52, 53
                              
