--------------------------------------------------------------------------------

WordNet Gloss Disambiguation Project
First release
04/01/2008

CONTACT: Helen Langone, hlangone@princeton.edu

--------------------------------------------------------------------------------

Release contents
----------------

This release, once extracted, is comprised of three subdirectories:

/WordNet-3.0/glosstag/merged    =WordNet glosses in merged format
/WordNet-3.0/glosstag/standoff  =WordNet glosses in standoff format
/WordNet-3.0/glosstag/dtd       =DTD describing the markup for the merged annotations


/WordNet-3.0/glosstag/merged

Contains WordNet glosses with all annotations combined in a single file. There 
are four XML files in this directory, which group the glosses by WordNet part
of speech. 

/merged/noun.xml
/merged/verb.xml
/merged/adj.xml
/merged/adv.xml

The XML data in these files conforms to the glosstag.dtd found in the /dtd 
directory in this release. See glosstag.dtd for detailed information about the 
encoding of the merged files.


/WordNet-3.0/glosstag/standoff

Contains WordNet glosses annotated using standoff markup in XCES format. Annotations 
for sense tags and other markup are stored in documents separate from the gloss
text. For the standoff annotation, the database of 117,659 WordNet glosses was 
split over 1,177 files of 100 synsets each. To make this more manageable, the files
are organized in 11 subdirectories of 10 sub-subdirectories each, and the final 8
sub-subdirectories in the 12th.

/standoff/00 - containing sub-subdirs 000-009
/standoff/01 - containing sub-subdirs 010-019
/standoff/02 - containing sub-subdirs 020-029
/standoff/03 - containing sub-subdirs 030-039
/standoff/04 - containing sub-subdirs 040-049
/standoff/05 - containing sub-subdirs 050-059
/standoff/06 - containing sub-subdirs 060-069
/standoff/07 - containing sub-subdirs 070-079
/standoff/08 - containing sub-subdirs 080-089
/standoff/09 - containing sub-subdirs 090-099
/standoff/10 - containing sub-subdirs 100-109
/standoff/11 - containing sub-subdirs 110-117

Each sub-subdir contains glosses and standoff annotations for 10 files of 100 
synsets each. There are seven files for each subgroup of 100 synsets. 

[prefix].txt = the gloss text
[prefix].wn = the header file, which contains the XCES header for the text
[prefix]-wngloss.xml = annotations for gloss structure
[prefix]-wnann.xml = token-level annotations
[prefix]-wnword.xml = sense tag annotations for single-word forms
[prefix]-wncoll.xml = sense tag annotations for multi-word forms, treating discontiguous spans as contiguous
[prefix]-wndc.xml = same as wncoll.xml, but using proposed markup for annotating discontiguous spans

The filename prefix consists of wsd- plus a 6 digit number that is the sub-subdir
name concatenated with a 3-digit number indicating 100-synset subgroup. For 
example, the first chunk of 100 synsets in the first sub-subdir (/000) would be 
named with a prefix of wsd-000000, the second chunk would have prefix wsd-000100,
the third wsd-000200, and so on.

The encoding for these files is detailed below.

There are indexes mapping WordNet 3.0 synsets, sense keys, and terms to filename 
prefix plus path to it. The indexes are found in 

/standoff/index.byid.tab
/standoff/index.bysk.tab
/standoff/index.bylem.tab
/standoff/index.bylem.noun.tab
/standoff/index.bylem.verb.tab
/standoff/index.bylem.adj.tab
/standoff/index.bylem.adv.tab

All indexes are tab-delimited, mapping an item to one or more filename prefixes.
The paths are relative from /standoff, eg., 00/000/, 00/001/, etc. The basic
file structure is: 

item<tab>path+prefix[<tab>path+prefix]*

index.byid.tab maps a WordNet synset by its id (consisting of {n,v,a,r}+offset) to 
the path+prefix of the files containing annotations for that synset. There is a 
one-to-one correspondence between synset id and path+prefix. Eg., 

n00001740<tab>00/000/wsd-000000
n00001930<tab>00/000/wsd-000000
n00002137<tab>00/000/wsd-000000

index.bysk.tab maps a WordNet sense key to the path+prefix of the files 
containing annotations for the sense key's synset. There is a one-to-one 
correspondence between sense key and path+prefix. Eg., 

abandonment%1:04:01::<tab>00/000/wsd-000100
abandonment%1:04:02::<tab>00/000/wsd-000300
abdominoplasty%1:04:00::<tab>00/000/wsd-000000
abort%1:04:00::<tab>00/000/wsd-000000
abscondment%1:04:00::<tab>00/000/wsd-000100
absence_without_leave%1:04:00::<tab>00/000/wsd-000100
absolute_space%1:03:00::<tab>00/000/wsd-000000
absolution%1:04:00::<tab>00/000/wsd-000300


index.bylem.tab maps WordNet terms to path+prefix. A "term" here is the WordNet
synset term lower-cased, and with underscores between words (for multi-word terms).
Since a term may appear in more than one synset, there is a one-to-many 
correspondence between term and path+prefix. Eg.,

abandonment<tab>00/000/wsd-000300<tab>00/000/wsd-000100
abdominoplasty<tab>00/000/wsd-000000
abort<tab>00/000/wsd-000000
abscondment<tab>00/000/wsd-000100
absence_without_leave<tab>00/000/wsd-000100
absolute_space<tab>00/000/wsd-000000
absolution<tab>00/000/wsd-000300
abstract_entity<tab>00/000/wsd-000000

The index.bylem.{noun,verb,adj,adv}.tab indexes are identical in format to 
index.bylem.tab, and contain mappings for WordNet terms broken down by their part 
of speech.


Document Encoding
-----------------

The annotated disambiguated glosses are provided in both standoff and merged 
formats. The former uses the XCES markup for standoff annotations, the XML 
schemas for which are found at http://www.xces.org/schema/2003/. While XCES is 
an emerging standard for representing corpus data, certain aspects of the gloss
annotations are not representable with it in its current form (0.4). The
gloss annotations contain markup for discontiguous spans of text, as when a 
WordNet collocation (multi-word form) is interrupted by a word (usually a
conjunction, or, in the case of phrasal verbs, an object). Examples of 
discontiguous collocations in the glosses are:

personal or business relationship = personal_relationship, business_relationship
canon and civil law = canon_law, civil_law
take a player out = take_out
put oneself forward = put_forward
pay them off = pay_off
bring something about = bring_about

With this project we are considering the body of glosses to be a corpus, however,
the data also contains information relevant to the WordNet synset. This information
does not fit neatly within the XCES standard. For this reason, and the fact that 
we allow for the annotation of discontiguous collocations, the merged files are 
encoded in a format specific to the project, described by glosstag.dtd.

The WordNet glosses are marked up with definition and example sentence boundaries,
and within these, text is tokenized and marked up with part of speech, potential
lemma forms, and a small set of semantic classes (indicating the token is punctuation, 
abbreviation, acronym, number, year, currency, or some kind of symbol). Collocations
are delimited, including markup to indicate discontiguous forms. Words and
collocations that have been disambiguated are further annotated with WordNet sense 
keys. 


Encoding of merged files
------------------------

Detailed information about the encoding of the merged files is found in 
glosstag.dtd. The dtd also lists values for part of speech and other attributes
that are found in the standoff annotation files. The following is a sample of 
the merged annotation for the noun synset at offset 00003553. Line numbers are
for reference.

1  <synset id="n00003553" ofs="00003553" pos="n">
2   <terms>
3    <term>whole</term>
4    <term>unit</term>
5   </terms>
6   <keys>
7    <sk>whole%1:03:00::</sk>
8    <sk>unit%1:03:00::</sk>
9   </keys>
10  <gloss desc="orig">
11   <orig>an assemblage of parts that is regarded as a single entity; "how big is that part compared to the whole?"; "the team is a unit"</orig>
12  </gloss>
13  <gloss desc="text">
14   <text>an assemblage of parts that is regarded as a single entity ; “ how big is that part compared to the whole ? ” ; “ the team is a unit ”</text>
15  </gloss>
16  <gloss desc="wsd">
17   <def id="n00003553_d">
18    <wf id="n00003553_wf1" lemma="an" pos="DT" tag="ignore">an</wf>
19    <wf id="n00003553_wf2" lemma="assemblage%1" pos="NN" tag="man">
20  <id id="n00003553_id.4" lemma="assemblage" sk="assemblage%1:14:01::"/>assemblage</wf>
21    <wf id="n00003553_wf3" lemma="of" pos="IN" tag="ignore">of</wf>
22    <wf id="n00003553_wf4" lemma="part%1|parts%1|part%2" pos="NNS" tag="un">parts</wf>
23    <wf id="n00003553_wf5" lemma="that" pos="WDT" tag="ignore">that</wf>
24    <wf id="n00003553_wf6" lemma="i%1|be%2" pos="VBZ" tag="un">is</wf>
25    <cf coll="a" id="n00003553_wf7" lemma="regard%2" pos="VBN" tag="un">
26  <glob coll="a" glob="man" id="n00003553_coll.a" lemma="regard_as%2" tag="man">
27   <id coll="a" id="n00003553_id.5" lemma="regard as" sk="regard_as%2:31:00::"/>
28    </glob>regarded</cf>
29    <cf coll="a" id="n00003553_wf8" lemma="as" pos="IN" tag="ignore">as</cf>
30    <wf id="n00003553_wf9" lemma="a" pos="DT" tag="ignore">a</wf>
31    <wf id="n00003553_wf10" lemma="single%1|single%2|single%3" pos="JJ" tag="un">single</wf>
32    <wf id="n00003553_wf11" lemma="entity%1" pos="NN" sep="" tag="auto">
33  <id id="n00003553_id.1" lemma="entity" sk="entity%1:03:00::"/>entity</wf>
34    <wf id="n00003553_wf12" pos=":" tag="ignore" type="punc">;</wf>
35   </def>
36   <ex id="n00003553_ex1">
37    <qf rend="dq">
38  <wf id="n00003553_wf13" lemma="how" tag="ignore">how</wf>
39  <wf id="n00003553_wf14" lemma="big%3|big%4" tag="un">big</wf>
40  <wf id="n00003553_wf15" lemma="i%1|be%2" tag="un">is</wf>
41  <wf id="n00003553_wf16" lemma="that" tag="ignore">that</wf>
42  <wf id="n00003553_wf17" lemma="part%1|part%2|part%4" tag="un">part</wf>
43  <wf id="n00003553_wf18" lemma="compare%2" tag="un">compared</wf>
44  <wf id="n00003553_wf19" lemma="to" tag="ignore">to</wf>
45  <wf id="n00003553_wf20" lemma="the" tag="ignore">the</wf>
46  <wf id="n00003553_wf21" lemma="whole%1|whole%3|whole%4" sep="" tag="auto">
47   <id id="n00003553_id.2" lemma="whole" sk="whole%1:03:00::"/>whole</wf>
48  <wf id="n00003553_wf22" sep="" tag="ignore" type="punc">?</wf>
49    </qf>
50    <wf id="n00003553_wf23" tag="ignore" type="punc">;</wf>
51   </ex>
52   <ex id="n00003553_ex2">
53    <qf rend="dq">
54  <wf id="n00003553_wf24" lemma="the" tag="ignore">the</wf>
55  <wf id="n00003553_wf25" lemma="team%1|team%2" tag="un">team</wf>
56  <wf id="n00003553_wf26" lemma="i%1|be%2" tag="un">is</wf>
57  <wf id="n00003553_wf27" lemma="a" tag="ignore">a</wf>
58  <wf id="n00003553_wf28" lemma="unit%1" sep="" tag="auto">
59   <id id="n00003553_id.3" lemma="unit" sk="unit%1:03:00::"/>unit</wf>
60    </qf>
61    <wf id="n00003553_wf29" tag="ignore" type="punc">;</wf>
62   </ex>
63  </gloss>
64 </synset>

Lines 2 through 12 contain synset terms, synset keys, and the original text
of the gloss from the WordNet synset. Lines 13 through 15 contain the tokenized
gloss text, unannotated. Line 16 starts the annotations. Automatic processing
of the glosses delimited the definition text (lines 17 to 35), and two example
sentences (lines 36-51, and 52-62), and tokenized the text. Each token is wrapped 
in either <wf> or <cf>, where <wf> is a word token, and <cf> is a token that is 
part of a collocation. The <glob> tag contains the main markup for a collocation, 
and also its sense tag, if disambiguated. The <id> carries sense tag information 
for disambiguated words (when child of wf) and collocations (when child of glob).

Each annotatable item has a unique id. Within example sentences, only the synset
terms were disambiguated.


Encoding of standoff files
--------------------------

In the standoff annotation files, the annotations are contained in separate XML
documents linked to the original gloss text, and also linked to the merged files 
via ids. For each [prefix].txt document containing gloss text, there are six XML 
files. We took the standoff format used by the second release of the ANC as the 
basis of our format, and modified it from there. In the ANC format, each standoff
annotation file is composed of a series of annotations consisting of one or more
features (zero or more, in our case). An annotation is represented by a <struct>
tag, and features as child <feat> tags. Each <struct> specifies an edge between
two nodes in the gloss text. Nodes are located between characters in the text, so
an edge between two nodes spans a stretch of text. The following example is from
the .txt file containing the gloss text for the synset above:

3             3                   3                   4                   4
7             8                   9                   0                   1
3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2
|a|n| |a|s|s|e|m|b|l|a|g|e| |o|f| |p|a|r|t|s| |t|h|a|t| |i|s| |r|e|g|a|r|d|e|d| 

The following annotation delimits the token "assemblage" as an edge between
nodes 376 and 386.

<struct from="376" id="n00003553_wf2" to="386" type="wf"/>


Header ([prefix].wn)

Contains information about the source data, and links to the annotation documents
for a filename prefix. For example:

<annotations>
  <annotation ann.loc="wsd-100000.txt">Document content</annotation>
  <annotation ann.loc="wsd-100000-wngloss.xml">WordNet glosses with definitions and examples delimited</annotation>
  <annotation ann.loc="wsd-100000-wncoll.xml">WordNet collocations (multi-word forms) and sense tags</annotation>
  <annotation ann.loc="wsd-100000-wnword.xml">WordNet single-word forms and sense tags</annotation>
  <annotation ann.loc="wsd-100000-wnann.xml">WordNet token-level annotations</annotation>
  <annotation ann.loc="wsd-100000-wndc.xml">WordNet markup for discontiguous collocations and sense tags</annotation>
</annotations>


Annotations for gloss structure ([prefix]-wngloss.xml)

The annotations in this file delimit gloss definitions and example sentences. The
following sample applies to the merged gloss above.


1  <?xml version="1.0" encoding="UTF-8"?>
2  <cesAna version="1.0.4" xmlns="http://www.xces.org/schema/2003" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.xces.org/schema/2003 http://www.xces.org/schema/2003/xcesAna.xsd">
.
.
.
9   <struct from="373" id="n00003553_d" to="432" type="def"/>
10  <struct from="433" id="n00003553_ex1" to="477" type="ex"/>
11  <struct from="478" id="n00003553_ex2" to="497" type="ex"/>


Token-level annotations ([prefix]-wnann.xml)

The annotations in this file delimit tokens as wf (word forms), cf (collocation
forms, punc (punctuation), or ignore (stoplist words). The features for each
annotation indicate the attributes from the merged file for the token, as well
as the token text. See glosstag.dtd for explanations of attribute values.

The following sample applies to the tokenized definition in the merged example 
above. 

1  <?xml version="1.0" encoding="UTF-8"?>
2  <cesAna version="1.0.4" xmlns="http://www.xces.org/schema/2003" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.xces.org/schema/2003 http://www.xces.org/schema/2003/xcesAna.xsd">
.
.
.
422 <struct from="373" id="n00003553_wf1" to="375" type="ignore">
423  <feat name="lemma" value="an"/>
424  <feat name="pos" value="DT"/>
425  <feat name="tag" value="ignore"/>
426  <feat name="text" value="an"/>
427 </struct>
428 <struct from="376" id="n00003553_wf2" to="386" type="wf">
429  <feat name="lemma" value="assemblage%1"/>
430  <feat name="pos" value="NN"/>
431  <feat name="tag" value="man"/>
432  <feat name="text" value="assemblage"/>
433 </struct>
434 <struct from="387" id="n00003553_wf3" to="389" type="ignore">
435  <feat name="lemma" value="of"/>
436  <feat name="pos" value="IN"/>
437  <feat name="tag" value="ignore"/>
438  <feat name="text" value="of"/>
439 </struct>
440 <struct from="390" id="n00003553_wf4" to="395" type="wf">
441  <feat name="lemma" value="part%1|parts%1|part%2"/>
442  <feat name="pos" value="NNS"/>
443  <feat name="tag" value="un"/>
444  <feat name="text" value="parts"/>
445 </struct>
446 <struct from="396" id="n00003553_wf5" to="400" type="ignore">
447  <feat name="lemma" value="that"/>
448  <feat name="pos" value="WDT"/>
449  <feat name="tag" value="ignore"/>
450  <feat name="text" value="that"/>
451 </struct>
452 <struct from="401" id="n00003553_wf6" to="403" type="wf">
453  <feat name="lemma" value="i%1|be%2"/>
454  <feat name="pos" value="VBZ"/>
455  <feat name="tag" value="un"/>
456  <feat name="text" value="is"/>
457 </struct>
458 <struct from="404" id="n00003553_wf7" to="412" type="cf">
459  <feat name="coll" value="a"/>
460  <feat name="lemma" value="regard%2"/>
461  <feat name="pos" value="VBN"/>
462  <feat name="tag" value="un"/>
463  <feat name="text" value="regarded"/>
464 </struct>
465 <struct from="413" id="n00003553_wf8" to="415" type="cf">
466  <feat name="coll" value="a"/>
467  <feat name="lemma" value="as"/>
468  <feat name="pos" value="IN"/>
469  <feat name="tag" value="ignore"/>
470  <feat name="text" value="as"/>
471 </struct>
472 <struct from="416" id="n00003553_wf9" to="417" type="ignore">
473  <feat name="lemma" value="a"/>
474  <feat name="pos" value="DT"/>
475  <feat name="tag" value="ignore"/>
476  <feat name="text" value="a"/>
477 </struct>
478 <struct from="418" id="n00003553_wf10" to="424" type="wf">
479  <feat name="lemma" value="single%1|single%2|single%3"/>
480  <feat name="pos" value="JJ"/>
481  <feat name="tag" value="un"/>
482  <feat name="text" value="single"/>
483 </struct>
484 <struct from="425" id="n00003553_wf11" to="431" type="wf">
485  <feat name="lemma" value="entity%1"/>
486  <feat name="pos" value="NN"/>
487  <feat name="tag" value="auto"/>
488  <feat name="sep" value=""/>
489  <feat name="text" value="entity"/>
490 </struct>
491 <struct from="431" id="n00003553_wf12" to="432" type="punc">
492  <feat name="pos" value=":"/>
493  <feat name="tag" value="ignore"/>
494  <feat name="type" value="punc"/>
495  <feat name="text" value=";"/>
496 </struct>


Single-word form annotations ([prefix]-wnword.xml)

The annotations in this file are for word forms that are disambiguatable (ie., are 
tokens that are not punctuation, and are not part of a collocation). Sense keys 
for disambiguated word forms are indicated by a feature.

The following sample applies to the merged gloss above.

1  <?xml version="1.0" encoding="UTF-8"?>
2  <cesAna version="1.0.4" xmlns="http://www.xces.org/schema/2003" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.xces.org/schema/2003 http://www.xces.org/schema/2003/xcesAna.xsd">
.
.
.
137 <struct from="376" id="n00003553_wf2" to="386" type="man">
138  <feat name="wnsk" value="assemblage%1:14:01::"/>
139  <feat name="text" value="assemblage"/>
140 </struct>
141 <struct from="390" id="n00003553_wf4" to="395" type="un">
142  <feat name="text" value="parts"/>
143 </struct>
144 <struct from="401" id="n00003553_wf6" to="403" type="un">
145  <feat name="text" value="is"/>
146 </struct>
147 <struct from="418" id="n00003553_wf10" to="424" type="un">
148  <feat name="text" value="single"/>
149 </struct>
150 <struct from="425" id="n00003553_wf11" to="431" type="auto">
151  <feat name="wnsk" value="entity%1:03:00::"/>
152  <feat name="text" value="entity"/>
153 </struct>


Multi-word form annotations ([prefix]-wncoll.xml)

The annotations in this file are for collocations. In order to conform to the 
current XCES standard, discontiguous collocations in this file are represented as
contiguous, spanning any intervening text. Therefore, this file is both correct
(conformant to the standard) and incorrect (some collocations are wrongly 
delimited).

As for single-word forms, sense keys for disambiguated collocations are indicated
by a feature.

The following sample applies to the merged gloss above.

1  <?xml version="1.0" encoding="UTF-8"?>
2  <cesAna version="1.0.4" xmlns="http://www.xces.org/schema/2003" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.xces.org/schema/2003 http://www.xces.org/schema/2003/xcesAna.xsd">
3   <struct from="404" id="n00003553_coll.a" to="415" type="man">
4    <feat name="wnsk" value="regard_as%2:31:00::"/>
5    <feat name="text" value="regarded as"/>
6   </struct>


Discontiguous collocation annotations ([prefix]-wndc.xml)

The annotations in this file are the same as that of [prefix]-wncoll.xml, except
that discontiguous collocations are represented using the markup proposed by 
Keith Suderman and Nancy Ide in their paper "Layering and Merging Linguistic 
Annotations" (http://acl.ldc.upenn.edu/W/W06/W06-2716.pdf). Since the markup
used in this file does not conform to the XCES standard, it will not validate
under the schema, nor be usable with software written based on the current standard.
We believe it might be useful for users who wish to write or modify their own 
software in order to make use of the correctly-delimited discontiguous
collocations.

The following sample is the representation of "neurological and visual disorders",
where "neurological disorders" is discontiguous. The contiguous collocation 
"visual disorders" is encoded as in the [prefix]-wncoll.xml file. First, two
pseudo nodes are created to reference the discontiguous parts (lines 37 and 38), 
and then an edge is created between the two pseudo nodes (lines 39 to 42).

37   <node id="n00337605_coll.b.1" ref="n00337605_wf18" type="part"/>
38   <node id="n00337605_coll.b.2" ref="n00337605_wf21" type="part"/>
39   <struct from="n00337605_coll.b.1" id="n00337605_coll.b" to="n00337605_coll.b.2" type="auto">
40    <feat name="wnsk" value="neurological_disorder%1:26:00::"/>
41    <feat name="text" value="neurological disorders"/>
42   </struct>
43   <struct from="2919" id="n00337605_coll.a" to="2935" type="auto">
44    <feat name="wnsk" value="visual_disorder%1:26:00::"/>
45    <feat name="text" value="visual disorders"/>
46   </struct>

The corresponding merged markup is this:

<cf coll="b" id="n00337605_wf18" lemma="neurological%3" pos="JJ" tag="un">
<glob coll="b" glob="auto" id="n00337605_coll.b" lemma="neurological_disorder%1" tag="auto">
 <id coll="b" id="n00337605_id.1" lemma="neurological disorder" sk="neurological_disorder%1:26:00::"/>
</glob>neurological</cf>
<wf id="n00337605_wf19" lemma="and" pos="CC" tag="ignore">and</wf>
<cf coll="a" id="n00337605_wf20" lemma="visual%3" pos="JJ" tag="un">
<glob coll="a" glob="auto" id="n00337605_coll.a" lemma="visual_disorder%1" tag="auto">
 <id coll="a" id="n00337605_id.2" lemma="visual disorder" sk="visual_disorder%1:26:00::"/>
</glob>visual</cf>
<cf coll="a,b" id="n00337605_wf21" lemma="disorder%1|disorder%2" pos="NNS" sep="" tag="un">disorders</cf>


Character encoding
------------------
All .txt files within the standoff dirs are encoded as UTF-16. All other files are
UTF-8.


Disclaimer
----------

While standoff annotations have many benefits, particularly the ability to isolate 
annotations of choice, it is not a well-supported format. Our standoff encoding is 
based heavily on the ANC format, but is not identical to it as our markup is 
necessarily different. Therefore, some tools that work with the ANC data may work 
with ours, but not all. We are supplying the data in this format as a service to 
users who are used to working with standoff annotations, and who will build or 
modify existing software to work with it. We are not supporting the ANC standoff 
annotation format, nor any software that uses or manipulates it, nor are we 
providing any tools ourselves.

The standoff annotations do not contain more, or better, information than the 
merged files. The annotations contained in them are identical to the merged data, 
just reformulated in a different way. If you have any doubts about which format 
to use, then use the merged files.


--------------------------------------------------------------------------
WordNet Gloss Disambiguation Project Copyright (c) 2008
by Princeton University. All rights reserved.