The RORIC-LING Bulletin

months 7 - 12

General Questions

Those unfamiliar with WordNet should read "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look at
                                                           http://www.cogsci.princeton.edu/~wn/
under "Publications".

If already familiar with WordNet, we recommend the book "WordNet: An Electronic Lexical Database", which is now available from MIT Press. The book includes articles describing the design and contents of WordNet (an update to Five Papers on WordNet as well as papers reporting on research done with WordNet in the areas of linguistics, information retrieval, word sense disambiguation, semantic concordance building, text analysis, and knowledge engineering). The book and CD-ROM can be purchased directly from MIT Press.

We do not know of any free XML documentation in Romanian on the web. However, a very good book, translated into Romanian, exists: 

Lee Anne Phillips, XML. Teora Publishing House, 2001

You should have downloaded a file called "wn16pc.exe". If you downloaded it correctly, you should be able to simply double-click on this file and it should extract itself. Then you must follow the instructions in the INSTALL.txt file to actually install the WordNet package.

Please look a
                                          http://www.cogsci.princeton.edu/~wn/doc.shtml
You should see a list of WordNet manuals, available there online.

Some WordNet-related projects are the following:

• Semantic networks - in languages other than English

• Web Interfaces - access WordNet over a network

• Local Interfaces - require files to be downloaded

• .NET

• C++

• COM

• dBase/MySQL

• Java

• Lisp

• Palm

• Perl

• Prolog

• Python

• Extensions - expand WordNet's features or integrate it into larger systems

More information on these projects can be found at
                                           http://www.cogsci.princeton.edu/~wn/links.shtml

The Global WordNet Association is a free, public and non-commercial organization that provides a platform for discussing, sharing and connecting wordnets for all languages in the world.

Questions Concerning the Program GenSynsets

The XML format facilitates, in this case, presenting the results of the program on the Web, and allows quicker access to information. XML (the Extensible Markup Language) was first ratified by the W3C (World Wide Web Consortium) as the standard for information exchange on the Internet in February 1998. XML specifies a rigorous, text-based way to represent the structure inherent in data so that it can be authored and interpreted unambiguously.

The default charset (when parameter CS is not specified in the command line) is iso-8859-1.

The I/O operations on files which enable working with data in various languages (coded with various character sets) have been written by using the Java specifications InputStreamReader, BufferedReader, and OutputStreamWriter, BufferedWriter respectively.

The program uses both Java functions for handling strings and user-defined methods for parsing rows of text read from the dictionaries, and for extracting strings that are significant for the algorithm (for instance, word translations).

The program defines a unique class, GenSynsets, which contains all variables and methods necessary in the implementation of the algorithms on which it is based. Of the defined methods we would like to mention those referring to I/O operations using the dictionary files, those for the labeling of the generated e-sets, the method which implements the backtracking-type strategy, sorting operations and the operations on strings.

The program finds the gloss corresponding to an English synset by using the getGloss() method which belongs to the Synset class.

In order to determine the final synsets (corresponding to the chosen foreign language) the program combines the e-sets labeled with the maximum value, corresponding to each eword, and eliminates all duplicate words. This combining is implemented by means of the backtracking method.

In this situation only those words of the synsets are used for which a translation exists (namely for which corresponding entries in the English-Foreign Language Dictionary exist).

Among the programming techniques which we have used the most important is the Backtracking method by means of which the fsynsets are generated, starting from the esets, and with elimination of duplicates.

Our program GenSynsets uses results which were obtained within the project JWordNet. JwordNet is a pure Java standalone object-oriented interface to the WordNet database of lexical relationships. It is intended for Java programmers who wish to write portable Java applications that use a local copy of the WordNet files, or who find JWordNet's object-oriented interface preferable to the procedural interface that the C library (and native method interfaces built on top of it) provide. It includes the Dictionary broker class, the IndexWord, Synset, Word, and Pointer domain entity classes, and the POS and PointerType enumeration holders.

In this case, going over all the synsets of the American WordNet is based on an enumeration and on the method synsets which belongs to the class DictionaryDatabase (JWordNet).

When running the program according to a list of synsets (specified by the corresponding offsets) two arguments are passed to the method: POS (part of speech)and the offset of a synset. Finding the synsets of the American WordNet relies on the method getSynsetAt()which belongs to the class FileBackedDictionary (JWordNet).

Questions Concerning the RORIC-LING Implementation of WN Algorithms

Because, for the sake of simplicity, within this demo we have worked with cleared, combined synsets where we have given up certain parameters like the sense number. The existing algorithm can be easily modified in order for it to work with synsets where this parameter exists and to take it into account. Also, in order to generate completely coded WN-type adjective clusters in a foreign language, one must first obtain all adjective synsets of that language, in order to see all existing meanings of a certain adjective in the target language. This is something we haven't done yet for Romanian, due to the existing incomplete bilingual dictionaries in electronic format. On the other hand, our claim was not to have generated clusters identical to those in WN, but just WN-type adjective synsets and clusters. Again, the existing computer programs can be easily modified in order to obtain the exact form of a WN adjective cluster.

As already mentioned, many adjectives are limited as to the syntactic positions they can occupy, and that limitation is usually coded in WordNet. Because it is a word-form limitation, it is coded for individual adjectives rather than for synsets. Consider, for instance, the cluster awake/asleep, both of which are limited to predicate position. Although these are the head words of the cluster, the limitation does not hold for all of the synonyms in the cluster. Therefore, the individual words so limited are all coded with (p). For adjectives limited to prenominal (attributive) position, the code is (a). And, finally, for those few adjectives that can appear only immediately following a noun, the code is (ip) for "immediately postnominal". This codification has not yet been implemented for Romanian adjective clusters.

Because this is not an exhaustive discussion concerning WordNet and/or the semiautomatic generation of WordNets for other languages. We have chosen to discuss the two basic WordNet structures - namely the hierarchy and the cluster - for which we have studied nouns and adjectives in WordNet. For a detailed discussion concerning all WordNet translation issues see the project BALKANet at
                                                                                                                                              http://www.ceid.upatras.gr/Balkanet

The main difficulties which occurred when automatically translating the English synsets into Romanian ones were generated by the so-called "false friends", by collocations, by loan translation, and by the fact that the polysemy of many English words is greatly superior to that of the corresponding Romanian words. For detailed explanations concerning all these phenomena please read the linguistic comments that you will find in the web page of the project.

General Questions Concerning WordNet

The lexical matrix is a matrix in which word forms are imagined to be listed as headings for the columns; word meanings are headings for the rows. An entry in a cell of the matrix implies that the form in that column can be used (in an appropriate context) to express the meaning in that row. If there are two entries in the same column, the word form is polysemous; if there are two entries in the same row, the two word forms are synonyms (relative to a context). For more information see "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look under "Publications" at the address
                                                                 http://www.cogsci.princeton.edu/~wn/

WordNet distinguishes between semantic relations and lexical relations; the emphasis is still on semantic relations between meanings, but relations between words are also included. However, WordNet is organized by semantic relations, which are indicated by pointers.

Synonymy is the central relation in WN. The definition of synonymy in terms of substitutability makes it necessary to partition WordNet into nouns, verbs, adjectives and adverbs.

As it is noted in "Five Papers on WordNet", "if concepts are represented by synsets, and if synonyms must be interchangeable, then words in different syntactic categories cannot be synonyms (cannot form synsets) because they are not interchangeable. Nouns express nominal concepts, verbs express verbal concepts, and modifiers provide ways to qualify those concepts. In other words, the use of synsets to represent word meanings is consistent with psycholinguistic evidence that nouns, verbs, and modifiers are organized independently in semantic memory".

Unlike synonymy and antonymy, which are lexical relations between word forms, hyponymy/hypernymy (also called subordination/superordination, subset/superset, or the ISA relation) is a semantic relation between word meanings.

Hyponymy is transitive and asymmetrical (Lyons, 1977, vol. 1), and, since there is normally a single superordinate, it generates a hierarchical semantic structure, in which a hyponym is said to be below its superordinate. Such hierarchical representations are widely used in the construction of information retrieval systems, where they are called inheritance systems (Touretzky, 1986): a hyponym inherits all the features of the more generic concept and adds at least one feature that distinguishes it from its superordinate and from any other hyponyms of that superordinate. This convention provides the central organizing principle for the nouns in WordNet.

A semantic relation in WordNet is the part-whole (or HASA) relation, known to lexical semanticists as meronymy/holonymy. The meronymic relation is transitive (with qualifications) and asymmetrical (Cruse, 1986), and can be used to construct a part hierarchy (with some reservations, since a meronym can have many holonyms). It is assumed that the concept of a part of a whole can be a part of a concept of the whole, although it is recognized that the implications of this assumption deserve more discussion than they receive within this framework. Meronymy is present in the organization of noun synsets.

As it is noted in "Five Papers on WordNet", "an important class of lexical relations are the morphological relations between word forms. Initially, interest was limited to semantic relations; no plans were made to include morphological relations in WordNet. As work progressed, however, it became increasingly obvious that if WordNet was to be of any practical use to anyone, it would have to deal with inflectional morphology. For example, if someone put the computer's cursor on the word trees and clicked a request for information, WordNet should not reply that the word was not in the database. A program was needed to strip off the plural suffix and then to look up tree, which certainly is in the database. This need led to the development of a program for dealing with inflectional morphology".

The prototypical definition points upward, to a superordinate term, not sideways to coordinate terms or downward to hyponyms. For instance, the definition of tree in standard paper dictionaries points to the superordinate term plant, but contains no information about coordinate terms. A dictionary definition draws some important distinctions and serves to remind the reader of something that is presumed to be familiar already; it is not intended as a catalogue of general knowledge.

The semantic relation that is represented in WordNet by '@->' has been called the ISA relation, or the hypernymic or superordinate relation (since it points to a hypernym or superordinate term); it goes from specific to generic and so is a generalization. The inverse semantic relation '~->' goes from generic to specific (from superordinate to hyponym)and so is a specialization.

As it is noted in "Five Papers on WordNet", "since a noun usually has a single superordinate, dictionaries include the superordinate in the definition; since a noun can have many hyponyms, English dictionaries do not list them (the French dictionary "Le Grand Robert" is an exception). Even though the specialization relation is not made explicit in standard dictionaries of English, it is a logical derivative of the generalization relation. In WordNet, lexicographers code the generalization relation '@->'explicitly with a labeled pointer between lexical concepts or senses. When the lexicographers' files are converted automatically into the lexical database, one step in this process is to insert inverse pointers for the specialization relation '~->'. Thus, the lexical database is a hierarchy that can be searched upward or downward with equal speed." Computer scientists call such hierarchies "inheritance systems", because they think of specific items inheriting information from their generic superordinates. All the properties of the superordinate are assumed to be properties of the subordinate as well; instead of listing those properties redundantly with both items, they are listed only with the superordinate and a pointer from the subordinate to the superordinate is understood to mean "for additional properties, look here".

WN is indeed a lexical inheritance system. A systematic effort has been made to connect hyponyms with their superordinates (and vice versa). In the WN data base, an entry for tree, for instance, contains a reference, or pointer '@->', to an entry for plant; the pointer is labeled "superordinate" by the arbitrary symbol '@'. In the database, the pointer '@' to the superordinate plant will be reflected by an inverse pointer '~' to tree in the synset for plant; that pointer is labeled "hyponym" by the arbitrary symbol '~'. The computer is programmed to use these labeled pointers to construct whatever information a user requests; the arbitrary symbols '@' and '~' are suppressed when the requested information is displayed. The synset for tree would look something like:
                                                                                                 {tree,plant,@ conifer,~alder,~…}
where the '…' is filled with many more pointers to hyponyms. The synset for plant would look something like
                                                                                                 {plant,flora,organism,@ tree,~…}.

The first person to make this claim explicit seems to have been Quillian (1967, 1968). Experimental tests of Quilian's proposal were reported in a seminar paper by Collins and Quillian (1969), who assumed that reaction times can be used to indicate the number of hierarchical levels separating two meanings.

An alternative conclusion - the conclusion on which WordNet is based - is that the inheritance assumption is correct, but that reaction times do not measure what Collins and Quillian, as well as other experimentalists, assumed they did. Perhaps reaction times indicate a pragmatic rather than a semantic distance - a difference in word use, rather than a difference in word meaning (Miller and Charles, 1991).

In WN the nouns are partitioned with a set of semantic primes, namely a relatively small set of generic concepts is selected and each one of them is treated as the unique beginner of a separate hierarchy. These multiple hierarchies correspond to relatively distinct semantic fields, each with its own vocabulary.

WN has adopted the following set of 25 unique beginners:

                                                                      {act, action, activity}                                        {natural object}
                                                                      {animal, fauna}                                                  {natural phenomenon}
                                                                      {artifact}                                                             {person, human being}
                                                                      {attribute, property}                                         {plant, flora}
                                                                      {body, corpus}                                                  {possession}
                                                                      {cognition, knowledge}                                   {process}
                                                                      {communication}                                               {quantity, amount}
                                                                      {event, happening}                                           {relation}
                                                                      {feeling, emotion}                                              {shape}
                                                                      {food}                                                                  {state, condition}
                                                                      {group, collection}                                            {substance}
                                                                      {location, place}                                                 {time}
                                                                      {motive}

The most important criterion in choosing these primitive semantic components is that, collectively, they should provide a place for every English noun. The resulting hierarchies vary widely in size and are not mutually exclusive - some cross-reference is required - but on the whole they cover distinct conceptual and lexical domains. They were selected after considering the possible adjective-noun combinations that could be expected to occur (an analysis carried out by Philip N. Johnson-Laird).

The hierarchies of nominal concepts in WN are said to have a level, somewhere in the middle, where most of the distinguishing features are attached. It is referred to as the basic level, and the nominal concepts at this level are called basic-level categories or generic concepts (Berlin, Breedlove, and Raven, 1966, 1973). Rosch (1975; Rosch, Mervis, Gray, Johnson, and Boyes-Braem, 1976) extended this generalization: for concepts at the basic level, people can list many distinguishing features. Above the basic level, descriptions are brief and general. Below the base level, little is added to the features that distinguish basic concepts.

As it is noted in "Five Papers on WordNet", "as the coverage of WordNet increased, it became increasingly obvious that alternative senses of a word could not always be identified by the use of synonyms. Rather late in the game, therefore, it was decided to include distinguishing features in the same way that conventional dictionaries do, by including short explanatory glosses as a part of synsets containing polysemous words. These are marked off from the rest of the synset by parentheses".

As it is noted in "Five Papers on WordNet", "meronyms are distinguishing features that hyponyms can inherit. Consequently, meronymy and hyponymy become intertwined in complex ways. For example, if beak and wing are meronyms of bird, and if canary is a hyponym of bird, then, by inheritance, beak and wing must also be meronyms of canary".

The connections between meronymy and hyponymy are complicated by the fact that parts are hyponyms as well as meronyms. The example which is given in "Five Papers on WordNet" is the synset {beak, bill, neb}, which is a hyponym of {mouth, muzzle}, which in turn is a meronym of {face, countenance} and a hyponym of {orifice, opening}. A frequent problem in establishing the proper relation between hyponymy and meronymy arises from a general tendency to attach features too high in the hierarchy. For example, if wheel is said to be a meronym of vehicle, then sleds will inherit wheels they should not have. Indeed, in WN a special synset was created for the concept {wheeled vehicle}.

Meronyms tend to occur most frequently in connection with words denoting physical objects. In WN, meronymy is found primarily in the {body, corpus}, {artifact}, and {quantity, amount} hierarchies.

The "part-of" relation is often compared to the "kind of" relation: both are asymmetric and (with reservations) transitive, and can relate terms hierarchically (Miller and Johnson-Laird, 1976). That is to say, parts can have parts: a finger is a part of a hand, a hand is a part of an arm, an arm is a part of a body: the term finger is a meronym of the term hand, hand is a meronym of arm, arm is a meronym of body. But the "part of" construction is not always a reliable test of meronymy. In many instances transitivity seems to be limited (Lyons, 1977).

For more information see "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look at
                                                           http://www.cogsci.princeton.edu/~wn/
under "Publications".

Winston et al.(1987) differentiate six types of meronyms: component-object (branch/tree), member-collection (tree/forest), portion-mass (slice/cake), stuff-object (aluminium/airplane), feature-activity (paying/shopping), and place-area (Princeton/New Jersey). Chaffin, Hermann, and Winston (1988) add a seventh: phase-process (adolescence/growing up). Meronymy is obviously a complex semantic relation - or set of relations. Only three of these types of meronymy are coded in WN: "is a component part of", "is a member of", and "is the stuff that it is made from". Of these three, the "is a component of" relation is by far the most frequent.

As it is noted in "Five Papers on WordNet", "semantic opposition is not a fundamental organizing relation between nouns, but it does exist and so merits its own representation in WordNet. For example, the synsets for man and woman would contain:
                                                { [man, woman,!], person,@ … (a male person) }
                                                { [woman, man,!], person,@ … (a female person) }
where the symmetric relation of antonymy is represented by the '!' pointer, and square brackets indicate that antonymy is a lexical relation between words, rather than a semantic relation between concepts.

The main semantic relations which are taken into consideration in WN with respect to nouns are hyponymy, , meronymy, and antonymy. When all these three kinds of semantic relations are included, the result is a highly interconnected network of nouns.

The adjective synsets in WN contain mostly adjectives, although some nouns and prepositional phrases that function frequently as modifiers have been entered as well. The discussion within RORIC-LING is limited to adjectives.

WN divides adjectives into two major classes: descriptive and relational. Descriptive adjectives ascribe to their head nouns values of (typically) bipolar attributes and consequently are organized in terms of binary oppositions (antonymy) and similarity of meaning (synonymy). Descriptive adjectives that do not have direct antonyms are said to have indirect antonyms by virtue of their semantic similarity to adjectives that do have direct antonyms. WN contains pointers between descriptive adjectives expressing a value of an attribute and the noun by which that attribute is lexicalized. Reference-modifying adjectives have special syntactic properties that distinguish them from other descriptive adjectives. Relational adjectives are assumed to be stylistic variants of modifying nouns and so are cross-referenced to the noun files. Chromatic color adjectives are regarded as a special case.

A descriptive adjective is one that ascribes a value of an attribute to a noun. That is to say, x is Adj presupposes that there is an attribute A such that A(x)=Adj. To say "The package is heavy" presupposes that there is an attribute WEIGHT such that WEIGHT (package) = heavy. Similarly, low and high are values for the attribute HEIGHT. WordNet contains pointers between descriptive adjectives and the noun synsets that refer to the appropriate attributes.

The semantic organization of descriptive adjectives is entirely different from that of nouns. There is no relation generating nominal hierarchies in the case of adjectives. The semantic organization of adjectives is more naturally thought of as an abstract hyperspace of N dimensions rather than as a hierarchical tree.

The basic semantic relation among descriptive adjectives is antonymy. The importance of antonymy first became obvious from results obtained with word association tests. The importance of antonymy in the organization of descriptive adjectives is understandable when it is recognized that the function of these adjectives is to express values of attributes, and that nearly all attributes are bipolar. Antonymous adjectives express opposing values of an attribute. For example, the antonym of heavy is light, which expresses a value at the opposite pole of the WEIGHT attribute. In WordNet, this binary opposition is represented by reciprocal labeled pointers: heavy!->light and light!->heavy.

Because many descriptive adjectives have no antonyms WN has introduced a similarity pointer and has used it to indicate that the adjectives lacking antonyms are similar in meaning to adjectives that do have antonyms. Gross, Fischer, and Miller (1989) proposed that adjective synsets be regarded as clusters of adjectives associated by semantic similarity to a focal adjective that relates the cluster to a contrasting cluster at the opposite pole of the attribute. Gross, Fischer and Miller distinguish direct antonyms like heavy/light, which are conceptual opposites that are also lexical pairs, from indirect antonyms, like heavy/weightless, which are conceptual opposites that are not lexically paired. Under this formulation, all descriptive adjectives have antonyms; those lacking direct antonyms have indirect antonyms, i.e., are synonyms of adjectives that have direct antonyms.

Some descriptive adjectives do not have direct antonyms. However, in the adjective organization of WN, those lacking direct antonyms have indirect antonyms, i.e., are synonyms of adjectives that have direct antonyms. Under this formulation, all descriptive adjectives have antonyms.

In WN those adjectives lacking direct antonyms have indirect antonyms, i.e., are synonyms of adjectives that have direct antonyms. Direct antonyms are represented by an antonymy pointer, '!->'; indirect antonyms are inherited through similarity, which is indicated by the similarity pointer, '&->'.

The basic model presented by the authors of WN with respect to adjectives consists of dividing adjectives into two major types, descriptive (which enter into clusters based on antonymy) and relational (which are similar to nouns used as modifiers). Without claiming complete coverage, the authors of WN believe that this model accounts for the majority of English adjectives.

A gradable adjective can be defined as one whose value can be multiplied by such adverbs of degree as very, decidedly, intensely, rather, quite, somewhat, pretty, extremely (Cliff, 1959).

Gradation must also be considered as a semantic relation organizing lexical memory for adjectives (Bierwisch, 1989). For some attributes gradation can be expressed by ordered strings of adjectives, all of which point to the same attribute noun in WordNet.

As it is noted in "Five Papers on WordNet", "it would not be difficult to represent ordered relations by labeled pointers between synsets, but it was estimated that not more than 2% of the more than 2,500 adjective clusters could be organized in that way. Since the conceptually important relation of gradation does not play a central role in the organization of adjectives, it has not been coded in WordNet."

The noun that names the attribute - e.g., LENGTH - and all the adjectives expressing values of that attribute (in this case long, short, lengthy, etc.) are linked in WordNet by a pointer.

In WN the opposition colored/colorless (cross-referenced to chromatic/achromatic) is used to introduce the names of colors. Hues are coded as similar to colored, and the shades of gray from white to black are coded as similar to gray, which is in a tripartite cluster with white and black, providing for a graded continuum.

Relational adjectives, which were first discussed at length by Levi (1978), mean something like "of, relating/pertaining to, or associated with" some noun, and they play a role similar to that of a modifying noun.

For example, fraternal as in fraternal twins relates to brother, and dental as in dental hygiene, is related to tooth.

The main differences are the following: 

1. Relational adjectives differ from descriptive adjectives in that they do not relate to an attribute. 

2. Relational adjectives do not refer to a property of their head nouns. 

3. Relational adjectives, like nouns and unlike descriptive adjectives, are not gradable. 

4. Relational adjectives do not have direct antonyms; although they can often be combined with non-, such forms do not express the opposite value of an attribute but something like "everything else". Since relational adjectives do not have antonyms, they cannot be incorporated into the clusters that characterize descriptive adjectives.

WordNet maintains a separate file of relational adjectives with pointers to the corresponding nouns. For more details see "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look at
                                                           http://www.cogsci.princeton.edu/~wn/
under "Publications".

WordNet maintains a separate file of relational adjectives with pointers to the corresponding nouns.

Some 1,700 relational adjective synsets containing over 3,000 individual lexemes are currently included in WordNet. Each synset consists of one or more relational adjectives, followed by a pointer to the appropriate noun.

For more details see "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look at
                                                           http://www.cogsci.princeton.edu/~wn/
under "Publications".

The numerals following certain items distinguish different subsenses or different privileges of occurrence - for example, the dried-up1 of a water hole in one synset and the dried-up2 of autumn leaves or fruit in another. Each of these cases, furthermore, contains parenthetical information designed to help distinguish these particular senses or indicate acceptable contexts.

As it is noted in "Five Papers on WordNet", "in addition to the lowercase within-cluster pointers, many head synsets contain pointers to other, related clusters. In the AWAKE/ASLEEP cluster, the capitalized pointer ALERT,& points to the head word of the ALERT/UNALERT cluster. These capitalized pointers are planned to serve as "see also" cross-references to related clusters.

The restricted within-cluster coding leads to a problem when closely related attributes are expressed by more than one pair of antonyms. In such cases, exactly the same set of synsets can be related to two different antonymous pairs, some of which are presently in different clusters. (Consider large/small and big/little). In such cases a single cluster has been created headed by both pairs, thus avoiding unnecessary redundancy. In addition, a particular synset can be coded with two pointers, one to its own cluster head, the other to the head of an outside cluster.

As it is explained in "Five Papers on WordNet", "verbs are divided into 15 files, largely on the basis of semantic criteria. All but one of these files correspond to what linguists have called semantic domains: verbs of bodily care and functions, change, cognition, communication, competition, consumption, contact, creation, emotion, motion, perception, possession, social interaction, and weather verbs. Virtually all the verbs in these files denote events or actions. Another file contains verbs referring to states, such as suffice, belong, and resemble, that could not be integrated into the other files. The verbs in this latter group do not constitute a semantic domain, and share no semantic properties other than that they refer to states. This file, whose organization resembles that of the adjectives in WordNet, consists of small semantic clusters. The division of verbs into 14 files corresponding to different semantic domains, each containing event and action verbs, and one file containing semantically diverse stative verbs reflects the division between the major conceptual categories EVENT and STATE found in Jackendoff's (1983) and Dowty's (1979) analyses".

The principle of lexical inheritance can be said to underlie the semantic relations between nouns, and bipolar oppositions serve to organize the adjectives. Similarly, the different relations that organize the verbs can be cast in terms of one overarching principle, lexical entailment (or strict implication).

The entailment relation between verbs resembles meronymy between nouns, but meronymy is better suited to nouns than to verbs. The following example concerning verbs is offered in "Five Papers on WordNet":

"Snoring or dreaming can be a part of sleeping, in the sense that the two activities are, at least partially, temporally co-extensive: the time that you spend snoring or dreaming is a proper part of the time you spend sleeping. And it is true that when you stop sleeping you also necessarily stop snoring or dreaming."

A verb X will be said to include a verb Y if there is some stretch of time during which the activities denoted by the two verbs co-occur, but no time during which Y occurs and X does not. If there is a time during which X occurs but Y does not, X will be said to properly include Y. A simple generalization of this would be the following: if X entails Y, and if a temporal inclusion relation holds between them, then people will accept a part-whole statement relating Y and X.

The sentence frame used to test hyponymy between nouns, An x is a y, is not suitable for verbs, because it requires that x and y be nouns. The semantic distinction between two verbs is different from the features that distinguish two nouns in a hyponymic relation.

The many different kinds of elaborations that distinguish a 'verb hyponym' from its superordinate have been merged into a manner relation that Fellbaum and Miller (1990) have dubbed troponymy (from the Greek tropos, manner or fashion). The troponymy relation between two verbs can be expressed by the formula To X is to Y in some particular manner.

For more details see "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look at
                                                           http://www.cogsci.princeton.edu/~wn/
under "Publications".

Troponymy is a particular kind of entailment, in that every troponym X of a more general verb Y also entails Y. Consider the pair limp-walk, which represents the example offered in "Five Papers on WordNet". The authors comment this example as follows: "The verbs in this pair are related by troponymy: to limp is also to walk in a certain manner; limp is a troponym of walk. The verbs are also in an entailment relation: the statement He is limping entails He is walking, and walking can be said to be a part of limping. Unlike the activities denoted by snore and sleep, or buy and pay, the activities referred to by a troponym and its more general superordinate are always temporally co-extensive, in that one must necessarily be walking every instant that one is limping. Troponymy therefore represents a special case of entailment: pairs that are always temporally co-extensive and are related by entailment".

Two kinds of entailment with temporal inclusion are accounted for in WN. One type of entailment is troponymy (limp-walk), while entailment without troponymy refers to pairs of verbs (snore-sleep) related only by entailment and proper temporal inclusion.

As it is noted in "Five Papers on WordNet", "many verb pairs in an opposition relation also share an entailed verb. For example, both hit and miss entail aim, because one must necessarily aim in order to hit or miss". In contrast to other kinds of entailment, "these verbs are not related by temporal inclusion. The activities denoted by hit (or miss) and aim occur in a sequential order: in order to either hit or miss , one must have aimed first; aiming is a precondition for both hitting and missing."  For more information see "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look at
                                                           http://www.cogsci.princeton.edu/~wn/
under "Publications".

The four types of entailment relations among verbs that have been taken into consideration in WordNet are the following:

• entailment + temporal inclusion + troponymy (limp-walk);

• entailment + temporal inclusion - troponymy (snore-sleep, buy-pay);

• entailment - temporal inclusion + backward presupposition (succeed-try, untie-tie);

• entailment - temporal inclusion + cause (raise-rise, give-have).

For more information see "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look at
                                                           http://www.cogsci.princeton.edu/~wn/
under "Publications".

As it is explained in "Five Papers on WordNet", to cover at least the most important syntactic aspects of verbs "WordNet includes for each verb synset one or several sentence frames, which specify the subcategorization features of the verbs in the synset by indicating the kinds of sentences they can occur in. This information permits one to quickly search among the verbs for the kinds of semantic-syntactic regularities studied by Levin and others."

For more information see "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look at
                                                           http://www.cogsci.princeton.edu/~wn/
under "Publications".

Here are the main verb files in the American WordNet:

• Verbs of bodily functions and care (sweat, shiver, faint, ache, tire, sleep, freeze);

• Verbs of change (change, alter, vary, modify);

• Verbs of communication (beg, order, thank);

• Competition verbs (fight, race);

• Consumption verbs (drink, eat);

• Contact verbs (scrub, wipe, squeeze);

• Cognition verbs (reasoning, judging, learning, memorizing);

• Creation verbs (engrave, sew, bake);

• Motion verbs (move, travel);

• Emotion verbs (amuse, encourage);

• Stative verbs (surround, cross, reach);

• Perception verbs (watch, spy, survey);

• Verbs of possession (hold, own, give, transfer, take, receive);

• Verbs of social interaction (impeach, excommunicate);

• Weather verbs (rain, thunder, snow).

For more information see "Five Papers on WordNet" which is available in PostScript, Acrobat (PDF), or as a compressed tar file containing the nroff source. Please look at
                                                           http://www.cogsci.princeton.edu/~wn/
under "Publications".