1: <HTML>
2: <HEAD>
3: <META NAME="Generator" CONTENT="Corel WordPerfect 8">
4: <TITLE>REQUIREMENTS OF TEXT PROCESSING LEXICONS</TITLE>
5: </HEAD>
6: <BODY TEXT="#000000" LINK="#0000ff" VLINK="#551a8b" ALINK="#ff0000" BGCOLOR="#c0c0c0">
7:
8: <P><U>REQUIREMENTS OF TEXT PROCESSING LEXICONS</U><A HREF="#N_1_"><SUP>(1)</SUP></A></P>
9:
10: <P>Kenneth C. Litkowski<A HREF="#N_2_"><SUP>(2)</SUP></A></P>
11:
12: <P>16729 Shea Lane, Gaithersburg, Md. 20760</P>
13:
14: <P>As text processing systems expand in scope, they will require ever larger lexicons along with a parsing capability for
15: discriminating among many senses of a word. Existing systems do not incorporate such subtleties in meaning for their
16: lexicons. Ordinary dictionaries contain such information, but are largely untapped. When the contents of dictionaries are
17: scrutinized, they reveal many requirements that must be satisfied in representing meaning and in developing semantic
18: parsers. These requirements were identified in research designed to find primitive verb concepts. The requirements are
19: outlined and general procedures for satisfying them through the use of ordinary dictionaries are described, illustrated by building frames for and examining the definitions of "change" and its uses as a hypernym in other definitions.</P>
20:
21: <P>1. <U>INTRODUCTION</U></P>
22:
23: <P>Five years ago, <A HREF="#bol">Bolinger</A> (1975, pp. 220-224), in discussing the increasing incorporation of meaning into linguistics, noted
24: that these efforts had not yet made use of the insights of lexicography. The few substantial efforts, such as those
25: spearheaded by Olney (<A HREF="#ol68">1968</A>, <A HREF="#ol72">1972</A>), <A HREF="#mel">Mel'cuk</A> (1978), <A HREF="#smi">Smith</A> (1972), <A HREF="#sim">Simmons</A> (1975), and <A HREF="#leh">Lehmann</A> (1976) made some
26: progress, but never came to fruition. Today, lexicography and its products, the dictionaries, remain an untapped resource of
27: uncertain value. Indeed, many who have analyzed the contents of a dictionary have concluded that it is of little value to
28: linguistics or artificial intelligence. Because of the size and complexity of a dictionary, perhaps such a conclusion is
29: inevitable, but I believe it is wrong. To view the real potential of this resource, it is first necessary to develop a
30: comprehensive model within which a dictionary's detail can be tied together. When this is done, the examination of
31: definitions makes it possible to identify some requirements for semantic representation of lexical entries and for semantic
32: parsers to be used in natural language processing systems. I describe herein what I have learned from this type of effort.</P>
33:
34: <P>The principal purpose of this paper is only to show that a dictionary can provide very useful insights about what should go
35: into a parser. I have not attempted to identify all the types of procedures that should go into a parser nor have I described a
36: comprehensive and self-contained system that demonstrates how everything can be tied neatly together in a computer
37: implementation. The specifications for such a system would be quite complex, as befits the complexity of a dictionary.
38: However, I have adopted many of the notions in a manual system used to search for primitive verb concepts. In this system,
39: I have developed and used a 400 page inverse dictionary, a thousand page coded dictionary, an elaborate index card filing
40: system, and complex listings of rules I have followed. Many of these rules are based on the general schema presented in this
41: paper. At this time, these notions have not only proved successful, but also have produced a wellspring of ideas yet to be
42: fully articulated. I hope that I can portray some inkling of the fascination that has led me ever deeper into the recesses of a
43: dictionary.</P>
44:
45: <P>It should be noted at the outset that the definitions in a dictionary were not developed to fit a grand scheme of semantic
46: representation, possessing self-contained logical consistency (if such exists). The definitions have many flaws and the
47: procedures I am following are uncovering many such flaws. In part, I expect that the rigorous approach necessary to extract
48: the meaning content of definitions can assist in bringing about greater self-consistency within the overall structure of a
49: dictionary.</P>
50:
51: <P>2. <U>GENERAL DESCRIPTION OF MY RESEARCH</U></P>
52:
53: <P>I began my research (see <A HREF="#lit">Litkowski</A> (1978) for a more complete description) with the objective of identifying primitive verb
54: concepts by following definitional paths within <EM>Webster's Third New International Dictionary</EM> (<A HREF="#w3">W3</A>). (All definitions quoted
55: in this paper are taken from W3.) To search for primitives, it was first necessary to develop a comprehensive framework
56: within which definitions could be analyzed. The theory of labeled directed graphs (digraphs) provided such a framework.
57: Using digraphs, I developed several increasingly detailed models of the semantic structure of a dictionary. In these models, a
58: point or node represents one or more definitions or concepts and a line or arc between these points represents a derivational
59: relationship between definitions. Using such models, theorems of digraph theory were used to predict the existence and form
60: of primitives within a dictionary. This justified continued effort to attempt to find such primitives. However, it immediately
61: became clear that this would entail the development of semantic representations for definitions and the development of a
62: semantic parser to transform definitions into these representations.</P>
63:
64: <P>The models showed that the big problem to be overcome in trying to find the primitives is the <U>apparent</U> rampant circularity
65: of defining relationships. To eliminate these apparent vicious circles, it is necessary to make a precise identification of
66: derivational relationships, specifically, to find the specific definition that provides the sense in which its definiendum is used
67: in defining another word. When this is done, the spurious cycles are broken and precise derivational relationships are
68: identified. Although this can be done manually, the sheer bulk of a dictionary requires that it be done with well-defined
69: procedures, i.e. with syntactic and semantic parsing. It is in the attempt to lay out the elements of such a parser that the
70: requirements for semantic representations have emerged. The knowledge thus gained, developed incrementally and
71: embodied in reduction rules, is then incorporated into procedures for the continued search for primitives. Thus far, these
72: procedures have been used to reduce the initial set of 20,000 verbs in <A HREF="#w3">W3</A> to fewer than 4,000, with the prospect of much
73: further reduction as the parsing principles are adopted. The search for primitives and the development of a semantic parser
74: are proceeding hand-in-hand.</P>
75:
76: <P>3. <U>GENERAL REQUIREMENTS OF A PARSER</U></P>
77:
78: <P>A lexical entry in a text processing system must contain information (1) to permit a parser to identify which sense of the
79: word has been used in an utterance and (2) which characterizes the meaning of that sense so that it can be used for
80: constructing the internal representation of the utterance. A syntactic parser, accessing lexical entries which contain only
81: syntactical information, will take us only part of the way. Elaborate and complex semantic representational and parsing
82: procedures are necessary for a full representation of an utterance.</P>
83:
84: <P>3.1 <U>Syntactic Requirements</U></P>
85:
86: <P>A parser must first be capable of dealing with the syntactic complexity of an utterance. In this paper, it is assumed that this
87: can be accomplished by an ATN-type parser and that semantic parsing principles can be integrated with the syntactic
88: procedures. However, it is recognized that the syntactic capabilities of the parser will continue to evolve; use of dictionary
89: definitions as a large corpus upon which to test syntactic parsing principles may assist in this evolution. Moreover, by
90: subjecting definitions to such a parser, it may be possible to identify and eliminate syntactic flaws overlooked by the
91: lexicographers. On the other hand, the incorporation of semantic parsing into a syntactic parser may necessitate greater
92: efficiencies for a parser. It is suggested below that what is included in a semantic representation of an entry in the lexicon
93: can be used to determine what parsing paths should be pursued.</P>
94:
95: <P>3.2 <U>Semantic Requirements</U></P>
96:
97: <P>The distinguishing characteristic of a semantic parser is that it must be capable of identifying which sense of a word is being
98: used in a particular utterance. Rieger (<A HREF="#rie">1977</A>, <A HREF="#rs">1979</A>) and <A HREF="#sma">Small</A> (1979) have argued that this can be accomplished with sense
99: discrimination nets, but feel that such nets must be developed by the AI community. In what follows, I suggest that nets can
100: be developed through the analysis of definitions in an ordinary dictionary.</P>
101:
102: <P><A HREF="#rie">Rieger</A> (1977) says that, to capture the meaning of a word, we need to identify all possible constructions in which it can
103: participate. He then develops sense selection networks for determining which of the possible constructions is utilized in a
104: particular instance. Such a network is a strategy for selecting an intended sense out of the mass of senses that a word might
105: have. <A HREF="#sma">Small</A> (1979) says that text processing requires complex interactions centered around word experts as the unit of
106: linguistic knowledge. The central parsing process involves understanding the sense or role of a word in a particular context.
107: A word expert is the same as Rieger's sense selection network and is cognizant of all possible contextual interpretations of
108: the word it represents. Each expert should be capable of sufficient context-probing to determine successfully its functional
109: or semantic role. A word expert can suspend its execution, stating conditions upon which it should be resumed (e.g. an
110: adjective should wait for its accompanying noun, which would provide attributes which will help search through the
111: adjective SSN). The parsing process results in a number of "concepts" (picture, event, and setting). Any word expert may
112: make reference to a central tableau of control state descriptions during the disambiguation process.</P>
113:
114: <P>A sense discrimination net consists of an ordered set of questions (the nodes of the network) and for each one the set of
115: possible answers to the question (the branches emanating from each node), with terminal nodes being the semantic
116: representations for a distinct sense. A word expert asks questions designed to identify its sense, but it can also develop other
117: questions which the parser must then attempt to answer. Some of these questions do both at the same time. (e.g. to
118: determine if a word is an adjective or a noun, the parser might need to inquire whether the word to the right is a noun; to go
119: further in analyzing the adjective, the parser would have to ask more detailed information about the sense of the noun.)</P>
120:
121: <P>Rieger says that the questions will fall into the following classes about: (1) adjacent words, (2) the syntax or semantics of
122: adjacent word senses, (3) invariant world knowledge, and (4) dynamic expectancies in the model. (In attempting to build
123: SSN's from dictionary definitions, I want to make clear that I do not want to put general world knowledge or dynamic
124: expectancies of an inference model into the semantic representations of the lexicon. However, it appears that definitions
125: already do contain significant amounts of world knowledge.)</P>
126:
127: <P>In this system, an SSN for a verb is described as a case framework, which is a specification of the syntax and/or semantics of
128: the concepts that can be associated with that verb. Each verb SSN will have to make reference to the entities which the verb
129: senses are capable of governing. At the bottom of the sense selection network will be the meaning case framework which
130: must reflect a semantically accurate labelling of all concepts that it binds. Cases in meaning case frameworks are mandatory
131: since it will otherwise be impossible to discriminate that sense.</P>
132:
133: <P>Word experts affect sense discrimination (both of the instant word and of other words) and augment the conceptual
134: information that constitutes the result of a parse. Small implies that complete disambiguation should take place by the time
135: that a period is reached, I do not think that this necessarily follows. This will be discussed further in section 4.5 (dealing with
136: ambiguity) and section 5 (dealing with multisentence processing).</P>
137:
138: <P><A HREF="#sma">Small</A> (1979) notes that the augmentation of conceptual information that results from a parser is the cutting edge of his
139: research because, although we may be able to complete some disambiguation, e.g. of a noun phrase, we may have to
140: characterize that noun phrase further. As will be shown below, this may entail characterizing a noun phrase as a form, an
141: appearance, a position, a quantity, a stage, a custom, a method, a tendency, or a property, thus going beyond the usual
142: notions of linguistic cases and yet being necessary to an ultimate disambiguation in an SSN. This is part of what I am
143: asserting can be done based on what is in the dictionary, using preposition definitions and perhaps also using constituents of
144: the definitions themselves to help make such characterizations.</P>
145:
146: <P>Rieger feels that sense selection networks can be developed by looking at the constructions in a single story, modeling them
147: into an SSN, going to the next story, augmenting the SSN as necessary, and building the vocabulary in this way. Although
148: this can be done, it will take a long time to build the lexicon in this manner. According to my thesis, this should be facilitated
149: somewhat by using the definitions of a word as found in a dictionary.</P>
150:
151: <P>Based on the foregoing, I believe that the following requirements must be satisfied by a semantic parser implemented
152: through sense selection nets. Diagnostic or differentiating components are needed for each definition. <U>Each definition must
153: have a different semantic representation</U>, even though there may be a core meaning for all the definitions of a word. Since
154: the ability to traverse a net successfully depends on the context in which a word is used, each definition, i.e. each semantic
155: representation, <U>must include slots perhaps with accompanying selectional restrictions) to be filled by that context</U>. The slots
156: will provide a unique context for each sense of a word. Context is what permits disambiguation. Since the search through a
157: net is inherently complex, <U>a definition must drive the parser</U> in the search for context which will fill its slots. These notions
158: are consistent with Rieger's; however, they were identified independently based on my analysis of dictionary definitions.
159: Their viability depends on the ability to describe procedures for developing sense selection networks with the desired
160: representations. This is discussed in section 4.</P>
161:
162: <P>3.3 <U>Representational Formalism</U></P>
163:
164: <P>Although I use a specific formalism for semantic representation in the examples discussed in the next section, it was not
165: developed for other than illustrative purposes. Any of a number of formalisms, such as those developed or used by Rieger
166: (1977), <A HREF="#bob">Bobrow</A> and Winograd (1977), or <A HREF="#nor">Norman</A> et al. (1975), may satisfy the needs described in the next section.
167: However, there are some basic requirements that any formalism must satisfy. For the most part, I follow Rieger in
168: representing a verb definition in the form of n-tuples, with the predicate first, followed by other information, which may
169: include obligatory cases that must be present in surrounding context, the use of sublists for providing selectional restrictions,
170: and certain conditions that must be satisfied. Rieger also permits multiple-component case frames where a definition involves
171: several meaning assertions as the terminal nodes of a sense selection network.</P>
172:
173: <P>Another concept is that of a "descriptor", which Rieger says is used to refer to a concept obliquely by describing it instead of
174: naming it, for use when it comes time to identify the concept at SSN application time or a specific candidate whenever an
175: actual model concept is required. He uses the following formalism:</P>
176:
177: <P>[*D* (var) (featurel) ... (featureN)]</P>
178:
179: <P>where (var) is an arbitrary reference name that satisfies the features. This formalism can be used to represent the unknown
180: subject X of a verb with the features identifying any characteristics that the subject must satisfy. (See <A HREF="#rie">Rieger</A> (1977, pp.
181: 18-21) for further details.) In this schema, sublists within the features may be used to indicate further selectional restrictions
182: that the context must satisfy.</P>
183:
184: <P>4. <U>DEVELOPMENT OF PARSER COMPONENTS</U></P>
185:
186: <P>General procedures for developing sense selection networks are described using the intransitive senses of the verb "change"
187: as examples. For this purpose, it is necessary to consider the intransitive definitions of "change" (shown in Table 1) and
188: those definitions of other verbs where "change" is used intransitively as the main verb (shown in Table 2).</P>
189:
190: <P>In discussing these general procedures, it should be noted that this has not been the principal purpose of my research, so I
191: have not tried to develop them systematically, although that could be done. Rather, I have developed such procedures
192: insofar as they have helped me move closer to the primitives. Since I have not at this time identified the primitives from
193: which all else is supposed to be derived (at least in theory), it goes without saying that any structures which I elaborate for
194: the definitions of "change" will not be complete or accurate. On the other hand, it should be noted that the analysis which I
195: have made shows further elements that have not been previously associated with this verb, and yet it has been accorded
196: primitive status by some. This is one reason why I would argue that many nuances not yet been captured in assertions about
197: meaning representation can be discovered from an analysis of dictionary definitions.</p>
198: <CENTER>
199: <a name="table1"></a><TABLE BORDER="1" ALIGN="CENTER" WIDTH="100%">
200: <TR VALIGN="TOP"><TD COLSPAN="2" WIDTH="112">
201: <P ALIGN="CENTER">Table 1</TD></TR>
202: <TR VALIGN="TOP"><TD COLSPAN="2" ALIGN="CENTER">Intransitive Definitions of "change"</TD></TR>
203: <TR VALIGN="TOP"><TD ALIGN="CENTER">Sense</TD>
204: <TD ALIGN="CENTER" WIDTH="392">Definition</TD></TR>
205: <TR VALIGN="TOP"><a name="def1"></a><TD>1</TD>
206: <TD>become different in one or more respects without becoming
207: something else</TD></TR>
208: <TR VALIGN="TOP"><a name="def1a"></a><TD>la</TD>
209: <TD>lose or acquire some characteristic, property, or tendency</TD></TR>
210: <TR VALIGN="TOP"><TD> </TD>
211: <TD>ALTER</TD></TR>
212: <TR VALIGN="TOP"><a name="def1b1"></a><TD>lb(l)</TD>
213: <TD>pass from one form, appearance, position, state, or stage to
214: another</TD></TR>
215: <TR VALIGN="TOP"><TD> </TD>
216: <TD>SHIFT</TD></TR>
217: <TR VALIGN="TOP"><a name="def1b2"></a><TD>1b(2)</TD>
218: <TD><U>obs</U> pale or blush</TD></TR>
219: <TR VALIGN="TOP"><a name="def1c"></a><TD>1c</TD>
220: <TD>increase or decrease</TD></TR>
221: <TR VALIGN="TOP"><a name="def1d"></a><TD>1d</TD>
222: <TD>adopt different customs, methods, or attitudes</TD></TR>
223: <TR VALIGN="TOP"><TD> </TD>
224: <TD><U>specif</U> experience a religious conversion</TD></TR>
225: <TR VALIGN="TOP"><a name="def1e"></a><TD>1e</TD>
226: <TD><U>of the moon</U> pass from one phase to another</TD></TR>
227: <TR VALIGN="TOP"><TD> </TD>
228: <TD><U>of the moon</U> <U>specif</U> pass through the phase of new moon</TD></TR>
229: <TR VALIGN="TOP"><a name="def1f"></a><TD>1f</TD>
230: <TD><U>chiefly dial</U> turn sour</TD></TR>
231: <TR VALIGN="TOP"><TD> </TD>
232: <TD><U>chiefly dial</U> become tainted</TD></TR>
233: <TR VALIGN="TOP"><a name="def1g"></a><TD>1g</TD>
234: <TD>shift one 's means of conveyance</TD></TR>
235: <TR VALIGN="TOP"><TD> </TD>
236: <TD>TRANSFER</TD></TR>
237: <TR VALIGN="TOP"><a name="def1h"></a><TD>1h</TD>
238: <TD><U>of the voice</U> shift to lower register</TD></TR>
239: <TR VALIGN="TOP"><TD> </TD>
240: <TD><U>of the voice</U> BREAK</TD></TR>
241: <TR VALIGN="TOP"><a name="def1i"></a><TD>1i</TD>
242: <TD><U>Brit</U> shift gears</TD></TR>
243: <TR VALIGN="TOP"><a name="def2"></a><TD>2</TD>
244: <TD>turn into or become something materially different from before</TD></TR>
245: <TR VALIGN="TOP"><a name="def2a"></a><TD>2a</TD>
246: <TD>undergo transformation or conversion - used with <U>into</U></TD></TR>
247: <TR VALIGN="TOP"><a name="def2b"></a><TD>2b</TD>
248: <TD>pass over from one character or state - used with <U>to</U></TD></TR>
249: <TR VALIGN="TOP"><TD> </TD>
250: <TD>undergo transition - used with <U>to</U></TD></TR>
251: <TR VALIGN="TOP"><a name="def2c"></a><TD>2c</TD>
252: <TD>undergo substantial substitution or replacement or be wholly
253: replaced</TD></TR>
254: <TR VALIGN="TOP"><a name="def3"></a><TD>3</TD>
255: <TD>disrobe and rearray oneself more suitably</TD></TR>
256: <TR VALIGN="TOP"><TD> </TD>
257: <TD>disrobe and rearray oneself more suitably in clothes suitable for
258: a social or formal occasion</TD></TR>
259: <TR VALIGN="TOP"><a name="def4a"></a><TD>4a</TD>
260: <TD><U>obs</U> accept something else in return</TD></TR>
261: <TR VALIGN="TOP"><a name="def4b"></a><TD>4b</TD>
262: <TD><U>obs</U> give up what one has in exchange - used with <U>for</U></TD></TR>
263: <TR VALIGN="TOP"><a name="def4c"></a><TD>4c</TD>
264: <TD>engage in giving something and receiving something in return</TD></TR>
265: <TR VALIGN="TOP"><TD> </TD>
266: <TD>EXCHANGE</TD></TR></TABLE>
267: </CENTER>
268: </P>
269: <BR WP="BR1"><BR WP="BR2">
270: <CENTER>
271: <a name="table2"></a><TABLE BORDER="1" ALIGN="CENTER" WIDTH="100%">
272: <TR VALIGN="TOP"><TD COLSPAN="2" ALIGN="CENTER" WIDTH="170">Table 2</TD></TR>
273: <TR VALIGN="TOP"><TD COLSPAN="2" ALIGN="CENTER">Intransitive uses of "change"</TD></TR>
274: <TR VALIGN="TOP"><TD ALIGN="CENTER">Use</TD>
275: <TD ALIGN="CENTER" WIDTH="444">Definition</TD></TR>
276: <TR VALIGN="TOP"><TD><a name="assibilate">assibilate</a> (vi)</TD>
277: <TD>change by introducing a sibilant sound</TD></TR>
278: <TR VALIGN="TOP"><TD><a name="become">become</a> (vi 2a)</TD>
279: <TD>change into being through taking on a new character or characteristic</TD></TR>
280: <TR VALIGN="TOP"><TD><a name="break5c">break</a> (vi 5c)</TD>
281: <TD>change sharply in purport, mood, or attitude</TD></TR>
282: <TR VALIGN="TOP"><TD><a name="break6b">break</a> (vi 6b)</TD>
283: <TD>change abruptly in line or set often with suggestion of opening</TD></TR>
284: <TR VALIGN="TOP"><TD><a name="caramelize">caramelize</a> (vi)</TD>
285: <TD>change to caramel or a caramellike substance or color</TD></TR>
286: <TR VALIGN="TOP"><TD><a name="chop">chop</a> (vi 3b)</TD>
287: <TD>change with or as if with the wind</TD></TR>
288: <TR VALIGN="TOP"><TD><a name="chch">chop and change</a> (vi 2)</TD>
289: <TD>change esp. pointlessly or capriciously</TD></TR>
290: <TR VALIGN="TOP"><TD><a name="coalify">coalify</a> (vb)</TD>
291: <TD>change into coal by the process of coalification</TD></TR>
292: <TR VALIGN="TOP"><TD><a name="comeover">come over</a> (vi la)</TD>
293: <TD>change from one side (as of a controversy) to the other</TD></TR>
294: <TR VALIGN="TOP"><TD><a name="comeround">come round</a> (vi 2)</TD>
295: <TD>change in direction or opinion</TD></TR>
296: <TR VALIGN="TOP"><TD><a name="curdle">curdle</a> (vi 1)</TD>
297: <TD>change into curd</TD></TR>
298: <TR VALIGN="TOP"><TD><a name="cut">cut</a> (vi 3g)</TD>
299: <TD>change in direction</TD></TR>
300: <TR VALIGN="TOP"><TD><a name="deform">deform</a> (vi)</TD>
301: <TD>change in shape </TD></TR>
302: <TR VALIGN="TOP"><TD><a name="devitrify">devitrify</a> (vi)</TD>
303: <TD>change from a vitreous to a crystalline condition usu. with loss of
304: transparency and luster</TD></TR>
305: <TR VALIGN="TOP"><TD><a name="differ">differ</a> (vi lb)</TD>
306: <TD>change from time to time or from one instance or occasion to another</TD></TR>
307: <TR VALIGN="TOP"><TD><a name="diphthongize">diphthongize</a> (vi)</TD>
308: <TD><U>of a simple vowel</U> change into a diphthong</TD></TR>
309: <TR VALIGN="TOP"><TD><a name="effloresce">effloresce</a> (vi 2a)</TD>
310: <TD><U>chem</U> change on the surface or throughout to a whitish mealy or
311: crystalline powder from the loss of water of crystallization on exposure
312: to the air</TD></TR>
313: <TR VALIGN="TOP"><TD><a name="fade">fade</a> (vi 6a)</TD>
314: <TD>change gradually in loudness or visibility - used of a motion-picture
315: image or of an electronics signal or image and usu. with <U>out</U> to specify
316: change from loud to soft or bright to dark and with <U>in</U> to specify change
317: from soft to loud or dark to bright</TD></TR>
318: <TR VALIGN="TOP"><TD><a name="flash">flash</a> (vi 8)</TD>
319: <TD><U>of a liquid</U> change suddenly or violently into vapor</TD></TR>
320: <TR VALIGN="TOP"><TD><a name="flop">flop</a> (vi 3)</TD>
321: <TD>change suddenly (as from one course to another)</TD></TR>
322: <TR VALIGN="TOP"><TD><a name="follow">follow</a> (vt 4b)</TD>
323: <TD>change in constant relation to</TD></TR>
324: <TR VALIGN="TOP"><TD><a name="gel">gel</a> (vi)</TD>
325: <TD>change into a gel</TD></TR>
326: <TR VALIGN="TOP"><TD><a name="gelatinize">gelatinize</a> (vi)</TD>
327: <TD>change into a jelly</TD></TR>
328: <TR VALIGN="TOP"><TD><a name="graduate">graduate</a> (vi 2)</TD>
329: <TD>change gradually</TD></TR>
330: <TR VALIGN="TOP"><TD><a name="hold">hold</a> (vi lb(1))</TD>
331: <TD>not change</TD></TR>
332: <TR VALIGN="TOP"><TD><a name="melt">melt</a> (vi la)</TD>
333: <TD>change from a solid to a liquid state usu. by the action of heat</TD></TR>
334: <TR VALIGN="TOP"><TD><a name="push">push</a> (vi 5b)</TD>
335: <TD>change in quantity or extent</TD></TR>
336: <TR VALIGN="TOP"><TD><a name="quarter">quarter</a> (vi 4)</TD>
337: <TD>change from one quarter to another - used of the moon</TD></TR>
338: <TR VALIGN="TOP"><TD><a name="range">range</a> (vi 6)</TD>
339: <TD>change within limits</TD></TR>
340: <TR VALIGN="TOP"><TD><a name="reform">reform</a> (vi)</TD>
341: <TD>change for the better</TD></TR>
342: <TR VALIGN="TOP"><TD><a name="resinify">resinify</a> (vi 1)</TD>
343: <TD>change into a resin</TD></TR>
344: <TR VALIGN="TOP"><TD><a name="rote">rote</a> (vi)</TD>
345: <TD>change by rotation</TD></TR>
346: <TR VALIGN="TOP"><TD><a name="run">run</a> (vi 11b)</TD>
347: <TD>change to a liquid state</TD></TR>
348: <TR VALIGN="TOP"><TD><a name="runinto">run into</a> (vt la)</TD>
349: <TD>change into</TD></TR>
350: <TR VALIGN="TOP"><TD><a name="solate">solate</a> (vi)</TD>
351: <TD>change to a sol</TD></TR>
352: <TR VALIGN="TOP"><TD><a name="specialize">specialize</a> (vi 3)</TD>
353: <TD>change adaptively</TD></TR>
354: <TR VALIGN="TOP"><TD><a name="transfer">transfer</a> (vi 2)</TD>
355: <TD>change from one vehicle or transportation line to another</TD></TR>
356: <TR VALIGN="TOP"><TD><a name="transform">transform</a> (vi)</TD>
357: <TD>CHANGE</TD></TR>
358: <TR VALIGN="TOP"><TD><a name="transship">transship</a> (vi)</TD>
359: <TD>change from one ship or conveyance to another</TD></TR>
360: <TR VALIGN="TOP"><TD><a name="turn3b1">turn (vi 3b(1))</a></TD>
361: <TD>change from ebb to flow or flow to ebb</TD></TR>
362: <TR VALIGN="TOP"><TD><a name="turn4c1">turn (vi 4c(l)</a>)</TD>
363: <TD>change from submission or friendliness to resistance or opposition -
364: usu. used with <U>against</U></TD></TR>
365: <TR VALIGN="TOP"><TD><a name="turn6b1">turn (vi 6b(1))</a></TD>
366: <TD>CHANGE - used with <U>into</U> or <U>to</U></TD></TR>
367: <TR VALIGN="TOP"><TD><a name="turn6b2">turn (vi 6b(2)</a>)</TD>
368: <TD>change to</TD></TR>
369: <TR VALIGN="TOP"><TD><a name="turnoff">turn off</a> (vi 2b)</TD>
370: <TD>change to a specified state</TD></TR>
371: <TR VALIGN="TOP"><TD><a name="waver">waver</a> (vi lb)</TD>
372: <TD>change between objects, conditions, uses, or otherwise</TD></TR>
373: <TR VALIGN="TOP"><TD><a name="weaken">weaken</a> (vi 2)</TD>
374: <TD>change from a complex to a simple sound (as from a diphthong to a
375: long vowel)</TD></TR>
376: <TR VALIGN="TOP"><TD> </TD>
377: <TD>change from a strong to a weak sound</TD></TR>
378: <TR VALIGN="TOP"><TD> </TD>
379: <TD>change from an open to a close vowel</TD></TR>
380: <TR VALIGN="TOP"><TD><a name="whiffle">whiffle</a> (vi lb(2))</TD>
381: <TD>change from one course or opinion to another as if blown by the wind</TD></TR></TABLE>
382: </CENTER>
383:
384: <BR WP="BR1"><BR WP="BR2">
385: <P>4.1 <U>Syntactic Rules and Usage Notes</U></P>
386:
387: <P>The first requirement for a sense selection network is that it should contain all the meanings of each set of homographs. For
388: example, the SSN for "change" should contain all its noun and verb definitions. (Since a dictionary may contain several
389: homographs in the same part of speech, e.g. "bore" has three distinct verb entries, all the definitions of each homograph
390: would have to be combined into one SSN.) The first task of a parser, then, is to identify the correct part of speech for each
391: word encountered in an utterance; in so doing, of course, the parser may have to deal with inflected forms of a word.</P>
392:
393: <P>To some extent, syntactic parsing may permit further discrimination in the SSN. In fact, it may eventually be possible to
394: group many definitions of a word according to the patterns of the syntactic context in which they can occur. This notion was
395: previously explored with some success (see <A HREF="#earl">Earl</A> (1973) for details and other references) under the rubric of "word
396: government." The extent to which this notion can be used for sense discrimination can be determined only after each SSN is
397: elaborated, i.e. only after determining how much sense discrimination must rely on semantic considerations. Clearly, if
398: syntactic parsing can do the job, a computer system will be much more efficient.</P>
399:
400: <P>Certainly, in the case of transitive and intransitive verbs or verbs which use particles, syntactic parsing will be very useful in
401: traversing the SSN. Many verbs have both transitive and intransitive definitions; for such verbs, answering the question,
402: through syntactic parsing, whether the verb has an object can provide one branching node in its SSN. There is also a large
403: number of transitive verbs in the dictionary with definitions which specify the object that must be present for them to be the
404: applicable sense. For example, "bail" has one sense, viz., "to clear (water) from a boat by dipping and throwing over the
405: side," which requires the object to be the word "water." In many cases, the object is specified generically, e.g. two senses of
406: "abandon" specify the object as "oneself," indicating that the object must be a reflexive pronoun; another sense of "bail"
407: specifies the object as "personal property," indicating that the object must satisfy this selectional restriction. (The nature and
408: treatment of selectional restrictions is discussed in section 4.4.) In these cases, questions about the object during syntactic
409: parsing can provide additional branching nodes in the SSN.</P>
410:
411: <P>Another significant class of definitions that may be recognized syntactically arises from verbs which take an adjective
412: complement. The applicable definitions of such verbs always end with the phrase "to be;" e.g. one sense of "feel" is defined
413: by "perceive oneself to be." Thus, for verbs with this type of definition (or for verbs defined by a verb which takes an
414: adjective complement), a syntactic question regarding the presence of an adjective complement can provide a branching
415: node.</P>
416:
417: <P>In <A HREF="#w3">W3</A>, many verb definitions have accompanying usage notes, which provide information about the use of the verb being
418: defined, usually in the form of a comment on idiom, syntax, semantic relationship, status, or various other matters. Of
419: interest here are those usage notes which identify a particular idiom in which the particular sense of the verb is used, an
420: accompanying particle (such as "up" or "out"), or an accompanying prepositional phrase. For example, 520 of the 788
421: senses of the verb "take," all of which would be included in a complete SSN for this verb, involve some peculiarity of usage
422: identified in the dictionary. Four senses of "change," labeled <a href="#def2a">2a</a>, <a href="#def2b">2b</a>, and <a href="#def4b">4b</a> in Table 1, have usage notes; three definitions in
423: which "change" is used (as shown in Table 2: <a href="#fade">fade</a>, <a href="#turn4c1">turn (vi 4c(1))</a>, and <a href="#turn6b1">turn (vi 6b(1)))</a> also have usage notes. The comments
424: made in these usage notes can be used to formulate branching questions for an SSN, although not as directly as perhaps
425: would be desired. These usage conditions do not specify that the presence of the idiom, particle, or preposition indicates the
426: applicability of the definition, but only that the absence of the condition indicates the nonapplicability of the definition.</P>
427:
428: <P>4.2 <U>Preposition Definitions</U></P>
429:
430: <P>Before continuing with the description of how to build a sense selection network, it is necessary to digress into a discussion
431: of prepositional definitions, since they will play a crucial role in attempting to develop semantic representations of
432: definitions. A preposition is defined as "a linguistic form that combines with a noun, pronoun, or noun equivalent to form a
433: phrase that typically has an adverbial, adjectival, or substantival relation to some other word." Prepositions are few in
434: number (I have identified 126 in <A HREF="#w3">W3</A>, half of which are phrases), but rich in significance for text processing, where they are
435: typically used to identify conceptual cases. However, from my examination of preposition definitions, I do not believe their
436: significance has been fully exploited.</P>
437:
438: <P><A HREF="#ben">Bennett</A> (1975) asserted that spatial and temporal prepositions (a high percentage of all prepositions) lead to 23 primitive
439: conceptual cases, even though in W3 the number of their definitions is at least two orders of magnitude higher The
440: difference seems to lie in the "apparent polysemy" which, as Bennett says, arises from the inclusion in prepositional
441: definitions of "redundant features already determined by the environment." In other words, many preposition definitions
442: contain information about the context surrounding the preposition. I believe such "redundancy" can be exp]oited in
443: developing a semantic parser which will have a much greater facility for the type of conceptual case resolution that Small is
444: concerned with.</P>
445:
446: <P>Like verbs, prepositions appear to form a closed system in which they are defined in terms of other prepositions. However,
447: unlike verbs, their primitives appear to be more easily identified. Of approximately 1400 definitions in W3, 70 percent are
448: defined in terms of other prepositions, 20 percent are defined only by usage notes, and 10 percent are defined by verb forms.
449: The usage note definitions, which have the appearance of primitives, uniformly begin with the phrase "used as a function
450: word to indicate." It is what follows the word "indicate" that can be used in developing the parser.</P>
451:
452: <P>As mentioned above, in its definition, a preposition forms a relation between its object and some other word. The nature of
453: this relation is what follows the word "indicate" in the usage notes. What I have found is that such relations follow certain
454: patterns which can be articulated in formal recognition rules (1) for inclusion in a semantic parser, (2) for developing a
455: semantic representation of verb definitions, and (3) for determining how to drive the parser. Usage note definitions of
456: prepositions may specify:</P>
457:
458: <UL>
459: <LI>a condition (i.e. selectional restriction) that the object of the preposition must satisfy (e.g. that it is an age, a time, a state,
460: or a group);</LI>
461:
462: <BR WP="BR1"><BR WP="BR2"> <LI>a condition that must be satisfied by the context surrounding the prepositional phrase (e.g. the presence of an action verb,
463: a specific type of action, something that is enveloped or covered);</LI>
464:
465: <BR WP="BR1"><BR WP="BR2"> <LI>the assignment of a characterization to the prepositional object (e.g. that it is a location, an instrument, a purpose or goal,
466: or a result (i.e. the traditional cases) or that the object is a thing observed as a spectator or a quantity of movement); and</LI>
467:
468: <BR WP="BR1"><BR WP="BR2"> <LI>the assignment of a characterization to some element of the context outside the prepositional phrase (e.g. the fact of
469: being present, of having parts or elements, or of being insertable into something else).</LI>
470:
471: <BR WP="BR1"><BR WP="BR2"></UL>
472:
473: <P>Every such definition does not contain all these specifications; at this time, I have not attempted an analysis of these
474: definitions into their components. However, examples of how I have used these notions are described in subsequent
475: sections.</P>
476:
477: <P>At this point, I will only make some general observations about what these definitions imply with respect to parsing and
478: semantic representation. In the first place, it appears that many words can be typecast with particular prepositional
479: definitions, e.g. some verbs can be characterized as governing the patterns embodied in certain prepositional definitions.
480: Such patterns are descernible from the definitions of such verbs.</P>
481:
482: <P>Furthermore, it should be clear from the four types of specifications mentioned above how these definitions can be
483: integrated into a general parser both for performing the parse and for building a semantic representation of what is being
484: parsed. By extension, these same considerations mean that it is possible to use specific prepositional definitions in parsing
485: verb definitions and in creating the semantic representations of those definitions either explicitly or in terms of frames with
486: slots accompanied by selectional restrictions that must be satisfied when a word is used in a particular utterance. These
487: issues are dealt with in more detail below, along with specific examples.</P>
488:
489: <P>4.3 <U>Predicates, Slots, and Selectional Restrictions</U></P>
490:
491: <P>The syntactic considerations described in section 4.2 clearly will not suffice for the construction of complete SSNs. Their
492: further elaboration requires the development of semantic questions for use at the branching nodes. To develop such
493: questions, it is necessary to know the semantic representations of the senses as they will appear at the terminal nodes of the
494: SSN. (This is a circular statement, since it is necessary to know how to discriminate among the senses before distinct
495: representations can be rendered. Therefore, the full elaboration of an SSN involves a process of iterative refinement of the
496: discriminatory and representational components. Moreover, accurate representations must eventually be given in terms of
497: primitive units; any intermediate representations must therefore be considered in this light.) In the following discussion, only
498: the development of representations for verbs will be considered, although, as will be seen, the reresentation of other parts of
499: speech is inextricably involved.</P>
500:
501: <P>The representation of a verb definition essentially involves the assignment or identification of (1) an appropriate predicate,
502: (2) the appropriate arguments or slots, and (3) selectional restrictions (if any) for each slot. The predicate and arguments
503: would be arrayed as an n-tuple, with the selectional restrictions placed in the appropriate slots. The representation of
504: particular definitions may involve a logical combination of more than one such n-tuple.</P>
505:
506: <P>In Rieger's system, a predicate is considered a label for the accompanying argument configuration, but it has no intrinsic
507: meaning. This is a convenient starting point for assigning a predicate, but in the case of analytic definitions (which consist of
508: a genus and differentiae), the predicate should be the ultimate generic term. For example, the definitions shown in <a href="#table2">Table 2</a>
509: can be assigned the predicate "change." This is more than just a label, but rather can be used to indicate that the basic
510: argument configuration and selectional restrictions for the particular definition come from the definitions for "change." This
511: is discussed in more detail below. (However, for the verb "change," the predicate "become different" will be used.)</P>
512:
513: <P>The argument configuration must be developed from an analysis of the definitions and usually requires an examination of the
514: definitions of the constituent words. To illustrate this process, definitions 1 and 2 of "change" will be used. For both
515: definitions, the first argument or slot will be used to indicate the subject of the verb; since the subject may be in the PAT or
516: AGT case (to be determined by the context), the corresponding slot for SUBJ will indicate that (PAT v AGT) is to be
517: assigned to SUBJ. The words "become different" in both definitions imply the presence of four slots: FROM-STATE,
518: TO-STATE, TIME1 and TIME2. However, since "different" is modified in two ways, some additional complexity is
519: introduced. In <a href="#def1">definition 1</a>, there is the notion that only an accidental attribute of the (PAT v AGT) "becomes different,"
520: while, in <a href="#def2">definition 2</a>, there is the notion that some essential attribute "becomes different," with the result that the (PAT v
521: AGT) no longer exists. The net effect of this distinction is that for definition 1, there must be a "FROM-STATE," a
522: "TO-STATE," and a "RESPECT" in which the change occurs, while for definition 2, there must be a "FROM-STATE"
523: (which in this case is the SUBJ of "change") and a "TO-STATE" which is the "RESULT" of the change. Possible semantic
524: representations of these two definitions are shown in Figure 1.</P>
525:
526: <a name="figure1"></a><TABLE BORDER="1" WIDTH="100%" CELLPADDING="1" CELLSPACING="1">
527: <TR VALIGN="TOP" align="center"><TD>Figure 1</TD></TR>
528: <TR VALIGN="TOP" align="center"><TD>Basic Frames for Definitions 1 and 2 of "change"</TD></TR>
529: <TR VALIGN="TOP"><TD><U>Definition 1</U>: become different in one or more respects without becoming something else
530: <pre>
531: [BECOME DIFFERENT (FROM-STATE NE. TO-STATE)
532: ((SUBJ) (PAT v ACT)
533: (* D * (PAT v ACT)
534: (ESSENTIAL ATTRIBUTES ...)
535: (ACCIDENTAL ATTRIBUTES ...)
536: ("RESPECT" (TIME1) (FROM-STATE ...)
537: (TIME2) (TO-STATE ...))))]
538: </pre>
539: <TR VALIGN="TOP"><TD><U>Definition 2</U>: become something materially different from before
540: <pre>
541: [BECOME DIFFERENT (FROM-STATE NE. TO-STATE)
542: ((SUBJ) (PAT v ACT) (TIME1) (FROM-STATE)
543: (* D * (PAT v ACT) ...))
544: (("RESULT") (TIME2) (TO-STATE)
545: (* D * ("RESULT" ...))]
546: </pre>
547: </TD></TR></TABLE>
548:
549: <P>These representations could appear at terminal nodes of an SSN and one could be the contribution made as a result of
550: parsing the verb "change," unless further analysis were to lead to one of the subsenses of these definitions.</P>
551:
552: <P>The final aspect of representing a verb definition requires the incorporation of selectional restrictions into the
553: representations. As with the predicate and the argument configuration, selectional restrictions on what can fill particular
554: arguments are derived from the definitional matter. As mentioned before, whatever representational formalism is used must
555: have a capability for identifying the selectional restrictions that must be satisfied by the context. For the subsenses of
556: definition 1 of "change," the selectional restrictions may be so detailed as to fill in some slots of the basic frame for definition
557: 1 or lead to the necessity for additional slots. As a result, using the terminology of <A HREF="#nor">Norman</A> et al. (1975), the concept
558: satisfying an argument of the basic frame may be completely determined in representing a subsense.</P>
559:
560: <P>For the most part, the subsenses of <a href="#def1">definition 1</a> follow the basic frame shown in <a href="#figure1">Figure 1</a> by providing information about the
561: "respect" in which the subject of the verb "becomes different." To determine that this is the case, it was first necessary to
562: examine the definitions of the main verbs of each subsense. In each instance, the examination showed that the notion of
563: "becoming different" is part of the meaning of the verb. Having arrived at this finding, it was then determined that most of
564: the remaining information in the subsenses pertains to the "respect" in which the change occurs. These "respects" are shown
565: in Table 3 for each subsense and would be used to replace the word "RESPECT" in the basic frame for definition 1. It
566: should be noted that, for subsenses <a href="#def1b2">lb(2)</a>, <a href="#def1c">1c</a>, <a href="#def1f">1f</a>, part of <a href="#def1g">1g</a>, part of <a href="#def1h">1h</a>, and <a href="#def1i">1i</a>, it was necessary to search for the "respect" in
567: the definitions of the subsense's constituents. (It should be added that it was this analysis of the subsenses that led to the
568: placement of the "RESPECT" slot under the slot for "ACCIDENTAL ATTRIBUTES" which in turn modifies the subject of
569: the verb. Each "respect" in which the change could occur was required, via the phrase "without becoming something else,"
570: not to change the essential nature of the subject.)</p>
571: <CENTER>
572: <TABLE BORDER="1" ALIGN="CENTER" WIDTH="100%">
573: <TR VALIGN="TOP"><TD COLSPAN="2" ALIGN="CENTER" WIDTH="141">Table 3</TD></TR>
574: <TR VALIGN="TOP"><TD COLSPAN="2" ALIGN="CENTER">Selectional Restrictions on "RESPECT" Slot of the Frame for Definition 1 of
575: "change"</TD></TR>
576: <TR VALIGN="TOP"><TD ALIGN="CENTER">Subsense</TD>
577: <TD ALIGN="CENTER" WIDTH="369">Selectional Restrictions</TD></TR>
578: <TR VALIGN="TOP"><TD><a href="#def1a">1a</a></TD>
579: <TD>characteristic, property, or tendency</TD></TR>
580: <TR VALIGN="TOP"><TD><a href="#def1b1">1b(1)</a></TD>
581: <TD>form, appearance, position, state, or stage </TD></TR>
582: <TR VALIGN="TOP"><TD><a href="#def1b2">1b(2)</a></TD>
583: <TD>facial complexion</TD></TR>
584: <TR VALIGN="TOP"><TD><a href="#def1c">1c</a></TD>
585: <TD>size, quantity, number, degree, value, intensity, power,
586: authority, reputation, wealth, amount, strength, etc.</TD></TR>
587: <TR VALIGN="TOP"><TD><a href="#def1d">1d</a></TD>
588: <TD>customs, methods, or attitudes <U>specif</U> religious attitudes</TD></TR>
589: <TR VALIGN="TOP"><TD><a href="#def1e">1e</a></TD>
590: <TD>phase of the moon</TD></TR>
591: <TR VALIGN="TOP"><TD><a href="#def1f">1f</a></TD>
592: <TD>capacity of being sour (e.g. disposition, taste, smell,
593: acidity)</TD></TR>
594: <TR VALIGN="TOP"><TD> </TD>
595: <TD>capacity of being tainted (e.g. subject to putrefaction,
596: corruption, moral contamination)</TD></TR>
597: <TR VALIGN="TOP"><TD><a href="#def1g">1g</a></TD>
598: <TD>means of conveyance</TD></TR>
599: <TR VALIGN="TOP"><TD> </TD>
600: <TD>vehicle or transportation line being used</TD></TR>
601: <TR VALIGN="TOP"><TD><a href="#def1h">1h</a></TD>
602: <TD>register of the voice</TD></TR>
603: <TR VALIGN="TOP"><TD> </TD>
604: <TD>voice's tone, pitch, or intensity</TD></TR>
605: <TR VALIGN="TOP"><TD><a href="#def1i">1i</a></TD>
606: <TD>method, tempo, or approach</TD></TR></TABLE>
607: </CENTER>
608: </P>
609: <BR WP="BR1"><BR WP="BR2">
610: <P>The subsenses may also provide further selectional restrictions about the direction of the change. These restrictions, as
611: shown in Table 4, would be added to the "FROM-STATE," the "TO-STATE," or as a relation between the two states.
612: Other information may add new arguments (as in definition <a href="#def1e">1e</a>) or give values to other slots (as in definitions <a href="#def1e">1e</a> and <a href="#def1h">1h</a>).</P>
613:
614: <P>Identification of the predicate, the argument pattern, and the selectional restrictions for all definitions in itself requires a
615: sophisticated semantic parser. Identification of the predicate can be accomplished in part by a taxonomic analysis of the type
616: proposed by <A HREF="#leh">Lehmann</A> (1976); for example, all the definitions in <a href="#table2">Table 2</a>, the intransitive uses of "change," could be assigned
617: the predicate "CHANGE." However, this is not valid, since ultimately the definitions in Table 2 should be assigned the
618: predicate "BECOME DIFFERENT" or whatever primitive turns out to be appropriate.</p>
619: <CENTER>
620: <a name="table4"></a><TABLE BORDER="1" ALIGN="CENTER" WIDTH="100%">
621: <TR VALIGN="TOP"><TD COLSPAN="2" ALIGN="CENTER" WIDTH="147">Table 4</TD></TR>
622: <TR VALIGN="TOP"><TD COLSPAN="2" ALIGN="CENTER">Other Selectional Restrictions of the Frame for Definition 1 of "change"</TD></TR>
623: <TR VALIGN="TOP"><TD ALIGN="CENTER">Subsense</TD>
624: <TD ALIGN="CENTER" WIDTH="351">Selectional Restriction</TD></TR>
625: <TR VALIGN="TOP"><TD>1a</TD>
626: <TD>becomes deprived of ("lose")</TD></TR>
627: <TR VALIGN="TOP"><TD> </TD>
628: <TD>comes to have ("acquire")</TD></TR>
629: <TR VALIGN="TOP"><TD>1b(2)</TD>
630: <TD>becomes red ("blush")</TD></TR>
631: <TR VALIGN="TOP"><TD> </TD>
632: <TD>becomes deprived of color or luster ("pale")</TD></TR>
633: <TR VALIGN="TOP"><TD>1c</TD>
634: <TD>becomes diminished ("decrease")</TD></TR>
635: <TR VALIGN="TOP"><TD> </TD>
636: <TD>becomes greater ("increase")</TD></TR>
637: <TR VALIGN="TOP"><TD>1e</TD>
638: <TD>SUBJ = moon</TD></TR>
639: <TR VALIGN="TOP"><TD> </TD>
640: <TD>(TIMEx) (THROUGH-STATE = new moon)</TD></TR>
641: <TR VALIGN="TOP"><TD>1h</TD>
642: <TD>SUBJ = voice</TD></TR></TABLE>
643: </CENTER>
644: </P>
645: <BR WP="BR1"><BR WP="BR2">
646: <P>In part, identification of the argument pattern can also be accomplished by taking advantage of the taxonomic relationships.
647: Thus, argument patterns for the verbs in whose definitions "change" is the main verb can be used as the starting point for
648: identifying the slots necessary to represent the definitions of "change." <A HREF="#leh">Lehmann</A> (1976) proposed to identify case argument
649: patterns based on an analysis of the uses of "high-level verbs;" for "move," 17 cases were identified in this way. However,
650: this approach does not ensure that the definitions of "move" will be represented nor does it identify the obligatory arguments
651: necessary to discriminate among the senses of "move."</P>
652:
653: <P>The development of procedures for identifying arguments and selectional restrictions ultimately must rely on the observation
654: of patterns and the development of procedures for recognizing those patterns. As noted in the previous section, definitions
655: of prepositions will play a significant role in the development of such recognition rules. For example, the phrase "in one or
656: more respects" in <a href="#def1">definition 1</a> of "change," combined with the fact that one definition of "in" is "with reference to," could
657: lead to the recognition rule that, whenever "change" (or any verb derived from it) is used in conjunction with a prepositional
658: phrase beginning with "in," the object of the preposition should replace the word "RESPECT" in the basic frame for
659: definition 1 of "change."</P>
660:
661: <P>Although the development of such rules is not the goal of my research, it has become clear that they are necessary for
662: developing sense selection networks. They are also necessary in identifying derivational paths within a dictionary in the
663: search for primitive verb concepts.</P>
664:
665: <P>4.4 <U>Building Nets and Driving the Parser</U></P>
666:
667: <P>The development of sense selection networks which incorporate semantic considerations should follow directly from
668: whatever semantic representations of the senses have been created. Essentially, each semantic representation would be a
669: terminal node of the SSN. However, because the set of definitions may have some internal ordering corresponding to
670: subsenses, it may be necessary to permit nonterminal nodes of the network to provide a distinct sense of the word being
671: parsed when it is not possible to reach a terminal node. Since an SSN is essentially an ordered set of questions leading to a
672: terminal node (or one that has an attached semantic representation), its development consists of identifying the questions and
673: putting them into the order in which the search through the net should be conducted. Although it would seem that the SSN
674: for each word will have to be developed on its own, the semantic representations themselves may possibly be used to
675: identify the questions and to determine the order of the search.</P>
676:
677: <P>Each semantic representation (as it should be developed, not necessarily as shown in <a href="#figure1">Figure 1</a>) is an ordered n-tuple, perhaps
678: with several sublists, which should contain all syntactic, contextual, and semantic information about the sense represented.
679: The first elements of the n-tuple could contain syntactic information (e.g. noun or verb), followed perhaps by contextual
680: information, and finally semantic information. The necessity of questions and branching nodes could conceivably be
681: recognized by comparing and contrasting the n-tuples corresponding to each sense to find the first differences between them.</P>
682:
683: <P>In the example of "change," the first difference would be whether the semantic representation corresponds to a noun or verb;
684: a second difference for the set of verb definitions would be whether the sense is transitive or intransitive, and so on down to
685: the representations of the subsenses where the difference would be, for definition 1, the "RESPECT" of the change. Thus,
686: the differences between the sense representations would identify where the branching nodes should be placed and could be
687: used to develop the questions that should correspond to each branch. Although I have not attempted to follow this
688: procedure, I believe that experience in analyzing such differences will make it possible to develop rules for identifying
689: questions to be used in the SSN.</P>
690:
691: <P>The questions that are developed for searching through an SSN will be essentially the same as those used in an ATN parser,
692: except that it will be necessary to add semantic paths to the syntactic ones. One difference will be that, instead of developing
693: an <EM>a priori</EM> model of the semantic grammar within which to conduct the search, each node will have to contain information
694: which will tell the parser what to look for next. For example, the fact that the subject of "change" may be in the PAT or
695: AGT case would require the parser to make particular searches in the context. Thus, in the use of "change" in defining
696: "<a href="#coalify">coalify</a>" (see Table 2), the agent of the change is specified as "the process of coalification," hence relegating the subject of
697: "coalify" to the PAT case. It is possible to conceive of a standard set of procedures for making such a search (see, for
698: example, <A HREF="#cha">Chafe</A> (1970, pp. 243-244)).</P>
699:
700: <P>When the question at a branching node requires a semantic resolution, the parsing requirements may be quite complex.
701: Some of the difficulties that may arise can be illustrated by considering how a search would try to identify the appropriate
702: sense of "change" in definitions where it is used as the main verb, i.e. those shown in <a href="#table2">Table 2</a>. When "change" is used in
703: defining another verb, three things may happen: (1) an argument of the frame for "change" may be given a value, (2) some
704: further selectional restrictions may be added to an argument slot, or (3) an additional slot may be created. These possibilities
705: may or may not help identify the applicable sense.</P>
706:
707: <P>Since all the definitions in Table 2 use "change" as the main verb, the first thing to examine is whether the contextual matter
708: of the definition provides a value that fills a slot in the basic frames for "change." The basic difference between definitions <a href="#def1">1</a>
709: and <a href="#def2">2</a> of "change" is that, in the latter instance, the subject of "change" is the "FROM-STATE" while the "TO-STATE" is
710: the object of a preposition, usually "into" or "to." Therefore, if there is an "into" in the context, there might be a presumption
711: that definition 2 is the applicable sense, such as in the definitions for "<a href="#become">become</a>," "<a href="#coalify">coalify</a>," "<a href="#curdle">curdle</a>," "<a href="#diphthongize">diphthongize</a>," "<a href="#flash">flash</a>,"
712: "<a href="#gel">gel</a>," "<a href="#gel">gelatinize</a>," "<a href="#resinify">resinify</a>," and "<a href="#turn6b1">turn (vi 6b(l))</a>." However, even in these instances, it would be necessary to compare the
713: subject and the object of "into" to determine if there has been an essential change in nature. For example, in the definition of
714: "coalify," the "process of coalification" is one in which (from the definition of "coalification") "vegetable matter" undergoes
715: a change "into coal;" this supports the choice of definition 2. On the other hand, the definition of "<a href="#caramelize">caramelize</a>" seems to lead
716: to the possibility that both definitions 1 and 2 are applicable because of the disjunction in the object of "to."</P>
717:
718: <P>In many of the definitions, the use of the verb "change" bears the same relation to definition 1 of "change" as its sub-senses,
719: i.e. these definitions (shown in Table 5) indicate more specifically the "respect" in which the change occurs. For these
720: definitions, it would be necessary to determine which subsenses would be applicable by comparing the "respect" indicated in
721: Table 5 to the "respect" indicated in each subsense. For those definitions which contain both "from" and "to" prepositional
722: phrases, such as those for "<a href="#comeover">come over</a>," "<a href="#devitrify">devitrify</a>," "<a href="#differ">differ</a>," "<a href="#melt">melt</a>," "<a href="#quarter">quarter</a>," "<a href="#transfer">transfer</a>," "<a href="#transship">transship</a>," and "<a href="#weaken">weaken</a>," the
723: inference can be made that definition 1 is intended, although further analysis would be necessary to determine which
724: "respect" and hence which subsense is intended.</p>
725: <CENTER>
726: <TABLE BORDER="1" ALIGN="CENTER" WIDTH="100%">
727: <TR VALIGN="TOP"><TD COLSPAN="2" ALIGN="CENTER" WIDTH="204">Table 5</TD></TR>
728: <TR VALIGN="TOP"><TD COLSPAN="2" ALIGN="CENTER">Selectional Restrictions on "RESPECT" (Slot Added in Uses of
729: "change")</TD></TR>
730: <TR VALIGN="TOP"><TD ALIGN="CENTER">Definition</TD>
731: <TD ALIGN="CENTER" WIDTH="248">Selectional Restriction</TD></TR>
732: <TR VALIGN="TOP"><TD><a href="#break5c">break (vi 5c)</a></TD>
733: <TD>purport, mood, or attitude</TD></TR>
734: <TR VALIGN="TOP"><TD><a href="#break6b">break (vi 6b)</a></TD>
735: <TD>line or set</TD></TR>
736: <TR VALIGN="TOP"><TD><a href="#comeround">come round (vi 2)</a></TD>
737: <TD>direction or opinion</TD></TR>
738: <TR VALIGN="TOP"><TD><a href="#cut">cut (vi 3g)</a></TD>
739: <TD>direction</TD></TR>
740: <TR VALIGN="TOP"><TD><a href="#deform">deform (vi)</a></TD>
741: <TD>shape</TD></TR>
742: <TR VALIGN="TOP"><TD><a href="#fade">fade (vi 6a)</a></TD>
743: <TD>loudness or visibility</TD></TR>
744: <TR VALIGN="TOP"><TD><a href="#push">push (vi 5b)</a></TD>
745: <TD>quantity or extent</TD></TR></TABLE>
746: </CENTER>
747: </P>
748: <BR WP="BR1"><BR WP="BR2">
749: <P>However, for many of the definitions, such as those for "<a href="#chch">chop and change</a>," "<a href="#graduate">graduate</a>," "<a href="#hold">hold</a>," and "<a href="#specialize">specialize</a>,"
750: identification of the appropriate sense of "change" is not possible from the given context. In these situations, all definitions
751: of "change" might apply and it would be necessary to await their use with further context before sense selection can take
752: place.</P>
753:
754: <P>The crux of these comments is that sense selection based on semantic questions inherently involves further computational
755: analysis which delves into the definitions of the words used in context with "change," The dictionary shows the complexity
756: of such processing, but at the same time it shows the availability of large amounts of information that will aid this process.</P>
757:
758: <P>To generalize from the observations made about disambiguation of "change" from whatever context in which it is used, e.g.
759: determining the "respect" in which something "becomes different" or determining the fact that the patient of "change"
760: "becomes something materially different from before," we are first of all involved in what Small calls "conceptual case
761: resolution," although in a sense that may go beyond what he envisioned. We are no longer dealing solely with identifying
762: cases like "agent," "patient," and "experiencer" or concepts like "setting," "event," and "picture." We are also concerned
763: with whether concepts like "phase of the moon," "tendency," "complexion," "quantity," "strength," "attitude," or "taste"
764: have been invoked by a particular context. Thus, we can say that semantic representation requires a deeper level of
765: "conceptual case resolution" and an even greater level of complexity. We need to recognize that, when this is done,
766: conceptual relationships between and among the definitions are being delineated. Hence, through this process, we are
767: capturing the factual world knowledge which is embodied in definitions and thus providing some part of what Rieger finds
768: necessary to incorporate separately in his AI Systems.</P>
769:
770: <P>4.5 <U>Ambiguity</U></P>
771:
772: <P>It should be emphasized that a correct sense selection net does not automatically ensure that the correct sense will be
773: identified. The context may simply be ambiguous. This is particularly noticeable in dictionary definitions where sufficient
774: context for sense discrimination is seldom available. I suspect that the same is true for a large number of utterances,
775: particularly if the context is limited to a single sentence. Clearly, it is necessary that a minimum context be present in order
776: to permit discrimination.</P>
777:
778: <P>When faced with ambiguity, a text processing system could try to find the plausible interpretation using world knowledge or
779: dynamic expectancies of an inference model. Although such systems have been used with some success in limited domains,
780: there are two significant difficulties that may arise. The first is that insufficient world knowledge has been provided to the
781: system. (It may even be that the requisite world knowledge does not exist.) The second difficulty is that no inference model
782: can be an accurate representation of how everybody reasons. (I suspect that each person follows a unique inferencing
783: system.)</P>
784:
785: <P>Based on the description of sense selection networks which has been laid out in the preceeding sections, it would seem that
786: another avenue for dealing with these difficulties has opened up. Since each SSN is designed to ask questions, perhaps it
787: would be desirable simply to go as far as possible in interpreting and then send back the questions that currently stymie
788: further disambiguation. This is, after all, what we would hope that a listener would do if we say something that does not
789: make sense.</P>
790:
791: <P>Such questioning could have great intrinsic value, because it would indicate (1) an inadequacy in the parsing system itself,
792: (2) the lack of specific world knowledge (either in the system or in the world), or (3) the necessily for making an inference.
793: Given the state of our knowledge about representing utterances, I would prefer a system that simply tries to capture what is
794: present in an utterance. Accurate sense selection networks for relevant parts of the lexicon make it possible to build
795: representations of scripts out of the components that we have at hand, rather than attempting to develop <EM>a priori</EM> inference
796: models.</P>
797:
798: <P>4.6 <U>Movement Toward Primitives</U></P>
799:
800: <P>As previously pointed out, it will eventually be necessary that semantic representations at the terminal nodes of an SSN be
801: given in terms of primitives. It was also mentioned that the procedures described in this paper have been used as part of a
802: research effort designed to move toward identification of primitive verb concepts. The full set of procedures are described
803: more fully in section 9 of <A HREF="#lit">Litkowski</A> (1978), but it will be useful to describe how the notions described here are incorporated
804: in that effort.</P>
805:
806: <P>The basic procedure used in moving toward identification of primitives is through the development and application of rules
807: which establish that particular words and definitions cannot be primitive. This requires a showing that a word or definition is
808: derived from a more primitive concept and that a primitive cannot be derived from it. The nonprimitives are then set aside
809: and further efforts focus on those words and definitions not yet eliminated.</P>
810:
811: <P>The notions in this paper are applied by trying to show an explicit derivational relationship between two definitions. If the
812: specific sense of the main verb of a definition can be identified and it can be shown that the definition contains differentiae
813: which provide a value to an unbound argument in the semantic representation for the main verb, then the definition in which
814: the binding takes place can he characterized as nonprimitive. For example, the definition of "<a href="#diphthongize">diphthongize</a>" (shown in Table
815: 2) gives a value to the SUBJ of the basic frame for "change" and is thus inferred to be nonprimitive.</P>
816:
817: <P>This notion of filling a slot is used more generally by developing recognition rules that identify particular word government
818: patterns. For example, if a verb has a definition with the phrase "with an instrument" (thus creating an "instrument"
819: argument) and is used in defining another verb accompanied by a "with" prepositional phrase whose object is defined as an
820: instrument, the latter definition is characterized as nonprimitive. Some verbs in the first category are "apply," "fasten," "cut,"
821: and "beat;" an example in the second category is the verb "knife."</P>
822:
823: <P>Other verb definitIons are characterized as nonprimitive when they contain an optional component, i.e. one not necessary to
824: discriminate among the senses of its main verb. Recognition rules are needed to identify different realizations of such
825: components, such as the "manner" component which is optional or fills a slot for such verbs as "move," "act," "perform,"
826: "utter," "express," and "behave." Other definitions are characterized as nonprimitive when recognition rules establish that the
827: definition consists of at least two distinct verb concepts. This is true of aspect verbs such as "cause," "cease," "begin,"
828: "attempt," "refuse," and "serve." Verb definitions of this type are very similar to those characterized as lexical relations by
829: <A HREF="#evens">Evens and Smith</A> (1978) or lexical functions by <A HREF="#mel">Mel'cuk</A> (1978).</P>
830:
831: <P>If relations or functions are used in representing definitions, it is important to understand that the function and the argument
832: are parts of the lexicon, rather than the argument alone. In other words, we would have to do more than indicate that one
833: definition is derived from another in building a semantic representation of an utterance. We would also have to represent the
834: operator which gives rise to the derivation. For example, if we have the use of the "cause" operator, we would have to
835: provide slots in our semantic representation for all the kinds of infor-mation which should be associated with the use of
836: "cause." (See <A HREF="#by">Byerly</A> (1979) for such a detailed specification.) Therefore, whenever we say that one entry in a dictionary is
837: derived from another by the application of some operator, we must be prepared to bring the representational contribution of
838: the operator to the construction of the semantic representation of the derived entry.</P>
839:
840: <P>5. <U>MULTISENTENCE PARSING</U></P>
841:
842: <P>It seems that semantic representations of definitions in the form described must ultimately constitute the elements out of
843: which semantic representations of multisentence texts must be created, perhaps with two foci: (1) describing entities
844: (centered around nouns) and (2) describing events (centered around verbs). In parsing a single sentence, it seems clear that
845: open variables, i.e. unfilled slots, will remain. Many such slots can be filled by later processing and parsing. Thus, at least
846: part of multisentence text processing must recognize this fact, strip away the arbitrary bounds (which some would say are
847: only convenient breath stops) of periods, and build semantic representations that deal with the entities and events by
848: collapsing sentences which are used only to fill in some slots not yet filled. If multisentence texts can then be studied
849: empirically, the structure of ordinary discourse will then be based on observations rather than theory.</P>
850:
851: <P>6. <U>CONCLUSIONS</U></P>
852:
853: <P>Although the paradigm presented in this paper is complex, I believe that it is nothing more than what the lexicons of present
854: AI systems are becoming I believe that more rapid progress can be made with an explicit effort to exploit and not to
855: duplicate the efforts of lexicographers.</P>
856:
857: <P><U>REFERENCES</U></P>
858:
859: <P><A NAME="ben"></A>Bennett, D.C. (1975). <EM>Spatial and Temporal Uses of English Prepositions: An Essay in Stratificational Semantics</EM>,
860: Longman Linguistics Library, Vol. 17, Longman, New York.</P>
861:
862: <P><A NAME="bob"></A>Bobrow, D.G. and T. Winograd (1977). "An overview of KRL, a knowledge representation language," <EM>Cognitive Science</EM>,
863: Vol. 1, No. 1, pp. 3-46.</P>
864:
865: <P><A NAME="bol"></A>Bolinger, D. (1975). <EM>Aspects of Language</EM> 2nd ed., Harcourt Brace Jovanovich, Inc., New York.</P>
866:
867: <P><A NAME="by"></A>Byerly, H. (1979). "Substantial causes and nomic determination," <EM>Philosophy of Science</EM>, Vol. 46, No. 1, pp. 57-81.</P>
868:
869: <P><A NAME="cha"></A>Chafe, W.L. (1970). <EM>Meaning and the Structure of Language</EM>, University of Chicago Press, Chicago.</P>
870:
871: <P><A NAME="earl"></A>Earl, L.L. (1973). "Use of word government in resolving syntactic and semantic ambiguities," <EM>Information Storage and
872: Retrieval</EM> Vol. 9, pp. 639-664.</P>
873:
874: <P><A NAME="w3"></A>_____ (1966). <EM>Webster's Third New International Dictionary</EM>, Encyclopaedia Britannica, Chicago.</P>
875:
876: <P><A NAME="evens"></A>Evens, M.W. and R.N. Smith (1978). "A lexicon for a computer question-answering system," <EM>American Journal of
877: Computational Linguistics</EM>, Microfiche 83 and Microfiche 81, Frames 16-24.</P>
878:
879: <P><A NAME="leh"></A>Lehmann, W.P. and R.F. Simmons (1976). <EM>A Proposal to Develop a Computational Methodology for Deriving Natural
880: Language Semantic Structures via Analysis of Machine-Readable Dictionaries</EM>, University of Texas, Austin, Texas,
881: September 28.</P>
882:
883: <P><A NAME="lit"></A>Litkowski, K.C. (1978). "Models of the semantic structures of dictionaries," <EM>American Journal of Computational
884: Linguistics</EM>, Microfiche 81, Frames 25-74.</P>
885:
886: <P><A NAME="mel"></A>Mel'cuk, I.A. (1978). "A new kind of dictionary and its role as a core component of automatic text processing systems,"
887: <EM>T.A. Informations</EM>, No.2, pp.3-8.</P>
888:
889: <P><A NAME="nor"></A>Norman, D.A., D.E. Rumelhart, and the LNR Research Group (1975). <EM>Explorations in Cognition</EM>, W.H.Freeman, San
890: Francisco.</P>
891:
892: <P><A NAME="ol68"></A>Olney, J., C. Revard, and P. Ziff (1968). <EM>Toward the Development of Computational Aids for Obtaining a Formal Semantic
893: Description of English</EM>, SP-2766/001/00, System Development Corporation, Santa Monica, California, 1 October.</P>
894:
895: <P><A NAME="ol72"></A>Olney, J. and D. Ramsey (1972). "From machine-readable dictionaries to a lexicon tester: Progress, plans, and an offer,"
896: <EM>Computer Studies in the Humanities and Verbal Behavior</EM>, Vol.3, No.4, November, pp. 213-220.</P>
897:
898: <P><A NAME="rie"></A>Rieger, C. (1977). <EM>Viewing Parsing as Word Sense Discrimination</EM>, TP-511, Department of Computer Science, University
899: of Maryland, College Park, Maryland, January.</P>
900:
901: <P><A NAME="rs"></A>Rieger, C. and S. Small (1979). <EM>Word Expert Parsing</EM>, TR-734, Department of Computer Science, University of Maryland,
902: College Park, Maryland, March.</P>
903:
904: <P><A NAME="sim"></A>Simmons, P.F. and R.A. Amsler (1975). <EM>Modeling Dictionary Data</EM>, Computer Science Department, University of Texas,
905: Austin, Texas, April.</P>
906:
907: <P><A NAME="sma"></A>Small, S. (1979). "Word expert parsing," <EM>Proceedings of the 17th Annual Meeting of the Association for Computational
908: Linguistics</EM></P>
909:
910: <P><A NAME="smi"></A>Smith, R.N. (1972). "Interactive lexicon updating," <EM>Computers and the Humanities</EM>, Vol.6, No.3, January, pp. 137-145.</P>
911:
912: <P><A NAME="N_1_">1. </A>Paper presented at the 18<SUP>th</SUP> Annual Meeting of the Association for Computational Linguistics, Philadelphia, PA 1980
913: (Extended abstract at pp. 153-4 of the Proceedings).</P>
914:
915:
916: <P><A NAME="N_2_">2. </A>Current address is CL Research, 9208 Gue Road, Damascus, MD 20872, with web address at http://www.clres.com. He may be reached by email at ken@clres.com.
917: </BODY>
918: </HTML>
919: