Download Understanding Word Meaning through Sequences: An Analysis of Word Expert Parsing and more Study notes Construction in PDF only on Docsity!
VIEWING Word EXPERT PARSING AS LINGUISTIC THEORY*
Department of Computer Science Mathematical Sciences B u i l d i n g U n i v e r s i t y o f Rochester Rochester, New York 14627
ABSTRACT
The Word Expert Parser is a computer program t h a t analyzes fragments o f n a t u r a l language t e x t i n order to e x t r a c t t h e i r meaning in c o n t e x t. The c o n s t r u c t i o n of the program has l e d to the development of a l i n g u i s t i c theory based on n o t i o n s orthogonal t o those t r a d i t i o n a l l y found a t the heart of such t h e o r i e s. Word Expert Parsing e x p l a i n s the understanding o f t e x t u a l fragments c o n t a i n i n g h i g h l y i d i o s y n c r a t i c elements, such as idioms, c o l l o c a t i o n s , c l i c h e s , and c o l l i g a t i o n s , a s w e l l a s l e x i c a l sequences t h a t c o n t a i n i n t e r e s t i n g s t r u c t u r a l phenomena. The theory p e r c e i v e s the i n d i v i d u a l word of language as the o r g a n i z i n g u n i t f o r l i n g u i s t i c knowledge, and views understanding as c o n s i s t i n g of lexical Interactions among procedural word e x p e r t s. T h i s paper d e s c r i b e s four classes o f l e x i c a l i n t e r a c t i o n r e q u i r e d t o e x p l a i n the understanding o f sentences i n c o n t e x t ,
Idiosyncratic Interaction, linguistic Interaction,
discourse interaction, and logical interaction.
The paper purposely a v o i d s programming d e t a i l s in order to focus on Word Expert Parsing as l i n g u i s t i c theory.
1. INTRODUCTION
The Word Expert Parser (UEP) is a computer program t h a t analyzes fragments of n a t u r a l language t e x t i n order t o e x t r a c t t h e i r meaning i n c o n t e x t. The system has been developed w i t h p a r t i c u l a r a t t e n t i o n p a i d t o the wide v a r i e t y o f d i f f e r e n t meaning r o l e s of words when appearing in combination w i t h o t h e r words. The t h e o r e t i c a l p o s i t i o n advanced by UEP about the nature of I n d i v i d u a l words is t h a t words have no meaning per se, but r a t h e r , t h a t fragments o f l e x i c a l items mean something through t h e i r I n t e r r e l a t i o n s h i p s. Furthermore, t h e c h a r a c t e r o f l e x i c a l r e l a t i o n s runs the gamut from the simple d i r e c t knowledge t h a t some word sequence r e p r e s e n t s some remembered concept, to the more a n a l y t i c a l knowledge t h a t p a r t i c u l a r k i n d s o f l e x i c a l sequences o f t e n represent c e r t a i n c l a s s e s o f conceptual n o t i o n s.
- The i n i t i a l research m o t i v a t i n g the ideas described In t h i s r e p o r t was supported by g r a n t NSG-7253 from the N a t i o n a l A e r o n a u t i c s and Space A d m i n i s t r a t i o n t o t h e U n i v e r s i t y o f Maryland. During the w r i t i n g of t h i s paper, the author was supported b y a F u l b r i g h t L e c t u r e s h i p i n A r t i f i c i a l I n t e l l i g e n c e at the U n i v e r s i t y de P a r i s , and computer f a c i l i t i e s were p r o v i d e d by the Institut
de Recherche et Coordination Acoustlqua/Musique.
The support of NASA, the F u l b r i g h t - H a y s program, and IRCAM are g r a t e f u l l y acknowledged.
Small
Groupe d 1 I n t e l l i g e n c e A r t i f i c i e l l c Departernent d ' I n f o r m a t i q u e U n i v e r s i t y Paris V I I I - Vincennes 93200 Saint Denis FRANCE
The e v o l u t i o n o f t h i s p e r s p e c t i v e s t a r t e d w i t h the o b s e r v a t i o n t h a t the understanding of a p a r t i c u l a r fragment of t e x t depends fundamentally on the disambiguation of the i n d i v i d u a l words composing i t. Knowing the c o n t e x t u a l meanings of the words is tantamount to understanding the meaning of the o v e r a l l fragment. Another way of saying the same t h i n g [Rieger, 1977] is t h a t language i n t e r p r e t a t i o n can be u l t i m a t e l y viewed as a process of word sense d i s c r i m i n a t i o n. U n f o r t u n a t e l y , t h i s p e r s p e c t i v e does not e l i m i n a t e t h e c l a s s i c problems of d e c i d i n g the nature of a d i s t i n c t word sense, the d i f f e r e n c e between d i f f e r e n t usages, word senses, and idioms, and so f o r t h. The s o l u t i o n to these problems comes in r e a l i z i n g t h a t the process of understanding the meaning of words in c o n t e x t does not r e q u i r e r e f e r e n c e to those n o t i o n s at a l l. The design of a p a r s i n g procedure based on d e t e r m i n i n g the meaning r o l e s of i n d i v i d u a l words in c o n t e x t has l e d to the Drthogonal l i n g u i s t i c n o t i o n s t h a t are the subject o f t h i s paper.
The o r g a n i z a t i o n of UEP is founded on the b e l i e f t h a t the grouping together of words to form meaningful sequences is an a c t i v e process which succeeds o n l y because of h i g h l y i d i o s y n c r a t i c a p p l i c a t i o n of l e x i c a l knowledge. That i s , we fragment t e x t and understand the meaning of the pieces because we know how the p a r t i c u l a r words i n v o l v e d i n t e r a c t w i t h each o t h e r.
Sometimes sequences of two or more words i n t e r a c t together to such an e x t e n t t h a t they seem to behave as a s i n g l e l e x i c a l item. L i n g u i s t s have l a b e l l e d such sequences Idioms. The n o t i o n to which t h i s d e f i n i t i o n g i v e s r i s e , however, causes several problems f o r l i n g u i s t i c t h e o r y. F i r s t o f a l l , r a r e l y does such a sequence hold together so t i g h t l y t h a t i t can b e t r u l y t r e a t e d t h e o r e t i c a l l y as a s i n g l e l e x i c a l item. Secondly, r a r e l y does such a sequence have a unique meaning. More o f t e n than n o t , the meaning of an i d i o m a t i c expression must be determined by d i s a m b i g u a t i o n. The sequence must be analyzed in c o n t e x t and be t r e a t e d by comprehension processes as being e i t h e r (a) a cohesive whole w i t h i d i o s y n c r a t i c meaning, or (b) a sequence having meaning through l e s s s p e c i f i c language knowledge. There is no a priori way of knowing the meaning of the sequence to be the one or the o t h e r.
The n o t i o n of i diom f a l l s at one end of a spectrum, an i d e a l i z e d end t h a t I c l a i m does not e x i s t. L e x i c a l seauences can be more or less Idiomatic, in the sense t h a t the process i n t e r a c t i o n s c o n s t i t u t i n g the understanding of them i n c l u d e s g r e a t e r or fewer i d i o s y n c r a t i c I n t e r a c t i o n s. The UEP way of l o o k i n g at the most i d i o m a t i c sequences i s t h a t the s p e c i a l
i n t e r a c t i o n s among the p a r t i c i p a t i n g words take p r i o r i t y over any other p o t e n t i a l i n t e r a c t i o n s i n v o l v i n g those words. The d i s a m b i g u a t i o n of i d i o m a t i c expressions, i. e. , the understanding o f the sequences as e i t h e r idioms or non-idioms (to use the popular d i s t i n c t i o n ) , g e n e r a l l y r e q u i r e s other process i n t e r a c t i o n s besides the s t r i c t l y w o r d - s p e c i f i c ones. The understanding of an idiom thus d i f f e r s i n s i g n i f i c a n t l y , from the p e r s p e c t i v e of UEP t h e o r y , from comprehension of any other k i n d (according to whatever c l a s s i f i c a t i o n scheme) of l e x i c a l sequence.
The n o t i o n t h a t a l l fragments of language are more or l e s s i d i o m a t i c , w h i l e r a d i c a l in some l i n g u i s t i c q u a r t e r s , has been p r e v i o u s l y suggested. In h i s i n t r o d u c t o r y textbook, Aspects of Language, Dwight B o l i n g e r asks "whether e v e r y t h i n g we say may be in some degree i d i o m a t i c — t h a t i s , whether t h e r e are a f f i n i t i e s among words t h a t continue to r e f l e c t the attachments the words had when we learned them, w i t h i n l a r g e r groups" [ B o l i n g e r , 1975J. A f t e r working w i t h i n what he c a l l s " t h e p r e v a i l i n g r e d u c t i o n i s m " , B o l i n g e r began to suggest a p o s i t i v e answer to h i s pedagogical q u e s t i o n , choosing to take "an i d i o m a t i c r a t h e r than an a n a l y t i c a l view" [ B o l i n g e r , 1979] of language. The c o n t r i b u t i o n o f a r t i f i c i a l i n t e l l i g e n c e I n g e n e r a l , and of Word Expert Parsing in p a r t i c u l a r , is to develop theory from t h i s i n f o r m a l view. The n o t i o n of process, and of process i n t e r a c t i o n , a l l o w s us to begin to do Just t h a t.
2. THE UORD EXPERT PARSER
The UEP computer system maintains l i n g u i s t i c knowledge across a community of word-based s t r u c t u r e s c a l l e d word experts, which represent the process of d e t e r m i n i n g the c o n t e x t u a l meaning and r o l e of the i n d i v i d u a l words. A word expert must not be thought of as a r e p r e s e n t a t i o n f o r the v a r i o u s meanings, r o l e s , and c o n t r i b u t i o n s of a word in c o n t e x t , but r a t h e r as a d e c l a r a t i v e r e p r e s e n t a t i o n (a network) of the process (which we s h a l l c a l l disambiguation) of d e t e r m i n i n g these t h i n g s. C e r t a i n l y , i t i s the meaning c o n t r i b u t i o n s of i n d i v i d u a l l e x i c a l items t h a t we wish to determine. Word e x p e r t s are b o t h data and process; they can be augmented, examined, and manipulated as data, y e t p a r s i n g takes p l a c e through t h e i r i n t e r p r e t a t i o n as program by an e x p e r t e v a l u a t o r , s i m i l a r to the EVAL of L i s p.
The d i s t r i b u t e d p a r s i n g scheme of UEP works as f o l l o w s. The UEP reader examines a word of t e x t and r e t r i e v e s i t s word e x p e r t from memory. The word e x p e r t s t a r t s e x e c u t i n g , t r y i n g t o determine the meaning r o l e o f i t s word i n c o n t e x t , i. e. , i n t e r a c t i n g w i t h other word e x p e r t s and w i t h h i g h e r - o r d e r system processes to a c q u i r e the a p p r o p r i a t e c o n t e x t u a l knowledge to make the c o r r e c t i n f e r e n c e s. F i n a l l y , a l l the word e x p e r t s f o r a p a r t i c u l a r fragment of t e x t come to mutual agreement on the meaning of the fragment, and the l o c a l d i s t r i b u t e d process t e r m i n a t e s. L o c a l , i n the sense t h a t as long as t h e r e remains i n p u t t e x t , the o v e r a l l p a r s i n g process c o n t i n u e s , w h i l e the d i s a m b i g u a t i o n o f i n d i v i d u a l l e x i c a l sequences making up the l a r g e r t e x t completes.
I n t e r a c t i o n , between i n d i v i d u a l s i n the w o r l d , or between d i s t r i b u t e d processes in a computer program, r e q u i r e s both (a) g i v i n g i n f o r m a t i o n and (b) r e c e i v i n g i n f o r m a t i o n. In UEP, the e x p e r t s exchange two kinds of i n f o r m a t i o n , c a l l e d concept structures and control signals. Concept structures represent human concepts, such as "a book", " g o i n g f i s h i n g " , " t h e box of candy 1 gave Joanie f o r V a l e n t i n e ' s day in 1981", "some b l u e p h y s i c a l
o b j e c t " , and the l i k e. Control s i g n a l s represent p r o c e s s i n g c l u e s , such as "expect a word t h a t can begin a l e x i c a l sequence t h a t can d e s c r i b e concept s t r u c t u r e X", "send me the concept s t r u c t u r e r e p r e s e n t i n g the agent of concept s t r u c t u r e Y or a s i g n a l saying you cannot", " w a i t a second and you w i l l be sent a concept s t r u c t u r e t h a t w i l l help y o u " , and s i m i l a r t h i n g s. The r e p r e s e n t a t i o n and use of concepts and s i g n a l s are described f u l l y in [Small, 1980].
3, LEXICAL INTERACTIONS
I use the term lexical interaction to denote the sending and r e c e i v i n g of c o n t r o l s i g n a l s and concept s t r u c t u r e s by word experts in UEP, This i n c l u d e s i n t e r a c t i o n s between i n d i v i d u a l experts, as w e l l as those between a word expert and another k i n d of model process ( e. g. , a mechanism i n f e r r i n g the goals of a d i a l o g u e p a r t i c i p a n t ). This paper discusses l e x i c a l i n t e r a c t i o n s b y p r e s e n t i n g four classes of r e q u i r e d i n t e r a c t i o n , and then arguing f o r the n e c e s s i t y and g i v i n g examples of each. The c a t e g o r i z a t i o n is by the k i n d of knowledge exchanged in the communication, and i n c l u d e s the f o l l o w i n g.
Idiosyncratic Interaction
Linguistic Interaction
Discourse Interaction
Logical Interaction
The l e a s t general c l a s s of l e x i c a l i n t e r a c t i o n s are considered i d i o s y n c r a t i c since they are w o r d - s p e c i f i c and a r i s e through simple r e c a l l memory. T h i s type of i n t e r a c t i o n permits the understanding of i d i o m a t i c fragments. General knowledge about the syntax and semantics of some n a t u r a l language g i v e s r i s e to linguistic i n t e r a c t i o n s , and are of course c r u c i a l to the understanding of l e x i c a l sequences not p r e v i o u s l y seen. Sometimes words i n t e r a c t w i t h processes t h a t monitor the development of an e n t i r e t e x t (or p a r t s t h e r e o f ) , o r the goals o f p a r t i c i p a n t s i n d i s c u s s i o n. These discourse i n t e r a c t i o n s are o f t e n necessary f o r the meaningful cohesion of l e x i c a l fragments. L a s t l y , but c e r t a i n l y not l e a s t i m p o r t a n t are the logical i n t e r a c t i o n s between words and the most general c o g n i t i v e processes. Perceptions about the w o r l d , b e l i e f s , inference-making s k i l l s , r o t e memory, and so f o r t h , a r e b a s i c to language understanding.
The c l a s s i f i c a t i o n of word fragments i n t o c a t e g o r i e s such as " i d i o m " , " c o l l o c a t i o n " , " c o l l i g a t i o n " , "noun phrase", "complement", and the l i k e , does not make sense in UEP theory. Rather, i n d i v i d u a l words are viewed as having c e r t a i n kinds and sequences of i n t e r a c t i o n s w i t h t h e i r neighbors to form meaningful pieces of t e x t. Fragments o f t e n d e s c r i b e d as " I d i o m s " are those t h a t are understood p r i n c i p a l l y through i d i o s y n c r a t i c l e x i c a l i n t e r a c t i o n s. A n o n - i d i o m a t i c s t r u c t u r e , diagnosed as a "noun phrase", is one t h a t i n v o l v e s mostly l i n g u i s t i c i n t e r a c t i o n s t o understand. A s o - c a l l e d "noun-noun p a i r " can be thought of as a l e x i c a l sequence comprehended w i t h the help of l o g i c a l i n t e r a c t i o n s , w i t h recourse to common sense memory and s k i l l s.
3.1 I d i o s y n c r a t i c i n t e r a c t i o n
Since the emphasis of the UEP research e f f o r t is to c o n s t r u c t a computer program to understand n a t u r a l language, we are not q u a l i f i e d to make
Noun Phrases
How can the p u r e l y l e x i c a l UEP system r e q u i r e no n o t i o n of h i g h order s t r u c t u r a l phenomena, y e t s t i l l b e a b l e t o account f o r them? The f o l l o w i n g example ( p r o v i d e d b y Y o r i c k w i l k s ) i l l u s t r a t e s the l e x i c a l i n t e r a c t i o n s r e q u i r e d t o a n a l y z e a n i n t e r e s t i n g fragment o f t e x t.
(3) " J o a n i e washes the c o l o r f u l d i s h e s u p. "
The d i f f i c u l t y w i t h t h i s fragment i s i n d e t e r m i n i n g t h a t the word dishes c o n t r i b u t e s to the meaning of t h e fragment t h r o u g h i n t e r a c t i o n s w i t h the two words t o i t s l e f t , b u t t h a t the word u p c o n t r i b u t e s by a s s o c i a t i o n w i t h the word washes, which precedes up by many i n t e r v e n i n g words. The reason t h a t I am a v o i d i n g the use o f t r a d i t i o n a l l i n g u i s t i c j a r g o n f o r d e s c r i b i n g t h i s phenomenon i s the f o l l o w i n g b e l i e f : An u n d e r s t a n d i n g of UEP r e q u i r e s the v i e w i n g o f language i n t e r p r e t a t i o n from the vantage p o i n t o f the i n d i v i d u a l word and i t s i n t e r a c t i o n s. An i m p o r t a n t way to achieve t h i s is to d e s c r i b e the a n a l y s i s process w i t h r e f e r e n c e t o the very n o t i o n s (not t h e t r a d i t i o n a l ones) around which i t i s o r g a n i z e d.
In the a n a l y s i s of the example fragment, UEP would f i n d the r e f e r e n t of Joanie and then proceed as f o l l o w s. The wash e x p e r t would b e g i n e x e c u t i n g , t r y i n g to d e t e r m i n e i t s own meaning r o l e in some l e x i c a l fragment, and at t h e same t i m e , t r y i n g to p r o v i d e i n f o r m a t i o n t o o t h e r l e x i c a l agents t o p e r m i t them to do the same. The meaning of wash in c o n t e x t depends on a number of f a c t o r s , i n c l u d i n g the n a t u r e o f the words succeeding i t , and t h e i r own a c t i o n s i n d e t e r m i n i n g t h e i r meaning and r o l e c o n t r i b u t i o n s in c o n t e x t. The wash e x p e r t must thus prepare f o r a number of c o n t i n g e n c i e s , or d i f f e r e n t t h i n g s t h a t c o u l d happen i n the t e x t , and then, w a i t t o see i f any o f them a c t u a l l y occur. I f the word up appears to t h e r i g h t of wash, f o r example, the words c o u l d choose to p a i r up i n t o a meaningful fragment (as in throw in above). Under c e r t a i n c o n d i t i o n s , the word up c o u l d appear l a t e r on in t h e t e x t , and s t i l l p a i r up w i t h wash (as must occur f o r c o r r e c t i n t e r p r e t a t i o n o f the example s e n t e n c e ).
What a r e the c o n t e x t u a l c o n d i t i o n s t h a t would p e r m i t t h i s? One of t h e c o n t i n g e n c i e s t h a t wash a n t i c i p a t e s i s t h e g r o u p i n g o f t h e words t o i t s r i g h t i n t o a meaningful fragment of t h e i r own ( i. e. , a concept structure). The wash e x p e r t knows t h a t (a) t h e n a t u r e of t h i s concept s t r u c t u r e may be i m p o r t a n t f o r i t s own sense d i s a m b i g u a t i o n , and (b) t h a t t h e word immediately f o l l o w i n g the meaningful fragment c o u l d p a i r u p w i t h i t. I n the Jargon of UEP, one of t h e e x p e r t s in such a meaningful fragment reports a concept s t r u c t u r e. Since t h e up e x p e r t does n o t r e p l y to dishes w i t h an a c c e p t a b l e message to c o n t i n u e the ongoing concept b u i l d i n g a c t i v i t y , the dishes e x p e r t r e p o r t s t h e s t r u c t u r e. I t i s t h i s r e p o r t t h a t t r i g g e r s some new p r o c e s s i n g by t h e wash e x p e r t , namely the e x a m i n a t i o n of t h e n e x t word ( i. e. , up).
The r e s t o f t h e a n a l y s i s t a k e s p l a c e p r e d i c t a b l y. The wash e x p e r t i n t e r a c t s w i t h up as i f u p o c c u r r e d t o i t s immediate r i g h t i n the t e x t. The p a i r i n g up of the two words r e s u l t s from mutual a c c o r d , and t h e wash e x p e r t c r e a t e s a concept s t r u c t u r e t o r e p r e s e n t t h e meaning o f the washing up of d i s h e s. Next wash o r g a n i z e s t h e conceptual o b j e c t , the c o 7 o r f u l dishes, i n t o t h e o v e r a l l meaning o f t h e sentence, and a g a i n w a i t s f o r t h i n g s t o happen. T h i s t i m e , t h e word e x p e r t f o r t h e p e r i o d at t h e end of the sentence executes, and t r a n s m i t s an a p p r o p r i a t e message. The wash e x p e r t a g a i n e x e c u t e s , c l e a n s u p i t s b u s i n e s s , and r e p o r t s
t h e concept s t r u c t u r e r e p r e s e n t i n g the meaning, i n c o n t e x t , o f t h e e n t i r e fragment (sentence).
P a s s i v e s and R e l a t i v e Clauses
Sentences in the p a s s i v e v o i c e and those c o n t a i n i n g r e l a t i v e clauses a r e s i m i l a r i n b e i n g complex s t r u c t u r a l phenomena i n n a t u r a l language, and o f t e n s u g g e s t i v e o f s e n t e n c e - l e v e l r u l e s a s l i n g u i s t i c e x p l a n a t i o n. Furthermore, t h e u n d e r s t a n d i n g of such c o n s t r u c t i o n s by the d i s t r i b u t e d word-based approach of wEP may be f a r f r o m e v i d e n t , e s p e c i a l l y c o n s i d e r i n g my c l a i m t h a t n o e x p l i c i t n o t i o n s o f s t r u c t u r e are r e f e r e n c e d b y t h e computer system or used in the t h e o r y. I n t e r p r e t a t i o n o f t e x t u a l fragments c o n t a i n i n g complex s y n t a c t i c s t r u c t u r e s takes p l a c e t h r o u g h complex p a t t e r n s o f l e x i c a l i n t e r a c t i o n s among the a p p r o p r i a t e word e x p e r t s. The words t h a t n o r m a l l y cue a reader about the presence of such s t r u c t u r a l r e l a t i o n s in a fragment a r e the ones in wEP t h a t c o o r d i n a t e the process of u n d e r s t a n d i n g them.
The a n a l y s i s of a p a s s i v e sentence i n v o l v e s l i n g u i s t i c i n t e r a c t i o n s among t h e word e x p e r t s f o r t h e s u f f i x en, t h e word by, and the o t h e r words composing i t. The f o l l o w i n g sentence has been parsed by the e x i s t i n g UEP system, and d i s c u s s e d at l e n g t h i n [Small, 1980].
(4) "The case was thrown out by f e d e r a l c o u r t. "
The en e x p e r t begins e x e c u t i n g b e f o r e throw, and t h e normal a t t e m p t s by t h e throw e x p e r t to c o o r d i n a t e the a n a l y s i s o f the fragment i n which i t p a r t i c i p a t e s a r e i n t e r c e p t e d by an. The a c t i o n s of en a l l o w throw to p a i r up w i t h o u t. as o u t l i n e d above f o r throw in and wash up, but i t s l e x i c a l i n t e r a c t i o n s t o d e t e r m i n e t h e n a t u r e o f t h e o b j e c t b e i n g " t h r o w n o u t " , and t h e agent d o i n g t h e " t h r o w i n g " a r e a l l i n t e r c e p t e d b y t h e e n e x p e r t , which p r o v i d e s throw w i t h t h e c o r r e c t r e p l i e s t o i t s q u e r i e s. Please r e f e r t o [Small, 1980] and [ S m a l l , 1981] f o r a f u l l e r d i s c u s s i o n.
R e l a t i v e c l a u s e s b e g i n n i n g w i t h the word who a r e analyzed by UEP t h r o u g h the i n t e r a c t i o n s among t h e who word e x p e r t and t h e e x p e r t s f o r the o t h e r words in t h e c l a u s e and t h e l a r g e r fragment c o n t a i n i n g i t. The f o l l o w i n g sentence i s a n example of such a fragment.
(5) "The man who throws the game l i k e s to l o s e. "
The who e x p e r t in t h i s sentence has the r e s p o n s i b i l i t y f o r i n t e r a c t i n g w i t h the word e x p e r t f o r likes to i n f o r m Tikes about the man d o i n g t h e " l i k i n g ". O r d i n a r i l y , t h i s e x p e r t would expect t o f i n d a meaningful l e x i c a l sequence t o i t s l e f t r e p r e s e n t i n g the needed concept. However, the p a r t i c u l a r s t r u c t u r e of the fragment means t h a t who must be at t h e o t h e r end of the r e l e v a n t l i n g u i s t i c i n t e r a c t i o n s o f likes, r a t h e r than the e x p e r t f o r t h e word t o i t s immediate l e f t , which would n o r m a l l y p e r f o r m the needed s e r v i c e.
The UEP i n t e r p r e t a t i o n of the example fragment proceeds as f o l l o w s. The word e x p e r t s f o r the and man agree to form a meaningful sequence and c o n s t r u c t a concept s t r u c t u r e t o r e p r e s e n t i t s meaning. The who e x p e r t begins e x e c u t i n g , g e t s h o l d of t h i s concept, and w a i t s f o r the throw e x p e r t t o s t a r t e x p l o r i n g the n a t u r e o f t h e l e x i c a l sequence o n i t s l e f t. I n a d d i t i o n , the who e x p e r t a n t i c i p a t e s t h a t another word e x p e r t f u r t h e r down the l i n e ( i n the example, t h e e x p e r t f o r 7 7/ces) w i l l a l s o seek out i n f o r m a t i o n about the sequence t o i t s l e f t , i n e x a c t l y the way t h a t throw does. The who e x p e r t , l i k e every word e x p e r t in UEP,
plans a s t r a t e g y to i n t e r a c t w i t h the e x p e r t s i n v o l v e d i n both i t s p r i o r c o n t e x t and i t s subsequent c o n t e x t , c o o p e r a t i v e l y t o i n t e r p r e t fragments of t e x t.
The throw e x p e r t begins e x e c u t i n g and i n v e s t i g a t e s the nature of the l e x i c a l sequence to i t s l e f t. The who e x p e r t p r o v i d e s the a p p r o p r i a t e I n f o r m a t i o n , i. e. , the concept s t r u c t u r e r e p r e s e n t i n g the men, and throw begins to disambiguate i t s meaning in c o n t e x t. The e x p e r t s f o r a and game m u t u a l l y agree on t h e i r l o c a l meaning, and through l i n g u i s t i c and i d i o s y n c r a t i c i n t e r a c t i o n s w i t h throw help i t determine i t s meaning as the " t h r o w i n g of a c o n t e s t ". The likes e x p e r t s t a r t s e x e c u t i n g , and i t s messages i n search of the person doing the " l i k i n g " are i n t e r c e p t e d by the who e x p e r t , which has been on the l o o k o u t f o r such i n t e r a c t i o n s since the b e g i n n i n g. Since the who e x p e r t knows the unique name of the concept s t r u c t u r e r e p r e s e n t i n g the man, it sends t h i s concept to likes , which proceeds n o r m a l l y , knowing n o t h i n g o f the s t r u c t u r a l c o m p l e x i t i e s preceeding i t.
The word e x p e r t s f o r b o t h throw and f o r likes can be expected to e x p l o r e the u n d e r l y i n g meaning of the l e x i c a l sequences preceding them. Note the way t h a t UEP a p p l i e s t h i s l i n g u i s t i c knowledge to the i n t e r p r e t a t i o n of fragments of n a t u r a l language t e x t c o n t a i n i n g these words. Rather than saying t h a t throw and Ukes a c t as f i n i t e verbs in c e r t a i n c o n t e x t s (which are d e s c r i b e d in some r e l a t i o n a l r e p r e s e n t a t i o n a l scheme, such as grammar r u l e s or l o g i c ) , we say i n s t e a d t h a t these words c a r r y on l i n g u i s t i c i n t e r a c t i o n s w i t h the a c t i v e processes modelling the other words making up the ( l o c a l l i n g u i s t i c ) c o n t e x t t o a r r i v e a t a m u t u a l l y acceptable c h a r a c t e r i z a t i o n o f t h e i r i n d i v i d u a l c o n t r i b u t i o n s to t e x t u a l meaning. The advantage of t h i s p e r s p e c t i v e comes from the f a c t t h a t l i n g u i s t i c i n t e r a c t i o n s c o n s t i t u t e but a p o r t i o n o f a l l p o s s i b l e l e x i c a l i n t e r a c t i o n s t h a t r e p r e s e n t i n UEP the process of understanding.
- 3 Discourse I n t e r a c t i o n s
While i t i s c l e a r t h a t c e r t a i n l e x i c a l sequences cannot be understood s o l e l y through recourse to syntax and semantics, namely those fragments f o r which i d i o s y n c r a t i c i n t e r a c t i o n s are r e q u i r e d ( i. e. , s p e c i f i c remembered c o n t e x t s ) , why do we need o t h e r k i n d s of general knowledge? We have a l r e a d y seen examples suggesting the answer to t h i s q u e s t i o n. I n t r y i n g t o understand the meaning of throw 1n the towel, the r e l e v a n t word e x p e r t s must f i n d out some t h i n g s about the person performing the d e s c r i b e d a c t i o n , b e f o r e knowing what a c t i o n h e i s i n e f f e c t p e r f o r m i n g.
I f the d i s c o u r s e d e s c r i b e s some s o r t o f c o m p e t i t i o n between two people (or teams), f o r example, throw in the towel could indicate a concession of d e f e a t by one of them. The f o l l o w i n g fragment i l l u s t r a t e s such a c o n t e x t u a l s i t u a t i o n.
(6) "Rick and Joanle p l a y chess. Rick throws i n the t o w e l. "
On the other hand, if t h e d i s c o u r s e has r e c e n t l y made reference to a p l a c e where one might dispose of a t o w e l , throw 1n the towel might be s i g n i f y i n g the p u t t i n g of some towel in t h a t p l a c e. The f o l l o w i n g example i l l u s t r a t e s t h i s case.
(7) "Joanie drops a penny in the p i t. Rick throws i n the t o w e l. "
I am not c l a i m i n g t h a t knowledge of the d i s c o u r s e
c o n t e x t is s u f f i c i e n t to disambiguate the meanings of the example sentence, but r a t h e r , t h a t such knowledge i s r e q u i r e d t o understand i t.
The d i s c o u r s e i n t e r a c t i o n s r e q u i r e d to i n t e r p r e t the above example take place (a) between the throw expert and a higher order process m o d e l l i n g the activity context, and (b) between the 1n e x p e r t an a process modelling the d i s c o u r s e focus of attention.* There are two aspects to the processing of the a c t i v i t y mechanism, the u n s o l i c i t e d sending o f c o n t r o l s i g n a l s t o i n d i c a t e the a n t i c i p a t i o n o f c e r t a i n a c t i o n s i n the t e x t and concept s t r u c t u r e s to represent them, and the more d a t a - d i r e c t e d i n t e r a c t i o n s w i t h word experts (and other understanding processes) to determine the n a t u r e of the a c t i o n s t h a t a c t u a l l y do occur. The throw expert must c a r r y on a c t i v i t y context Interactions to determine if the discourse could be seen as d i s c u s s i n g some c o m p e t i t i v e a c t i v i t y. If so, the "concession of d e f e a t " i n t e r p r e t a t i o n of the example sentence is p l a u s i b l e. The in expert carries on rocus or attention interactions to find out if some l o c a t i o n has r e c e n t l y been described in the t e x t in which something might be thrown.
While the UEP system has been d i r e c t e d toward the understanding of fragments of t e x t o c c u r r i n g in t e x t u a l d i s c o u r s e , the issues a r i s i n g i n the i n t e r p r e t a t i o n of d i a l o g u e are very s i m i l a r. The d i f f e r e n c e between the two tasks i n v o l v e s the n a t u r e o f d i s c o u r s e i n t e r a c t i o n s. I n i n t e r p r e t i n g fragments of d i a l o g u e from the vantage p o i n t of one of the p a r t i c i p a n t s , word experts must i n t e r a c t w i t h model processes m o n i t o r i n g the goals of the o t h e r p a r t i c i p a n t. The f o l l o w i n g example (provided by James A l l e n ) i l l u s t r a t e s the q u e s t i o n.
(8) "When is the Windsor t r a i n? "
I n t r y i n g t o understand t h i s q u e s t i o n from the p e r s p e c t i v e of the person at the i n f o r m a t i o n desk of a t r a i n s t a t i o n , the q u e s t i o n could be d i r e c t e d a t e l i c i t i n g e i t h e r o f two pieces o f i n f o r m a t i o n [ A l l e n , 1978], i. e. , the time o f the next a r r i v a l from Windsor, or the time of the next departure to Windsor.
By saying t h a t the Windsor train is a "noun-noun p a i r " , we get nowhere in t r y i n g to understand i t. In UEP, the word experts f o r Windsor and t r a i n would i n t e r a c t l o c a l l y and determine the range of p o s s i b l e i n t e r p r e t a t i o n s f o r the fragment. In the case of t e x t u a l d i s c o u r s e , the train expert would c a r r y on d i s c o u r s e i n t e r a c t i o n s w i t h the a c t i v i t y process t o f i n d out i f d i s c u s s i o n o f some p a r t i c u l a r t r a i n were a n t i c i p a t e d in the t e x t. In the case of d i a l o g u e , these i n t e r a c t i o n s would occur between train and an Intention mechanism, which might determine t h a t the speaker in the d i a l o g u e is concerned w i t h the t r a i n s coming from Windsor, and not w i t h the t r a i n s l e a v i n g f o r Windsor. I f the processes modelling the a c t i v i t y c o n t e x t o r the speaker i n t e n t i o n s cannot p r o v i d e help to the t r a i n e x p e r t , the word e x p e r t s f o r the sequence would c o n s t r u c t a concept s t r u c t u r e to represent the d i s j u n c t of the two p o s s i b i l i t i e s , but continue t o await the i n f o r m a t i o n t h a t would decide between them.
- The term activity context d e s c r i b e s a n o t i o n s i m i l a r to the scripts of Schank and Abelson [1975] and to the rrames of Charniak [1977]. The n o t i o n of rocus or attention has been taken d i r e c t l y from the work of Grosz [1977].
The understanding of these fragments is c o o r d i n a t e d in WEP by the word e x p e r t f o r the a f f i x 1ng. The 1ng expert i n t e r a c t s l i n g u i s t i c a l l y w i t h the e x p e r t s f o r the words around i t , h e l p i n g them form meaningful sequences, and c a r r i e s on l o g i c a l i n t e r a c t i o n s w i t h the b e l i e f m o d e l l i n g process t o determine the r e l a t i v e p l a u s i b i l i t y o f the two p r o p o s i t i o n s p o s s i b l y s i g n i f i e d b y the l a r g e r sequence. In the f i r s t case above, the ing e x p e r t begins e x e c u t i n g a f t e r the and man have a l r e a d y s t a r t e d c o n s t r u c t i n g a concept s t r u c t u r e to represent the meaning of the man. It a w a i t s the r e p o r t of t h i s concept s t r u c t u r e , as w e l l as the one to be r e p o r t e d by the t i g e r word e x p e r t. Furthermore, 1ng c a r r i e s on l i n g u i s t i c i n t e r a c t i o n s w i t h eat to a r r i v e c o o p e r a t i v e l y at a concept s t r u c t u r e r e p r e s e n t i n g i t s meaning. The i n g e x p e r t then has a p l a u s i b i l i t y interaction with the belief modeller, and c o o r d i n a t e s the remainder of the understanding process based on t h i s i m p o r t a n t knowledge.
4. S U M M A R Y
Word Expert Parsing is a l i n g u i s t i c theory based on a l e x i c a l o r g a n i z a t i o n of l i n g u i s t i c knowledge represented p r o c e d u r a l l y in word e x p e r t s. The comprehension of fragments of n a t u r a l language t e x t is viewed as a process of word i n t e r a c t i o n s , where a c t i v e l e x i c a l agents cooperate to form meaningful sequences o f i n t e r r e l a t e d l e x i c a l items. L e x i c a l i n t e r a c t i o n s are o f four types, idiosyncratic, linguistic, discourse, and logical. I d i o s y n c r a t i c i n t e r a c t i o n s a l l o w UEP to e x p l a i n the understanding of i d i o m a t i c (more or l e s s i d i o m a t i c ) l e x i c a l sequences, by comparing new sequences w i t h e x p l i c i t l y remembered ones ( c a l l e d prerabs by Bolinger [1979]). L i n g u i s t i c i n t e r a c t i o n s enable the use of s y n t a c t i c and semantic g e n e r a l i z a t i o n s t o i n t e r p r e t fragments, and d i s c o u r s e i n t e r a c t i o n s p r o v i d e word e x p e r t s w i t h knowledge of d i s c o u r s e a c t i v i t i e s and f o c i o f a t t e n t i o n. L o g i c a l i n t e r a c t i o n s a l l o w word e x p e r t s to use knowledge about the r e a l - w o r l d , e s p e c i a l l y about the m u l t i p l e p e r s p e c t i v e s o f i n d i v i d u a l conceptual o b j e c t s w i t h i n i t and the r e l a t i v e p l a u s i b i l i t y o f p r o p o s i t i o n s about i t.
S. ACKNOWLEDGEMENTS
The t h e o r y presented in t h i s paper was o r i g i n a l l y conceived i n wonderful c o o p e r a t i o n w i t h Chuck Rieger over the past several years. Some of the s e n s i b l e p e r s p e c t i v e s in the paper have b e n e f i t t e d from much a p p r e c i a t e d w r i t t e n and spoken suggestions and c r i t i c i s m s of Y o r i c k W i l k s , Dick Hudson, Dwight B o l i n g e r , Pat Hayes, and James A l l e n. The Groupe d ' I n t e l l i g e n c e A r t i f i c i e l l e o f the U n i v e r s i t y P a r i s V I I I - Vincennes has p r o v i d e d an e x c e l l e n t environment f o r r e s e a r c h. Thanks to P a t r i c k Greussay, Harald Wertz, Daniel Goosens, Annette C a t t e n a t , and Gerard Paul. Many e x t r a thanks t o P a t r i c k f o r a l l the personal and r o f e s s i o n a l help he has g i v e n in making my year at incennes both v a l u a b l e and e n j o y a b l e.
6. REFERENCES
Allen, James (1978), Recognizing Intention in Dialogue, Technical Report, Department of Computer Science, U n i v e r s i t y o f Toronto.
B o l i n g e r , Dwight (1975), Aspects of Language, Harcourt Brace Jovanovich.
B o l i n g e r , Dwight (1979), Meaning and Memory, in Experience Forms, Haydn ( e d. ) , Mouton.
Charniak, Eugene (1977), Ms. Malaprop, A Language Comprehension Program, F i f t h I n t e r n a t i o n a l J o i n t Conference o n A r t i f i c i a l I n t e l l i g e n c e.
Chomsky, Noam (1965), Aspects of the Theory of Syntax, MIT Press.
Dik, Simon C. (1978), F u n c t i o n a l Grammar, North Holland.
F i l l m o r e , Charles J. (1968), The Case f o r Case, in Universa Is in L i n g u i s t i c Theory, Each and Harms ( e d s. ) , H o l t.
Grosz, Barbara (1977), The Representation and Use of Focus in Dialogue Understanding, Technical Note #151, S t a n f o r d Research I n s t i t u t e.
Hudson, Richard A. (1979), Pan-Lexical ism, Working Paper, Department of Phonetics and L i n g u s t i c s , U n i v e r s i t y College London.
Jackendoff, Ray (1972), Semantic I n t e r p r e t a t i o n in Generative Grammar, MIT Press.
Jackendoff, Ray (1976), Toward an Explanatory Semantic Representation, Linguistic Inquiry, v. 7 , n. 1.
Kaplan, Ronald M., and Joan W. Bresnan (1980), Lexical-Functional Grammar: A Formal System for Grammatical Representation, Occasional Taper #13, MIT Center f o r C o g n i t i v e Science.
Marcus, M i t c h e l l P. (1979), An Overview or a Theory of Syntactic Recognition for Natural Language, AI Memo #531, MIT A r t i f i c i a l I n t e l i g e n c e Laboratory.
Miller, George A. (1978), Semantic Relations among Words, in L i n g u i s t i c Theory and P s y c h o l o g i c a l R e a l i t y , H a l l e , Bresnan, and M i l l e r ( e d s. ) , MIT Press.
Mitchell, T. F. (1971), Linguistic "goings on"; collocations and other lexical matters arising on the s y n t a c t i c r e c o r d , Archivum L i n g u i s t i c u m , v. 2.
Rieger, Chuck (1977), Viewing Parsing as yord Sense D i s c r i m i n a t i o n , in A Survey of L i n g u i s t i c Science, Dingwall ( e d. ). Greylock P u b l i s h e r s.
Riesbeck, C h r i s t o p h e r K., and Roger C. Schank (1976), Comprehension by Computer: Expectation-Based Analysis of Sentences in Context, Research Report #78, Department of Computer Science, Yale U n i v e r s i t y.
Schank, Roger C., and Robert P. Abel son (1975), Scripts, Plans, and Knowledge, Fourth International J o i n t Conference o n A r t i f i c i a l I n t e l l i g e n c e.
Small, Steven (1980), Word Expert P a r s i n g : A Theory of Distributed Word-Based Natural Language Understanding, Technical Report #954, Department of Computer Science, U n i v e r s i t y of Maryland.
Small, Steven (1981), Toward a Cognitive Mechanics for Distributed Modelling, Technical Report, Department of Computer Science, U n i v e r s i t y of Rochester ( t o appear).
Smith, Edward E., Lance J. Rips, and Edward J. Shoben (1974), Semantic Memory and Psychological Semantics, in The Psychology of L e a r n i n g and M o t i v a t i o n , v. 8, Bower ( e d. ) , Academic Press.
W i l k s , Yorick (1980), Some Thoughts on Procedural Semantics, C o g n i t i v e Studies Centre Report # 1 , U n i v e r s i t y of Essex.
