perm filename CHAP4[4,KMC]5 blob
sn#014428 filedate 1972-11-28 generic text, type T, neo UTF8
00100 PROBLEMS OF NATURAL LANGUAGE UNDERSTANDING IN TELETYPED INTERVIEW DIALOGUES.
00150
00200
00300 By `natural language` I shall mean everyday American English
00400 such as is used by readers of this book in ordinary conversations.
00500 It is still difficult to be explicit about the processes which
00600 enable hummans to interpret and respond to natural language.
00700 Philosophers, linguists and psychologists have speculated about
00800 and investigated natural language with various purposes and few
00900 useful results. Now attempts are being made in artificial intelligence to write
01000 algorithims which `understand' what is being expressed in natural
01100 language utterances.
01200 During the 1960's when machine processing of natural language
01300 was dominated by syntactic considerations, it became clear that
01400 this approach was insufficient. The current view is that to unDerstand
01600 what utterances say, knowledGe about linguistic syntax and semantics
01700 must be combined with knowledge about an underlying conceptual
01800 structure containing a world-model and an ability to draw inferences.
01900 How to achieve this combination efficiently represents a huge task for
02000 both theory and implementation.
02100 Since the behavior being simulated by our paranoid model is the
02200 linguistic-conceptual behavior of paranoid patients in a psychiatric
02300 interview, the model must have some ability to process and respond to
02400 natural language input in a manner indicating the underlying pathological
02500
02600 to develop a method for understanding everyday Englisg sufficient
02700 for the model to behave conversationally in a paranoid way in a
02800 circumscribed situation. What is said in this situation is far
02900 icher than what is said in conversations with a block-stacking
02950 i2900
03000 robot but its requirements for constructing an interpretation
03100 of an input are not as complex as trying to understand anything
03200 said in English BY anybody in any dialogue situation.
03300 WE TOOK A PRAGMATIC APPROACH WHICH CONSIDERED "UNDERSTANDING"
03400 TO REPRESENT "GETTING THE MESSAGE" OF AN UTTERANCE BY
03500 GLEANING SOME {NOT ALL} OF THE RELATIONS BETWEEN THEM.
03600 THIS STRAIGHTFORWARD APPROACH TO A COMPLEX PROBLEM HAS ITS
03700 DRAWBACKS, AS WILL BE SHOWN, BUT WE WERE STRIVING FOR A
03800 SUFFICIENCY TO DEMONSTRATE PARANOIA RATHER THAN COMPLETE
03900 COMPREHENSION OF ENGLISH.
04000 LINGUISTIC APPROACHES CITE TRADTIONAL PROBLEMS WITH
04100 AMBIGUITY, AS ILLUSTRATED IN THE FOLLOWING EXAMPLE FROM
04200 WILKS { }. SUPPOSE I WALKED UP TO YOU, A STRANGER, ON
04300 THE STREET ON SUNDAY MORNING AND SAID
04400 {1} `HE FELL WHILE GETTING TO THE BALL'.
04500 ADMITTEDLY THIS IS A STRANGE SCENE AND IN THIS SITUATION
04600 YOU WOULD THINK ME TO BE CRAZY, HUNGOVER AND MAYBE STILL
04700 DRUNK. BUT THE EXAMPLE IS NO MORE WEIRD THAN THE ISOLATED
04800 EXAMPLES DISCUSSED IN THE LINGUISTICS LITERATURE. SUPPOSE
04900 FURTHER THAT IN YOUR PERSONAL `DICTIONARY' THE WORD `BALL'
05000 HAS AT LEAST TWO SENSES, {A} A SPHERICAL PHYSICAL OBJECT
05100 USED IN A GAME, AND {B} A FORMAL DANCE. {IT PROBABLY HAS
05200 ALSO A THIRD SENSE AS A VERB BUT WE WILL IGNORE THIS MORE
05300 OR LESS RECENT EXAMPLE OF SEMANTIC SHIFT}. HAVING NO
05400 FURTHER INFORMATION IN THIS SITUATION AND ATTEMPTING TO
05500 CONSTRUCT AN INTERPRETATION OF MY UTTERANCE, YOU WOULD BE
05600 PUZZLED AS TO WHETHER I WAS REFERRING TO A BALL GAME OR A
05700 DANCE. IF WE THEN CONTINUED ON OUR RESPECTIVE WAYS, SAYING
05800 NOTHING ELSE, YOUR PUZZLEMENT WOULD CONTINUE AND EVEN INCREASE
05900 --I DON'T KNOW WHAT HE WAS REFERRING TO NOR WHY HE EVEN SAID
06000 THAT TO ME.
06100
06200 THE AMBIGUITY ARISES BECAUSE OF THE TWO WORD SENSES FOR
06300 BALL EACH OF WHICH WOULD GIVE THE UTTERANCE A MEANINGFUL
06400 INTERPRETATION. BUT THE EXAMPLE IS EXTREMELY FORCED AND
06500 ARTIFICIAL. SUCH ISOLATED UTTERANCES CANNOT BE DISAMBIGUATED
06600 {UNIQUEATED IS A BETTER TERM} BUT THIS IS NO HANDICAP FOR
06700 ORDINARY HUMAN CONVERSATIONS IN WHICH AMBIGUITIES HARDLY ARISE
06800 AT ALL. BBESIDES THE UTTERANCE ITSELF, EXTRA INFORMATION IS
06900 USUALLY AVAILABLE IN THE FORM OF CONTEXTUAL AND SITUATIONAL
07000 KNOWLEDGE. EVEN BETTER, ONE CAN ALWAYS ASK. IF I HAD SAID
07100 ONLY UTTERANCE {1} TO YOU, YOU COULD SIMPLY ASK:
07200
07300 {2} `WHAT DO YOU MEAN?'
07400
07500 AND MY REPLY WOULD INDICATE SOMETHING ABOUT A GAME OR A
07600 DANCE OR WHO 'HE' WAS.
07700 UTTERANCES OCCUR IN CONVERSATIONS WHICH TAKE PLACE IN
07800 SOCIOPSYCHOLOGICAL SITUATIONS. THE COMMUNICANTS HAVE ROLES
07900 AND INTENTIONS TOWARDS ONE ANOTHER. IF THE SITUATION IS THAT
08000 OF A MEDICAL OR PSYCHIATRIC INTERVIEW BETWEEN DOCTOR AND
08100 PATIENT AND THE DOCTOR ASKS:
08200 {3} "HOW MUCH DO YOU DRINK?"
08300 WE KNOW FROM THE NATURE OF THE SITUATION THAT DRINK MEANS
08400 `DRINK ALCOHOL' AND DOES NOT REFER TO A TOTAL FLUID INTAKE.
08500 DIALOGUES REPRESENT CONNECTED DISCOURSE IN WHICH ALL
08600 THE UTTERANCES, EXCEPT PERHAPS FOR OPENING GREETINGS, ARE
08700 CONNECTABLE TO PREVIOUS UTTERANCES. CONTEXTS AND SUB
08800 CONTEXTS, TOPICS AND SUBTOPICS SURROUND ANY GIVEN UTTERANCE
08900 AND ACTIVATE RELEVANT WORD SENSES SUCH THAT ALTERNATIVE
09000 SENSES DO NOT ARISE IN THE COMPREHENSION PROCESS. IN
09100 SPOKEN DIALOGUES INTONATIONS AND WORD EMPHASES ARE FURTHER
09200 MEANS FOR AVOIDING AMBIGUITIES. BUT THE CONNECTED DISCOURSE
09300 OF DIALOGUES BRINGS PROBLEMS OF ITS OWN TO THE ALGORITHMIST
09400 WHOSE PROGRAM MUST KEEP TRACK OF WHAT IS GOING ON AND WHAT
09500 HAS BEEN SAID BEFORE. FOREMOST IS THE PROBLEM OF IMAPHORIC
09600 REFERENCE.
09700
09800 ANAPHORA
09900
10000 AN ANAPHORIC REFERENCE IS A WORD OR PHRASE WHICH REFERS
10100 BACKWARDS {USUALLY, I.E. THERE ARE SOME RARE FORWARD REFERENCES}
10200 TO SOMETHING IN A PREVIOUS UTTERANCE. A COMMON EXAMPLE IN
10300 INTERVIEW DIALOGUES IS THAT OF PRONOUNS.
10400 {4} PATIENT - MY FATHER WAS AN ALCOHOLIC.
10500
10600 {5} DOCTOR - WERE YOU VERY CLOSE TO HIM?
10700
10800 WHERE THE TERM `HIM' IN {4} OBVIOUSLY REFERS TO `FATHER' IN
10900 {3}. IT IS NOT TOO DIFFICULT A PROBLEM FOR A PROGRAM TO MAKE
11000 THE CORRECT ASSIGNMENTS IN PERSONAL PRONOUN ANAPHORA. OF
11100 GREATER COMPLEXITY ARE UTTERANCES CONTAINING THE WORDS `IT'
11200 OR `THIS'. FOR EXAMPLE, SUPPOSE THE INTERVIEW CONTINUED AFTER
11300 {5} AS FOLLOWS:
11400
11500 {6} PATIENT - YES I WAS, EVEN THOUGH HIS DRINKING
11600 UPSET ME.
11700
11800 {7} DOCTOR - HOW DID IT UPSET YOU?
11900
12000 HERE THE COMPREHENSION ALGORITHM MUST GRASP THAT THE `IT' OF
12100 {7} REFERS TO THE TOPIC OF THE FATHER'S DRINKING IN {6}.
12200 FURTHER, IF THE DIALOGUE CONTINUED:
12300
12400 {8} PATIENT - IF EMBARASSED ME WHEN MY FRIENDS SAW
12500 HIM DRUNK.
12600
12700 {9} DOCTOR - DO YOU THINK HE SENSED THIS?