00100	    		CHAPTER FOUR
00200	 SPECIAL PROBLEMS FOR COMPUTER UNDERSTANDING OF NATURAL LANGUAGE 
00300		 IN TELETYPED PSYCHIATRIC INTERVIEWS
00400	
00500	
00600	By `natural language` I shall mean everyday American English such  as
00700	is  used  by  readers  of  this book in ordinary conversations. It is
00800	still difficult to be  explicit  about  the  processes  which  enable
00900	hummans  to  interpret and respond to natural language. Philosophers,
01000	linguists and psychologists have investigated natural  language  with
01100	various purposes and few useful results.  Now attempts are being made
01200	in artificial intelligence to write  algorithims  which  `understand'
01300	natural   language  expressions.     
01310		During the 1960's when
01400	machine processing of natural language  was  dominated  by  syntactic
01500	considerations,  it  became  clear that syntactical information alone
01600	was  insufficient  to  comprehend   the   expressions   of   ordinary
01700	conversations. The current view is that to understand what is said in
01800	linguistic expressions, syntax and semantics must  be  combined  with
01900	beliefs  from an underlying conceptual structure having an ability to
02000	draw inferences. How to achieve this combination efficiently  with  a
02100	large  data-base  represents  a  monumental  task for both theory and
02200	implementation.  
02210		Since the behavior being simulated by our
02300	paranoid  model  is  the language-behavior of a paranoid patient in a
02400	psychiatric interview, the model must have an  ability  to  interpret
02500	and respond to natural language input  sufficient only to demonstrate
02600	language-behavior characteristic of the paranoid mode.  How  language
02700	is  understood  depends  on  the  intentions  of  the  producers  and
02800	interpreters  in  the  dialogue. Thus  language  is   understood   in
02900	accordance with the participant's view of the game being played. Our purpose was to develop a
03000	method for understanding everyday English sufficient for the model to
03100	communicate linguistically in  a  paranoid  way  in the  circumscribed
03200	situation of a psychiatric interview.
03400	We did not try to construct a general-purpose algorithm  which  could
03500	understand  anything  said  in  English  by anybody to anybody in any
03600	dialogue situation. (Does anyone believe it possible?)
03700		We took as a pragmatic measure of "understanding" the ability
03800	of the algorithm to `get the message' of an expression by trying to classify
03900	the imperative or directive intent of the interviewer,i.e.what effect he is
04000	trying to bring about in the interviewee relative to the topic.  This
04100	straightforward  approach  to a complex problem has its drawbacks, as
04200	will be shown, but we strove for a highly individualized idiolect sufficient
04300	to  demonstrate  paranoid  processes of an individual in a particular
04400	situation rather than for a general supra-individual or ideal  comprehension
04500	of  English.  If the language-understanding process interfered  with
04600	demonstrating the paranoid processes, we would consider it  defective
04700	and  insufficient  for  our  purposes.             (Insert from Machr
04800	here)
04900		Some special problems  a dialogue algorithm must cope with in a
05000	psychiatric      interview      will      now      be      discussed.
05010	
05100	QUESTIONS
05200	
05300		The  principal  sentence-type used by an interviewer consists
05400	of a question. The usual wh- and yes-no questions must be  recognized
05500	by  the  language-analyzer.  In  teletyped  interviews a question may
05600	sometimes be put in declarative form followed by a question  mark  as in:
05700		(1) PT.- I LIKE TO GAMBLE ON THE HORSES.             	
05800		   DR.- YOU GAMBLE?
05900	
06000	Particularly difficult are `when' questions which  require  a  memory
06100	which  can  assign  each  event a beginning, end and a duration. Also
06200	troublesome are questions such as `how often',  `how  many',  i.e.  a
06300	`how' followed by a quantifier.
06400		In constructing a simulation  of  a  thought  process  it  is
06500	arbitrary  how  much  information  to represent in memory. Should the
06600	model know what is the capital of Alabama? It is trivial to store a lot of facts. We took the position  that
06700	the  model  should  know  only what we believed it reasonable to know
06800	about a few hundred topics expectable  in  a  psychiatric  interview.
06900	Thus  the  model  performs  badly  when  subjected  to baiting `exam'
07000	questions designed to test its informational limitations rather than to seek useful
07100	psychiatric information.
07200		IMPERATIVES
07300	
07400		Typical imperatives in a  psychiatric  interview  consist  of
07500	expressions like:
07600		(2) DR.- TELL ME ABOUT YOURSELF.
07700		(3)  DR.-  LETS  DISCUSS  YOUR  FAMILY.  Such imperatives are
07800	actually interrogatives to the interviewee about the topics they  refer  to.  Since
07900	the  only  physical  action  the  model  can  perform  is to `talk' ,
08000	imperatives  should  be  treated   as   requests   for   information.
08100	DECLARATIVES
08200	
08300		In  this  category  we lump everything else. It includes
08400	greetings, farewells, yes-no type answers, existence  assertions  and
08500	predications made upon a subject. 
08700	
08800	AMBIGUITIES
08900	
09000		Words have more than  one  sense,  a  convenience  for  human
09100	memories  but  a struggle for language-analysing algorithms. Consider the
09200	word `bug' in the following expressions:
09300		(4) AM I BUGGING YOU?
09400		(5) AFTER A PERIOD OF HEAVY DRINKING HAVE YOU  FELT  BUGS  ON
09500	YOUR SKIN?
09600		(6) DO YOU THINK THEY PUT A BUG IN YOUR ROOM? In (4) the term
09700	`bug'  means  to  annoy,  in  (5) it refers to an insect and in(6) it
09800	refers to a microphone used for hidden survellience. Some common words  like
09900	`run' have fifty or more common senses. Context must be used to carry
10000	out disambiguation, as described in 00.0. Also we have the  advantage
10100	of an idiolect where we can arbitrarily restrict the word senses. One
10200	characteristic of the paranoid mode is that no matter in  what  sense
10300	the interviewer  uses  a word, the  patient may  idiosyncratically
10400	interpret  it in some  sense relevant to his  pathological  malevolence
10500	beliefs.   		ANAPHORIC REFERENCES
10600	
10700		The common anaphoric references consist of the pronouns `it',
10800	`he', `him', `she', `her', `they', `them' as in:
10900		(7) PT.-HORSERACING IS MY HOBBY.
11000		(8) DR.-WHAT DO  YOU  ENJOY  ABOUT  IT?  The  algorithm  must
11100	recognize  that the 'it' refers to `horseracing'. More difficult is a
11200	reference more than one I/O pair back in the dialogue as in:
11300		(9) PT.-THE MAFIA IS OUT TO GET ME.
11400		(10) DR.- ARE YOU AFRAID OF THEM?
11500		(11) PT.- MAYBE.
11600		(12) DR.- WHY IS THAT? The `that' of (12) does not  refer  to
11700	(11)  but  to  the  topic  of  being  afraid  which  the  interviewer
11800	introduced in (10). Another  pronominal  confusion  occurs  when  the
11900	interviewer uses `we' in two senses as in:
12000		(13) DR.- WE WANT YOU TO STAY IN THE HOSPITAL.
12100		(14) PT.- I WANT TO BE DISCHARGED NOW.
12200		(15) DR.- WE ARE NOT COMMUNICATING. In (13)  the  interviewer
12300	is  using  `we' to refer to psychiatrists or the hospital staff while
12400	in (15) the term refers to the interviewer and patient.
12500	
12600		TOPIC SHIFTS
12700	
12800		In the main a psychiatric interviewer is in control of the
12900	intervie. When he has gained sufficient information about a topic,
13000	he shifts to a new topic. Naturally the algorithm must detect this
13100	change of topic as in the following:
13200		(16) DR.- HOW DO YOU LIKE THE HOSPITAL?
13400		(17) PT.- ITS NOT HELPING ME TO BE HERE.
13500		(18) DR.- WHAT BROUGHT YOU TO THE HOSPITAL?
13600		(19) PT.- I AM VERY UPSET AND NERVOUS.
13700	
13800		(20) DR.- WHAT TENDS TO MAKE YOU NERVOUS?
13900		(22) PT.- JUST BEING AROUND PEOPLE.
14000		(23) DR.- ANYONE IN PARTICULAR?
14100		In (16) and (18) the topic is the hospital. In (20) the
14200	
14300	topic changes to causes of the patient's nervous state.
14400		When a topic is introduced by the patient as in (19),
14500	a number of things can be expected to be asked about it. Thus 
14600	the algorithm can have ready an expectancy list which , combined
14700	with an anaphora list, allows it to determine whether the topic
14800	introduced by the model is being responded to or whether the interviewer
14900	in continuing with the previous topic.
15000		Topics touched upon previously can be re-introduced
15100	at any point in the interview. The memory of the model is responsible
15200	for knowing what has been discussed.