{
    "paper_id": "J78-2002",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T03:04:52.407630Z"
    },
    "title": "THE D,ERIVATICN OF M S W E R S FROFl LOGICAL FORMS I I J A QUESTION AHSWERIIIG SYSTEW",
    "authors": [
        {
            "first": "Fred",
            "middle": [
                "J"
            ],
            "last": "Damerau",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Thomag J Watson Research Center Yorktown Heights",
                "location": {
                    "settlement": "New York"
                }
            },
            "email": ""
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "This papex descrsbes how the process 05 g e n e~t i n g a response g i v e n an underlying representation f o x an i n p u t q u e s t i o n is accomplished in t h e Transformatioaal Question Rnswering [ P A ? s y s t e m under development a t IBM Research, a b r i e f description a f which is g i v e n .",
    "pdf_parse": {
        "paper_id": "J78-2002",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "This papex descrsbes how the process 05 g e n e~t i n g a response g i v e n an underlying representation f o x an i n p u t q u e s t i o n is accomplished in t h e Transformatioaal Question Rnswering [ P A ? s y s t e m under development a t IBM Research, a b r i e f description a f which is g i v e n .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "The last formal level o f representation in this system is called a l o g i c a l form. The b a s~c method of evaluation of logical forms is the generate and t e s t \" paradigm, used, f o r q~& n p l s in the LUNAR system (Woods, Kaplan and Nash-Webber, 1 9 7 2 1, a l t h b u g h t h a t implementation must be fairly efficient in order t e be j~a c t i c a l on a moderate size d a t a base. The 1 Semant~c, interpreter I <-----Sernhntic r u l e s ( f a r a l l ' X I 1 5",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "-----------------,,J <--I 7 1 I L o g i c a l form(s) 1 I f ---------1 I IEvaluatosl <------------ Data",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "basic idea is t o keep track of t h e e q u i v a l e n c e . r e l a t i o n s h i p s botueen the variables in the l o g i c a l fcrh and a s s o c i a t e d c o n s t a n t s , and use t h i s i n f o r m a t i o n t o dexive from the",
                "sec_num": null
            },
            {
                "text": "( s e a t % ' X 3 8 ' ( q u a n t i t y",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "basic idea is t o keep track of t h e e q u i v a l e n c e . r e l a t i o n s h i p s botueen the variables in the l o g i c a l fcrh and a s s o c i a t e d c o n s t a n t s , and use t h i s i n f o r m a t i o n t o dexive from the",
                "sec_num": null
            },
            {
                "text": "( s e t x 'X34 '(and ( t e s t i c t '9",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "X38",
                "sec_num": null
            },
            {
                "text": "v ( T~~~~~ X 3 4 ' 1 9 7 6 ) ' = ) ( p a r c e l X 3 4 ) 1 1 ( g r e a t e x t h a n XI15 ?25) 1 ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "X38",
                "sec_num": null
            },
            {
                "text": "t h a t , we s t i l L have a considerable way t o go,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "p t o g r s m as w e l l as in the L I S P d e f i n i t i o n 3 f the r u n c t j o n",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "a b j e c t i n the answer s e t , t h i s p r o b l e m naturally does not",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {},
        "ref_entries": {
            "FIGREF0": {
                "num": null,
                "text": "l o g i c a l f o~m s and t h e process hy which candidate s e t s are computed from these forms @re described in c o n s i d e s a b l e d e t a i l * We believe it shoufd n o t be necessary for a c o m p u t a t i o n a l linguistics p r o j e c t t o describe o p e r a h i o n s beyopd the l a s t lev'el of f orma1 representation in ozdex 5~1 an o u t s i d e r t o understand exactly how a system operates s u f f i c i e n t l y well t h a t he.can paedict its b e h a v i o r . Although we have attempted t o achieve ~h i r p a p e r d e s c r i b e s how t h o p r o c e s s o f g e n e r a t i n g a r a s p o n s e g i v e n a n t l~l d e~l y i n g xrsprg!senT:amn fair a n i 1 1 p u t q u e s t i o n i s ~c c c m~l i s h e d A n t h e ~r a n s f o r p n t i o n n l E u c s t~o n A n s w~r i h g ( T P ) 1 s y s trrp u n d o r' c o~~t i n u~n g dr,u@kopmmc?nt a t I A N R e s e a r c h , . TQA h a s beert, o p e r a t i o n a l . 1 a l a b o r n t s z y m o d e f o r s e v e r a l yeers.. T h e s y s t e m is noid i n s t a l l e d i n t h e o f f i c e o f the p l a n n i n g d e p a r t m e n t o l a s m a l l c i t y u h e r e it is u s e d t o a c c e s s t h e f i l e o f l a n d u s e f o x e a c h p a r c e l o f l a n d I n t h e c i t y , ( a b o u t 1 0 , 0 0 0 p a r c e l s ~i t h 4 0 p i e c a s o f d a t a f o r e a c h p a r c e l 1 . The sysytcm is trnlilcrgoin,g r n~d i f i c a t i o n s nncl i n~p x o v s n e n t p x i s x t o a f o r m a l e v a 1 u a t i 0 1 1 s t a g e I A g e n e r a l i z e d f l o w d i a g r a m o f t h & TQA s y s t e m is g i v e n in F i g u r e 1 . Input, f r o m a d i s p l a y d e v i c e o r t y p e w r i t e r -l i k e t e r m i n a l , is f e d to t h e p r e p r o c e s s o x , w h i c h s e g m e n t s 5lle i n p u t c h a r a c t e r stsing a n t o w o r d s a n d p e r f o r m s l e x i c a l l o o k u p . The p r o c e s s of l o o k u p i s c o m p l i c a t e d s o m e w h a t b y a p r o v i s i o n f o r s y n o n y m a n d p h r a s e , x e p l a c e m e n t . W o r d s l i k e \"car\" and \" a u t o m o b i l e w a r e c h a n g e d t o \" a u t o \" , a n 8 strings l i k e \" g a s s t a t i o n w a r e frozen i n t o single l e x i c a l u n i t s ,",
                "type_str": "figure",
                "uris": null
            },
            "FIGREF1": {
                "num": null,
                "text": "t p u t f r o m t h e l e x i c a l l o o k u p i s a l~s t of t z a n s l e a c h t r e e , contaznzng padrk o f s~, e r ! c h i n . b r m a t x o r r , g y n t a c t z c faaturcs and scrmantlc f e a t u t : e s , R S r e q u~r~d A d e s c z~~t~r~n~ o f t h e , l e x i c a l cornpol>cetlt, now a b s u l~t~ l n rts d e t a r l b u t s t i l l valid 111 main o u t l i n e A S g l v a n ln R o b~n s a r r ( t r e e s is i n p u t t o e; s e t 0 3 sFtt~$m t z~~~~!~~a t x~~ d e s c r i b~d l n P l a t h t 1 9 7 4 I . These t~Q * n s f a r n a t~o n s a p~r a t c ! arr g d j h c a n t 1 6~i c~b i t 8 1 1 9~ t o d e a l w i t h p a t f c r n s 0 5 C~.~S S~~L P~S , o r d i n a l n u m b e r s , s t r a n d e d p r e p o J i t % o n s , and t h e l i k e . T h e e f f e c t of t h l s p n a s e is to reduce the nurnber o f s u r f a c e u r f a c e t r e e s i l l r e s u l t -L n an u n d e r l y i n g s t r u c t u r e . This s t r u c t u r e i t s e l f 1s input once agaln t o t h e t k a n s f o x m a t i o n a l recognize^:, uslng a ( s m a l J 1 set of grammar r u l e s t a i l o r e d t o a s p e c i T i c data b a s e t o In f o r m , but r e -f l e c t the p a t i c u l a r rneanmg constraints resulting f r o m the f o r m a t and c o n t e n t of a g i v e n data base. The q u e r y s t x u c t u r e t r e e 1s p r o c e s s e d b y a Knuth-style semantic i n t e x p r e t e r , PeteLck t 1 9 7 7 1 , p r o d u c i n g a l o s i c a l f o r m . A l o g i c a l f o r m can b e s t be t h o u g h t o f , The p r o c e s s of answer e x t r a c t i o n from the data base is a c c r~a p l i s h e d b y a cotnbinatWn of L I S P a n d FLII programs, d e s c r i b e d b e l o w , and an experimental relational d a t a base manactenrent s y s t e m c a l l e d R e l a t i o n a l S t o r a g e System ( R S S ) information f r o m s u c h a data b a s e . All t h e p r o c e . s s i n g m o d u l e s are under the c o n t r o l of a d r i v e r m d u l e , which maintains cornmfinication with the u s milliseconds .of computez time u s e d u p t o the point shown, on a n I B M S / 3 7 0 Model 1 6 8 The s t r U c t u r e s printed are a b r a c k e t t e d terminal s . t t i n g representation o f s t r u c t d r e s which are s t o r e d and manipulated as trees b y t h e",
                "type_str": "figure",
                "uris": null
            },
            "FIGREF2": {
                "num": null,
                "text": "programsc The nonterminal nodes 0 % t h e t r e e , l inf brmatian that: id not shown here .,The number 5 9 1 is a land use code which. in t h e d a t a b a s e , i n d i o a t e s a d r u g s t o r e r and th# l o n g numbers in bhe ansNez are the p a r c e l identifi6rs, ( w a r d -b l o c k -l o t ) .Ffom t h i s b z i c d description, it s h o u l d be a p p a r e n t t h a t the T Q A s y s t e m , c o n s i d e x e da s a b l a c h b o x , is sirn'ilar t d many sthers:. 1.n p a r t i c u l a r , there is a d e s i 9 n a t e d l e v OF L O G I C A L FORM!$ Th.e basic method o f e v a l u a t i o n of l p g i c a l forms is the \" g e n e r a t e and t e s t v paradigm u s e d , f o r e x a m p l e , i.n t h e t h e w o r s t c a s e , all t h e o b j e c t s hnonn t o t h e s y s t e m , to see, if they s a t i s f y . tile query p x e d ' i c a t e s . It is computationally i m p r a c t i c a l e x c e p t foq s m a l l data bases. Our current* v a r i a n t 05 t h i s metllad js much m o r e e f f i c i e n t . The b a s i c i d e a is t o k e e p t r a c k of t h e e q u i v a l e n c e r e l a t i b n s h i p s between the v a x i a b l e s in the l o g i c a l form and associated c o n s k a n t s , and use this i n f o r m a t i o n to d e r i v e t h e e x t e n s i o n s c# the p x e d i c a~t e s c o n t a i n e d i n the l o g i e a l f o r m f % a m tne a a t a b a s e . A s i m i l a r -p x o~o s a l h a s been made b y Reeiter(19761 We do not how.ever, m & R e such e x t e n s i v e use of quekp trgnsformations a s Reiter o u t l i n e d . Some o f t h e predicates a r e p e r f e c t l y o r d i n a r y l i k e q r e a t i ! r t h a n . Some are quantifiers, like f o x :~t d e a s t , which Cakes a limit argument n, an argument P A G E 11 which is a s e t ? and a p r o p o s i t i o n e , and which is true j which is t r u e j u s t in ease its s i n g l e ergurnen-t is a p a r c e l identifier. The h a i n d a t a base related p t e d i c a t e is named t e s t f c t . Referring to Figure Z I it is seen t h a t Cestfot h a s three arguments, The first is w c o n s t a n t or a v a r i a b L e which w i l l be r e p l a c e d b y a c o n s t a n t b e f a y e evaluation, the second argument is a llst whose memtbers determine a p a r t i c u l a r : data base value, and the t h i r d is an o p e r a t o r s p e c i f y i n g t h e r e l a t i o n which must h o l d between t h e i i r s t argumentsand the data base v a l u e d e t e x m i n e d by the 'second argument. The d a t a base oan be t h o u g h t of as a c o l l e c t i o n Q + b i n a r y relations, all shirring the same key. In our applicq,tiob, name and the* k e y which identifies a va1u.e in the r e l a t i o n . The k e y a c t u a l l y has two parts. The second p a r t is a y e a h now unused, a l t h o u g h since the f i l e s in n h i c h we are currently i n t e r e d e d are changed on a yearly basis, we a n t i c i p a t e maintaining and accessing h i s t o r i c a l data. The f i r s t part of the key is the account number m e n t i o n e d above. In gener a 1 , number of s i m u l i f i c a t l o m can be, and in p a r t h a v e b e e n , c a r r i e d o u t on l o g i c a l i o r m s p r i o r t o e v a~h a t i o n . Some pxedicates, f o r e x a m p l e , are essentiazly e m p t y f o r purposes 0 2 e v a l u a t i o n , in that t h e y always e v a l u a t e to true,. c e s s o r assumes t h A t the contents of the % a x e s f i e l d a the s e t x f u n c t i o n i s e v a l u a t e d as soon as the s e m a n u c i n t e r p r e t e x h a s discovered t h a t it has no iree v a r i a b l e s * using the s t a n d a r d e v a l u a t i o n mechanism, and t h e v a l u e , i . e . , a s e t , is substituted f o r s e t x a x p r e s s i o n . Our s y s t e m perFormk s i m p l -i f i c a t i o n s 04' this R i n d in i t s normal mede ( a l t h o u g h it can a l s o d e l a y ~1 1 e v a l u a t i o n s q n t i l a comple%e form has-been b u i l t ) , s o that t h e final logkcal foxm seen b y t h e r e t r i e v a l furlotions during p~o c e S s i w i s usually t h a t shown ia Figure 3 , where the i n n e i set^ has b e e n r e p l a c e d hy the s a t i s f y -i i n g s e t viz t h e p a r c e l i d e n t i f i e r s of the s e t of drug s t o~e s .~ L+ker all t h e a p p l k c @ l e s i m p & i f i r a t h o n s have been donel t h t~ resulting form is p a s s e d t o the e v a l u a t i o n function, E,V,ALU. The P r e -e v a l u a t o r It m i g h t seem t h a t s i n c e the s y s t e m has been w r i t t e n in LISP, it w o u l d 0nJ.y be n e c e s s a r y t o d e f i n e the a p p r o p z i a t e functions and then call 'the r e g u l a x LISP evaluator, ~n s t e a d of a special e v a l u a t o r like E V A L U . WI~ile Chis would be possible, t h e aifficulty w i t h such an a~p r o a c~h can r e a d i l y b~ s e e n b y considering the e m b e d d e d s e t x in F i g u r e 2. The d e s i r e d set o f X 7 s is t h a t set of p a r c e l identifiers f o which t h e a s s o c i a t e d l a n d use c o d e is \" 5 9 I n . t e s t f c t is a predicate which i s t r u e for t h e a p p r o p r i a t e X 7 s r but wha$ 1s the c a n d i d a t e s e t of X75 which should b e t e s t e d ? A t w u r s t , the system m i g h t c o n s i d e r t h e s e t of a l l o b j e c t s it knows a b o u t . A S a b e t t e r c h o i c e , the s y s t e m cou3d i n f e r from the s y n t a x of t e s t f c t t h a t +he c a n d i d a t e s are all m e m b e r s of the s e t of parce'l identifiers, but s { t i l l t h e r e are a l m o s t 1 0 , 0 0 0 P A G E 14 what axe t h e h e l g n t s 0 2 t h e a r u g stores ? 2930 S U R F A C E S ' T R U G T U R E S : 1 . t ( < M M SO'MEI ( T H I N G X I ) ) B E ( T I ? & ( ( H E I G H T SLIT) [ O F ~T M E [ I : D R U G , S T O R E 5 9 1 1 X 7 ) l ' l l ) ? ) 1 . ~B I L &DENTICAL ('THE ( X 4 ( * R F H E I G H T X 4 [ T H E [(DRUG-STORE 5 9 1 ) X7)l 13D k ) ) ) I ( W Q , S O F l E ) ( T . H I N~; x i ) ) RD) 6 5 9 9 Q U E R Y SITRUCTURES: 1 (THE eX4 ( * B D HEIGHT X 4 ( T H E :(DRUG-STORE 5 much b e t t e r a p p r o a c h is t o a t t e m p t ' t o c o m p u t e the e x t e n s i o n of those p r e d i c a t e s for which t h e v a r i a b l e b e i n g s o u g h t IS an a r g u~e n t Again r e f e x r~n g to Figure 2, a PAGE 15 r e a s o n a b l e set tin fact the perfect set) of candidates f o r X'7 can be found by. Looking i n t l l c data base f o r t h a t s e t of i d e n t i f i e r s fox which t h e l a n d u s e c o d e is 5 9 1 If the data b a s e i s p r o p e r l y organized, s u c h a search can b e v e r y z a s t Not a l l p r e d i c a t e s are s o s i m p l e h o w e v e r . The remainder o f this s e c t i o n will describe. i n some detail llbw c a r a d i d a t e sets f o r more complicated p % e d i c a t e s are rived at. Once can'di3ate s e t s hav.e b e e n computei~ t h e EVALU function can invoke *he LISP e v a l u a t o r o d t l l a logical form. T~E ! alternative of i n c l u d i n g a candidate generatow 'in t h e s e t x program and a12 the ~d t e n t i a l t o p level p r e d i c a t e s and t h e n a p p l y i n g t h e LISP E V A L f u n c t i o n directly s e e m s much l e s s a t t r a a t i v e . As a pxeliminary, notice that w e need o n l y i p s u x e that candidate s e t s have b e e n e s t a b l i s h e d fd'r a l l the setx variables in a logical for111. T h i s is so Because, while e a c h quantifier has an associated v a r i a b -l e , the domain o f that the l o g i c a l 5orm f o z the question, \"What drug s t o r e s are l o c a t e d i n wazd 8 ? v The p r d d i c a t e of t h e s-e& is \" a n d v T , b u t f o r pu~pposes o'f: r a t i o n of the re-evaluation function. Pre-evaluation is accomplished b y a f u n c t i o l i EVALUA., which t a k e s a l o g i determines the t y p e of form w i t h which it is dealing and calls an appropriate s p e c i a l i s t roufihe If as in the c a s e of the llandlr of Figure 4 , the logical form being considered c o n t a i n s more than ohe component form, , l e f t -t o -r i m h t process. The function always z e t u r n s nil, a woxk b e i l l y a . c c z l i i i p l~~. .~, hy changes t o g l o b a l v a x i a b l e s . Among t h e s e are a LISP variable which PAGE 1 7 c o n t a i n s a l i s t of all set* v a r i a b l e s in the logical foxm, a LISP v a r i a b l e which lists each query v a r i a b l e for which a v a l u e has been founiir and i t s value, and a LISP v a x , i a b l e which k e e p s track 0% the e q u a l i t y r e l a t i o n s h i p s which have been d i s c o v e r e d between query v a z i a b l e s for which a value is yet to be found. Operation of the aZgorithm can be better understood by considering somewhat more complicated e x a m p l e s than t h o s e s e e n p t e v i o u s l y . When E V A L U R is given the l o g i c a l 9 o r m 03What psrcels h a v e an area exceeding 5",
                "type_str": "figure",
                "uris": null
            },
            "FIGREF3": {
                "num": null,
                "text": "it c a l l s the s e t x s p e c i a l i s t , which adds X2 to the ( n u l l ) list of s e t variables and the (null) list of query v a k i a b l e s , and c a l l s EVALUA w i t h the a s s o c i a t e d s e t x p r e d i c a t e , :'andvr. As mentioned., t % i s s i m p l y results in t w o PAGE 13 c a l l s to E V A L U A , the S i r s t o f which causes the q u a n t i f i e r S p e d l a l i~t t o be i n v o k e d . (The second call, when m a d e , w i l l not c a u s e any change t o the global lists oE c a is n o t u s e f u l f o r p u r p o s e s of r e t r i e v a l . ) X 3 9 is a d d e d to t h e l i s t of q u e r y v a r i a b l e s , and the domain argument of the quantifier is i n s p e c t e d . When this is s e e n te b e an i n s t a n c e of s e t x rather -t;han a list of constants, two actions are taken. Notice that whatever the domain, o f X 3 9 19, it is a s u b s e t ( p e r h a p s not a p r o p e r s u b s e t , ) specialist fox t e s t f c t . Since there axe t w o v a r i a b l e s in t e s t f c t , X5 and X2, for whfch v a l u e s are unknown, a s c a l l to t h e data base c a n n o t yet b e made. The i n s t a n c e of testfct is p l a c e d -o n a l i s t of p e n d i n g l a t a b a s e calls, The s p e c i a l i s t for numeric p r e d i c a t e s , finding t h a t one argument i s a v a r i a b l e and the o t h e t a constafbct, causes a can be used by t h e d a t a base componen* L o narrow i t s search j u s t a s w e l l a s a constant or l i s t of c o n s t a n t s , and i s therefore a c c e p t a b l e a s the v a l u e o f a candidate l i s t -. These changes t o the v . a r i a b l e i i s t s cause t h e list of pending data base calls to be inspected a n d , s i n c e only o n e v a r s a b l e is now unknown in the stacked t e s t f c t , a call to the d a t a b a s e is made for those pascels with an area g r e a c e l s which have an area e x c e e d i n g 5 5 0 , 0 0 0 squaxe feet. This list is t h e n a s s i g n e d as the value of the c a n d i d a t e s e t f oxm as an argtrment . The case of n e s a t i v e s ., The p r e d i c a t e w n o t w p d e n o t e d in our s y s t e m b y n o t * to distinguish it f r o m the LISP not, p r e s e n t s s p e c = a l problchs f o r the k i n d of s y s t e m o u t l i n e d above. # s i m p l e exa!nple 05 t h e difficulty can be seeh L~I What drug s t o r e s are n o t in C r a f",
                "type_str": "figure",
                "uris": null
            },
            "FIGREF4": {
                "num": null,
                "text": ". . . . . . . . . . . . . . . . . .. . . .",
                "type_str": "figure",
                "uris": null
            },
            "FIGREF5": {
                "num": null,
                "text": "which corresponds t o the question \"What drug s t o r e s a r e not l o c a t e d In traffic zone 6 ? \" and v a r i a n t s t h e r e o f . \"When the t e s t f c t specialist is g i v e n the first h a l f of the anq in this form, a l o n g with i n f o r m a t i o n t h a t therq is a dominat3ng no**, it could in p r i n c i p l e g e n e r a t e a data base c a l l . , since there i s on1.y one unassigned v a y l a h l e . The e f f e c t would be the r e t r i e v a l of a l l p a r c e l i d e n t i f i e r s of p a r c e l s n b t l o c a t e d in t r a f f i c zone 6 . This is a subsmt;a~.rtial f r a c t i o n of the dadta' base, and would require i n~r d k n a t e amounts of time and s t o r a g e s p a c e to h a n d l e Notice t h a t the o t h e r half of the and d l 1 a l s o provide a candidate l i s t for the v a r i a b l e L3, presumabfy much smaller in s i z e . It appeaxs to be t h e c a s e ? from our s o f a r l i r n i % e d experience, that questions containing o n parcel id en ti fie^. When the second h a l f of the and of Figure 6 is ~X Q G~S S~~, and a value found f o x X 3 , the d e f e r r e d t e s t f c t w i l l be u n s t a u k e d , r e s u F t i n g in a data base c a l l , and c a u s i h g a r e t r i e v a l b a s e d on t h a t list ok i d e n t i f i e r s rather than on the n e g a t e d v a l u e . This data is also an illustration of why, as was m e n t i o n e d above, the l o g i c a l form as a whole m u s t i n g e n e r a l be e v a l u a t e d b y the LISP e v a l u a t o r . In this case, t h e c a n d i d a t e set f a r X3 d e r i v e d from t h e second c l a u s e of the PAGE 22 and is a s u p e r s e t 05 the answer s e t whicn c a n anl'in those c a s e s where At is u l~n e d e s s a t y , h u t thnt is p u r e l y an implementation d e u i s i o n The rl-ot-f of F i g u r e 7 p r e s e n t s a d L f f e r e n t kind o f pxoblem f t a w many banks have a h e i g h t not e x c e e d i n g 9 floors ? a t i v e must be p a s s e d i n s i d e the q u a n t i 4 i e~ s i n c e the a l t e r n a t i v e of & i n d i n g a l l buildings g r e a t e r than 5 stories in lreight and t h e n g e t t i n g t h e complement s e t with r e s p e c t t o c all b u i l d i n g s is e x t r e m e l y unattractive c o n b p u t a t i o n a l l y . In the sgcond placer a s e a r c h q u a l i f i e r of \"(= 5\" does not intuitively seem t o b a much w o r s e than '9 5 \" . at l e a s t in the absence o f data base d i s t r i b t t t i b n a l s t a t i s t l c s .",
                "type_str": "figure",
                "uris": null
            },
            "FIGREF6": {
                "num": null,
                "text": "Oux pxesent e y s t e m does thisl although experience hay show t h a t all i n s t a n c e s of t -e s t f c ? dominated b y no= should be d e f e r r e d , a s a t h e c a s e s of\"v-=\" , f o r e f f i c i e n c y r e a~o n s . Other s p e c i a~i s b s M o s t of t h e important s p e c i a l i s t r o u t i n e s in Ehe pre7evaluator have a l r e a d y been mentioned. There are a few othezs which s h o u l d be noted. In principle, one could, g i v e n a p r e d i c a t e XiHe \"SCHOOL(X)tlr g e n e r a t e a l i s t 0 5 schools. Tn the p z e s e d t applioation, this would n o t be u s e f u l , b u t might in soqe other. The s o l e uses at present a,re q g e n e r a t o r f o r The p r o p o s i t i o n \" [ Q U A~T I T Y x s l v l is t r u e if #is e q u a l ko the c a r d i n a l i t y of the set, 5 : The a s s o c i a t e d specialist hhs the obvious functiorr; of determining when g is an i n s t a r s i a b l e s can o n l y x e f e r t o individuals, the p r e d i c a t e rvMEMBER'T a r s o i s in this class e . g., g i v e n ( M E M B~R X 3 ( S Srom instances 05 testfct with an o p e r a t o r of \" -= \" are d e f e r r e d u n t i l enough information i s available t o e x e c u t e the q u e s y u s i n g a l i s t of p a r c e l identifiers. Some o t h e r s t e p s have also been taken t o reduce daka base a c c e s s time and s the r e l a t i o n h~~~~w d i v i d e s t h e p a r c e l s of the city i n t o 6 c l a s s e s , while the r e l a t i o n \"XUC\" [ L a n d Use C o d e ) divides the p a r c e l s inte several hundred classes. If there is no i n t r i n s i c r e a s o n f o r o r d e r i n g the i n s t a n c e s of t e s + f c t differently, t h e o n e w i t h lvLUC'q w i -1 1 o c c~ ear lie^ in the logical f o r m J (cf. Figure 4 1 . The pre-evaluation s p e c i a l i s t P A G E 25 Sox t e s t f a t m a k e s use of this ordering in t w o w a y s . If a Gariable has b e e n a s s i g n e d a list of identifiers containing f e w e x membersthan some thxeskold x, is currently smt t o 2 5 , but can e a s i l y be c h a n g e d ) , t h e n a r e t r i e v a l wlll alwa-ys be made using the l i s t of i d e n t i f i e r s r a t h e r than b y a constant compared. t o data base v a l u e s . to the ,.test-Ect specialist uill look up the ward of the foux drug s t o r e s instead of Einding the l~u n d r e d s of p a r c e l s in w a r d 2 . I n some instances, varticularly f a r r e l a t i o n s like Land Use C d e , this may result in mor& d9ta b a s e a c c k s s e s t h a n r e t r i e v i n g a new s e t of keys depending on v a l u e , b u t t h e improvement cannot b e l a r g e . In many o%,ber i n s t a n c e s , there i s a b i g r e d u c t i o n in a c c e s s e s . If the c a q d i d a t e s e t is l a x g e r than.25, r e t r i e v a l w i l l be made using the oonstant, but the length o f the c u r t e n t c a n d i d a t e l i s t is used to limit the number of a c c e s s e s . Thus? if the c u r r e n -f ;~ candidate l i s t is 5 0 , the data base Any number of other e f f i c i e n c y measures c o u l d be a d o p t e d ? and more may be necessary than we now h a v e . For the moment, of loQica1 forms is q u i t e straightforward. H i d d e n semantic e f f e c t s are discussed in the n e x t s e c t k o n ; here we are m a i n l y concerned w i Those members of the c a n d i d a t e s e t fox which the predicate e v a l u a t e s to true are p l a c e d in the solutlon s e t . O p e r a t i o n of the q u a n t i f i e r p r e d i c a t e s is similar to t h a t Sometimes the final f a g i c a l f o r m t o b e e v a l u a t e d bears no o b v i o u s relation t o t h e input q u e s t i . o n , a s in Figure 8 . The u s u a l reason i s t11a.t: a l a r g e amaurP of e v a l u a t i o n was done i n g interpretation. because foxm c o n t a i n e d no free v a r z a b l e s . The &uLL l o g i c a l f0r.m corresponding to Figure 8 Are there more than 25 p a r c e l s in the Carhart neighborhood.?",
                "type_str": "figure",
                "uris": null
            },
            "FIGREF7": {
                "num": null,
                "text": "i g u r e 9 i s % g i v e n in figure 9.P A G E 23The e v a l u a t i g n of t h e p r e d i c a t e t e s t ' f c t is n o t as n b b i o u s as that of t h e othGrs One of t h e d e s i g n g o a l s in thep r o j e c t has been t o make it r e I a t i L o 1 y e a s y t o m o v e from one data base t o another. As p a s t o f t h a t ef5brt, we h a v e a t t e m p t e d t o make the LISP programs, a s c -o n t s n s t e d t b",
                "type_str": "figure",
                "uris": null
            },
            "FIGREF8": {
                "num": null,
                "text": "been t o d e f i n e a l i s t s t r d c t u r e , e s s e n t h l l y n e s t e d b i n a r y r e l a t S a n S , into which the z e a l data st!zucture i s m a p p e d . R e s t r u c t u r i n g i s accomplished b y the PL/I p r o g r a m which serves as the L I S P -RSS i n t e r f a c e . A t the s a m e tune. as the P L / I p r o g r a m r e t u r n s . v a f u e s t o t h e t e s t g c t s p e c i a l i s t d u r l n g tile pre-evaluation phase, it $oxmatS the c o r r e s p o n d i n g d a t a b a s e i t e m s into the sbandard s t r u c h r e a n d w r i t e s them onto a disk f i e In e f f e c t c r e a t i n g a s u b -d a t a base 5 o r t h e particular q u e r y . 0x11~ t h e s u b -d a t a base is u s e d durlng evalugtion o f g l o g l c a l forms, to find v a l u e s corresponding to keys in t h e instances of t e s t f c t . In addition to isolating the X I S P p r o g r a m s from \u20ache z e a l d a t a s t r u c t u r e , this + a c t l c m a k e s it unnecessary f o r any p r o g r a m s called b y t h e e v a l u a t o r to r e -a c c e s s the full d a t a base, w i t h a consequent e f f i c i e n c y gain. s undergone t h e Least m o d i f i c a t i o n in our chang'e of data b a s e from b u s m e s s s t a t i s t i c s t o p l a n n i n g data. Except f o x improvements made t o i n c r e a s e the e f f i c i e n c y of processes which will bw used t o compute the answer t o a query s h o u l d be o b v i o u s at the l e v e l of e v h e r the query structuze or the logical form. We have not, however, been zompletely successful in accomplishing this. In some c a s e s , w e can s e e how it might be done and have n~t g o t t e n around to d o i n g it b e c a u s e of more urgent concexns. In other cases, we can see h~w to ds it, b u t not how Lo do it efficiently. In a f e w cases, it is not c l e a r what Vo do. A p~r o x i r n a t i o n . Consider the sexkence and corresponding l o g i c a l f o r m shown i n F i g u r e J O . ., g and 8 are approxima-t;ely equal to 2 , 14 a h d 1 8 are a p p r o x i m a t e l y equal to 1 0 0 0 . Whether +h$s d d i n i t i a n P s An a d d i t i o n a l example is L A S T Y E A R which is d e f i n e d to be the p r e v i o u s y e a r . Many o t h e z d e f i n i t i o n s of this k i n d have been elimihated in the current v e r s i o n df the s y s t e m . .Answers. It is n o t always obvious what constitutes t h e answer t o 9 q u e s t i o n . C o n s i d e r t h e example i n Figure 1 5 . Both the E n g l i s h question in i t s l i t e r a l reading and the l o g i c a l form would seem t o i m p l y tkat the question would be answered b y p r e s e n t i n g only the numbers in the right hand column of the tahle which is a c t u a l 3 . y p r i n t e d a s an answer. h h a t only n 3.ist o f wards s h o u l $ be t h e o u t p u t , and given \"What is the c o m b i n e d floor area 0 2 t h e drug only a s i n g l e number representAng t n e t o t a l is the d e s i r e d In what wards axe t h e drug s t g s e s l o c a t e d ? r s y s t e m d o e s not a s y e t answer this questioh or i t s analogues, klthougth t h i s is planned f o r l a t e r in t h e d i o u s for questions of t h i s t y p e . , For y e s / n o questions, and far q u e s t i o n s in which t h e r e is adly one PAGE 33 arise, and t h e a p p r a p r i a t e answer is e a s d l y produced.. We h a v e not y e t concexned a u x s e L v e s w i t h adding an E n g l i s h response g e n e r a t o r t b the TQA s y s t e m . In t h e a p p l i c a t i o n s e n v i s i o n e d at p r e s e n t , such a capability d o e s n8t seem t o be c r i t i c a l . We are a b l e t o manage w i t h short answers f r o m t h e d a t a base and with canned information and e s r o r messages. In spite of t h i s omission, it s h in i t , more than we would l i k e . Whether i t i s p o s s i b l e t o achieve a l e v e l af formal representation which would make t h i s unnecessazy is stir1 u n c l e a r . Moreover, e v e n if i-1; weze p a s s i b l q , i t is n o t clear whether such a s o l u t i o n would b e efficient enauyh, or e v e n if St would be m o r e pexspicuous than the current system We i n t e n d to p r o c e e d a s f a r as we than t a b l e s . This is n o t t o i m p w t h a t we regard m o d i f i c a t i o n 0 9 a t a b l e whose size is on the order of a g r a m m a r a s t r l v l a l ; quite the c o n t r a r y . Nonetheless, we b e l i e v e it is e a s i e r t o change-a grammax or PAGE 39 a semantic interpreter e x p r e s s e d in t a b l e form than it is t o change a Special parser o x a s p e c i a l i n t e r p r e t e r . In e s s e n c e , w e b e l i e v e it should n o t be n e c e s s a r y for a computational linguistics p r o j e c t to d e s c r i b e o p e r a t i o n s Beyond the last l e v e l of farma1 r e p r e s e n t a t i o n in order for an o u t s i d e r t o Andexstand e d a c t l y how 'a s y s t e m o p e z a t e s .",
                "type_str": "figure",
                "uris": null
            }
        }
    }
}