{
    "paper_id": "J75-4016",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T02:40:37.963124Z"
    },
    "title": "",
    "authors": [],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "T h i s paper outlines t h e stucture and operation o f the 1 ingui s t i c component from a language generation system in an interactive program. The component receives messages describing what i s t o b e s a i d f o r m u l a t e d i n t h e representation of the main prograr and produces fluent English utterances appropriate t o t h e current discourse situation. T h e component is data-directed and uses a procedural grammar, organized as a set of strategies. Interactive, speclalist prograas presently under developwent will heed to produce fluent, intentional English utterances in responce to particular, complex t i t u t l o n s. This creates a requlroaont for language generating facilities that I s not faced in transformational grarapar, rochanical translation programs, or paraphrase generating programs. As a component of rn interactive, specialist program, the production of t h e English must be driven directlr by the communicative intentions of the program and by the discourse situation, We can imagine t h a t the overall program consist, \u00a7 o f a number of cooperating modulesf o r parsing and interpreting what i s said t o i t , ior solving ptoblens in its domain, for managing i t s renary, and, i n particular, f o r generating u t t e r a n c e s t o c o m~u n i c a t e w l t h I t s users* This generation component can be p r o f i t a b l y v i e w e d as having three aspects or msub-corponentsw. 1) Situation/doaain specie1 i s t s t h a t a r e activated when t h e program recognizes what situation it i s in+ They then decide what message will be produced. They will decide what e f f e c t on t h e listener is",
    "pdf_parse": {
        "paper_id": "J75-4016",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "T h i s paper outlines t h e stucture and operation o f the 1 ingui s t i c component from a language generation system in an interactive program. The component receives messages describing what i s t o b e s a i d f o r m u l a t e d i n t h e representation of the main prograr and produces fluent English utterances appropriate t o t h e current discourse situation. T h e component is data-directed and uses a procedural grammar, organized as a set of strategies. Interactive, speclalist prograas presently under developwent will heed to produce fluent, intentional English utterances in responce to particular, complex t i t u t l o n s. This creates a requlroaont for language generating facilities that I s not faced in transformational grarapar, rochanical translation programs, or paraphrase generating programs. As a component of rn interactive, specialist program, the production of t h e English must be driven directlr by the communicative intentions of the program and by the discourse situation, We can imagine t h a t the overall program consist, \u00a7 o f a number of cooperating modulesf o r parsing and interpreting what i s said t o i t , ior solving ptoblens in its domain, for managing i t s renary, and, i n particular, f o r generating u t t e r a n c e s t o c o m~u n i c a t e w l t h I t s users* This generation component can be p r o f i t a b l y v i e w e d as having three aspects or msub-corponentsw. 1) Situation/doaain specie1 i s t s t h a t a r e activated when t h e program recognizes what situation it i s in+ They then decide what message will be produced. They will decide what e f f e c t on t h e listener is",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "choice of plan is determined by the character o f the event's \"actionw.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The T r~n s l a t i o n Prooesa",
                "sec_num": null
            },
            {
                "text": "The action is \" < f i t person into full schedulerw, and i t will have two relevant properties in the lexicon: \"plan*, and \".appingW.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The T r~n s l a t i o n Prooesa",
                "sec_num": null
            },
            {
                "text": "klan is e i t h e r the nane of a standard p l a n t o be used; or an actual p l a n , over. The p l a n is now as below.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The T r~n s l a t i o n Prooesa",
                "sec_num": null
            },
            {
                "text": "partially",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The T r~n s l a t i o n Prooesa",
                "sec_num": null
            },
            {
                "text": "n 4 e -l (clause trans1 particle) slats fmntfngs \"8aybeW subject twinston> vg node-2 (verb-group par tic1 el slots modal \"canw pre-vb-adv n i l it does nothing, nothing has y e t been printed beyond the verb group, and other heuristics will be free t o a p p l y t o choose the proper position.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The T r~n s l a t i o n Prooesa",
                "sec_num": null
            },
            {
                "text": "Since ( p e r s o n being t a l k e d about, is here equal to the student, t h e person t h e prograa is talking with, i t i s realized as the pronoun \"youw and t h e particle is dlsplaccd.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The T r~n s l a t i o n Prooesa",
                "sec_num": null
            },
            {
                "text": "Going irom <31-10-75,9a~-12arn> t o w t o m~r r~~ ~o r n i n g * m y be little more t h a n table lookup by a wtfme\" coBposer that hat been designed to know the formats of the time expressions inside the scheduler. ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The T r~n s l a t i o n Prooesa",
                "sec_num": null
            },
            {
                "text": "p l a n n f n p , i f f o r no other reason than t h a t the M e s s a g e can o n l y be",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "Barganing Between Goalsw. i h the proceedings n f IJCAI-4, available from the MIT A I l a b , McDonald, D. (1975) The Design o f a Program f o r Generating p a t~.~~ Language",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Genesereth",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "; A Macsyma Advisor",
                        "suffix": ""
                    },
                    {
                        "first": "Mac",
                        "middle": [],
                        "last": "Project",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Mit",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Cambr L D G E",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Mass",
                        "suffix": ""
                    },
                    {
                        "first": "N",
                        "middle": [],
                        "last": "Goldman",
                        "suffix": ""
                    }
                ],
                "year": 1974,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Genesereth, M. (1975) A MACSYMA Advisor. Project MAC, MIT, Cambr l d g e , Mass, Goldman, N. (1974) \"Computer Generation of Natural Language fro. a Deep Conceptual Basee. memo AIM-247, Stanford A r t l f l c i a l I n t e l l i g e n c c Lab., Stanford, Calif, Goldstein, 1. (1975) \"Barganing Between Goalsw. i h the proceedings n f IJCAI-4, available from the MIT A I l a b , McDonald, D. (1975) The Design o f a Program f o r Generating p a t~.~~ Language. unpubl ished Master's Thesl s, MIT D e p t . of El ectlcsl Engi neerf ng.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "Understanding Natural Language",
                "authors": [
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Winograd",
                        "suffix": ""
                    }
                ],
                "year": 1972,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Winograd, T. (1972) Understanding Natural Language. Academic Press, New York, NY.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "uris": null,
                "type_str": "figure",
                "text": "desired, and exactly what o b j e c t s and relations a r e t o be nentioned. F o t example, an a p p o i n t~e n t scheduling program might be told to *scm~ule a group meeting for F r i d a y w and then find t h a t a critical aentber o f the group i s uncxvailable, The situation specialists in t h e scheduling prograr a r e t h e ones t o decide whether i t is more a p p r o p r i a t e t o s i m p l y say \" 1 can'tR, O F w h e t h e r t o v o l u n t c p informationw I can't; Mitch won't be back u W l t Mondayn. 2) Models of t h e audience and t h e d i s c o u r s e s i t u a t i o n to u s e i n construct2ng utterances. There m u s t be a r e c o r d o f t h e p a s t conversation to gulcfe in the selection a f p r o n o u n s , A l s o , the program must have nodels of, and heuristics about what the audience a l r e a d y knows a n d t h e r e f o r e doesn't have t o be t o l d . T h i s Informtion l a y be very specific and domain dependent. Fot exarple, In chess, one can say \"the white queen could take e knightn. T h e r e is no need t o say \"a black k n l p h t w , because t h i s information is supplied by inferences from what one knows about c h e s sinferences that the spcarer assures the listener shares. 3) Llnpu!rtic know1 edge about how to construct understandable utterances in the English Isnpuagc. Obviously, t h i s lnformatlon vill Include a lexicon assoclatlng objects and relations from t h e min program with rtrate&i@~ for realizing them in English (particular words, p h r a s e s , syntactic constructions, etc.). There is also a tremendous amount of informatian which describes the characteristics of the English language and the conditions o f i t s use. It specifies rhe allowable arrangements of strategies and what niodlfications or alternatives t o them nay be appropriate in particular circumstances. Of the three aspects just described, my work has concentrated on the t h i r d . What follows i s drawn from sy thesis McDonald ' 7 5 ) and from ongoing research. The Lingufetio Component The 1 inguistic knowledge required f o r generating utterances i s put into one component whose job i s t o take a message from t h e sltuatlon specialists and cpnstruct a t r a n s l a t l w of t h a t message In English. The messages are in the representation used by the main program and t h e s 1 tuation specialists. Tho translation is done by a data-directed process wherein the elenents and structure o f the message i t s e l f provide the control. The design of the 1 ingui stics component was arrived a t independent of any particular main program, f o r the simple reason that no programs o f adequate complexity were available a t the time. However, a t the p r e s e n t t i m e a g r a m m a r and f c x l c o n is being d e v e l o p e d t o use with a t l e a s t two prograRs being developed by o t h e r people a t MIT. They a r e an a p p o i n t m e n t s c h e d u l i n g program (Goldstein ' 7 5 ) and an advisor t o a i d users of MACSYMA (Genesereth '75). The s h o r t d i a l o g below i s an example of t h e degree of f l u e n c y we arc hoping t o eventually achieve. The d l a l o g ts between a scheduling prograa a c t i n g as an appolntaent secretary (P), and a s t u d e n t ( 5 ) .",
                "num": null
            },
            "FIGREF1": {
                "uris": null,
                "type_str": "figure",
                "text": "I want t o see Professor Winston sometime in the next few days. (PI H e r s p r e t t y busy a l l week. Can i t wait? (S) No, it can't, A l l I need i s h i s signature on a f o r b (PI Well, maybe he can squeeze you in tommorrow ~o r n l n g . Give me your name and check back in an hour. Messages Using the current message f o r a a t and Ignoring t h e d e t a i l s of t h e schedulerts representation, t h e phrase \"maybe he can squeeze y o u in t o~m o r r o w~ could have c o w from a Pessage like t h i s one, p u t t o g e t h e r by one of t h e situation specialists. Message-1 features. ( prediction ) event (event actor (Winston) action ( f i t p e r s o n -i n t o d u l l schedule> t h e (31-10-75, gar-12am) hedge <fs possible) aim-at-audience hedge Messages have features describing t h e program's communicative intentions what sort of u t t e r a n c e is t h i s t o be; what effect i s it t o have. Messages l i s t the objects t o be desert-bed (the r i g h t hand column) along with annotations f o r each object ( l e f t hand co1u.n) to show how they relate t o t h e rest of the message. The phrases on the r i g h t in angle brackets represent actual s t r u c t u r e s from t h e s c h e d u l e r wf t h t h o s e The Lexicon Translatl~n ftor tho internal reprcscntai ton o f s coaputcr program",
                "num": null
            },
            "FIGREF2": {
                "uris": null,
                "type_str": "figure",
                "text": "has t h e same sort o f p r o b l e w as translating b e t w~e n t w o natural languages. T k same concepts nay not be a v a i l a b l e a s prim1 t i v o s in both rcpresents.tions, and the conventions o f t h e target Isnguape my require additional information t h a t was not in the source.Generally speaking translation cannot be one f o r one.",
                "num": null
            },
            "FIGREF3": {
                "uris": null,
                "type_str": "figure",
                "text": "English phrase is best for a particular element in a program's message will depend on what i s in t h e r e s t of the message and of what the external c o n t e x t is. In such circunstances, translation by tablelookup is inadequate. In this component, in order t o allow a l l factors t o be c o n s i d e r e d , t h e t r a n s l a t i o n o f e a c h element ias d o n e b y individualized procedures called wcorposersH.For each main program t h a t t h e linguistic component becomes associated with, a lexicon must be created which will list the elements of the rain program's representation that c o u l d appear in a message (1. C. w p r C d l~t l~n H , \" e v e n t w , w < W i n s t o n P , e t c . ) . W i t h each element i s recorded the composer t h a t will be run when the time comes t o produce en English description f o r it (examples will be given s h o r t l y ) . Some conposers nay be applicable f o r a whole c l a s s o f elements, such as \"eventsw. They would know the structure t h a t a l l events have in common ( e . g . actor, actlon, tine) and would k n o w how t o i n t e r p r e t t h e idiosyncratic d e t a i l s o f each event by using d a t a in the lexicon associated with them. The Grammarstrategies The bulk o f the g r a u a r con&ists o f \"strategiesw. Strategies arc associated with particular languages rather than w i t h particular main prograas as composers are. A g i v e n s t r a t e g y may be used f o r several d i f f e r e n t purposes. A typical case i s the strategy u s e -s , ? m p~r e s e n ttense: a clause in the simple present (\"prices risew) my be understood as future, cond'itlonal, or timeless, according t o what other phrases arc present, Each composer ray know of several strategies, or Coabinatlons of strategies which it could use in dcscrlbing an ele.cnt f r o m the message, It w i T 1 choose between the& according t o the c o n t e x tusually d e t a i l s o f the element or syntactic constraints iaposed by previously selected strategies. T h e strategies themselves do no reasoning; t h e y are implemented as functions which the corposers call t o do a l l the actual corsst'ruction o f the utterance.",
                "num": null
            },
            "FIGREF4": {
                "uris": null,
                "type_str": "figure",
                "text": "t t h i s p o i n t . the out1 ine of t h e data-dr Iven translation process can be su~rnriztd. A message is glven f o r translation. The e l e~e n t s of t h e aessage are associated i n a lexicon w i t h procedures t o d e s c r i b e the.. The procedures a r e run; they c a l l grs~latica1 strategies; and t h e strategies construct the English utterance.Of course, i f t h i s were a l l t h e r e was t o it, t h e process would never run, betause a l l of the subprocesses aust be throughly coordinated i f they are n o t to \" t r i p over t h e i r own feet\", or, for t h a t ~a t t e r , i f ordinary human beings are t o bo able t o design the..Ina system where t h e knowledge o f what to do i s d i s t r i b u t e d over a l a r g e number of separate procedures, control structure assumes central. iaportance. Plans Before d t~c r i b i n g t h e control structure, I m u s t l a y o u t some additional aspects o f the design of t h e ilngulstlcs coaponent. There i s no i n t e r l i n g u a or intermediate level of structure c o~p a r a b l c to the dkep structures of Transformttlonal G r a~s a r , o r the s e w n t l c n e t s of S l~a o n s (73) or Goldban ( 7 4 ) . Detorminln~ t h e appropriate sutface s t r u c t u r e , however, r e q u i r e s examined one piece ax a t h e . The e n t i r e utterance must be organized before a detailed analysis and translation can get underway. As this is done, t h e w p r o t o -u t t e r a n c e w i s r e p r e s e n t e d i n t e r n s o f a s o r t of scaffoldinga representatiqp of the ultimate surface structure tree insofar as its d e t a i l s a r e known w i t h e x t e n s i v e annotation, explicit and implicit, t o p o i n t out where elements that are n o t y e t described may be positioned, and t o implement the graamatical restrictions on possible future details as dictated by what has already been done. The scaffolding t h a t i s constructed in the translation o f each message i s called i t s w p 4 a 0 w . Plans are made up o f syntactic nodes of the usual s o r tclauses, noun groups, e t c ,and nodes may have features in t h e Banner of tysteni; grammar Winograd ' 7 2~ Nodes have subplans c o n s i s t i n g o f a l i s t of named slots marking the possible potftlons f o r sub-constituents, given in t h e order ~f t h o eventual surface structure. Possible slots would be w s u b J e c t w , *.%in v e r b w , \"noun headw, wpre-verb-adverbw, and so on. The syntactic node t y p e s will each have a nuaber of -possible plans, corresponding\" t o t h e d i f f e r e n t possible arrangements or sub-consti tuents that may occur wl t h t h o d i f f e r e n t combinations o f f e a t u r e s t h a t t h o Mdc may have, D e p e n d i n g on t h o r t a g e o f t h e t r a n s l a t i o n p r o c e s s , a s l o t may bc \" f l 1 l o d w w i t h a pointer t o an lnternal o b j e c t f r o . the message, a syntactic node, a word or idior, a r nothing. The translation proaaea Tho translation i s done i n two phases. The second phase does n o t begin until the f i r s t i s coapletely finished. During the f i r s t phase, a p l a n is s e l e c t e d and t h e eleaents o f t h e message are transferred, l a r g e l y untouched; t o the s l o t s of the plan and features added t o i t s nodes. During the second phase, the plan is wwalktdR topdown and from l e f t t o r i g h t . Conpostrs f o r mssage e l e~e h t s in t h e plan's s l o t s are activated t o produce English descriptions f o r the elcments as t h e y are reached in turn. B o t h processes a r e data-directed, t h e f i r s t by t h e particular contents o f the message and t h e second by the structure o f the p l a n and t h e contents o f i t s s l o t s . There are sound linguistic reasons f o r t h i s two stage processing. Most parts of a lessage Bay be translated in t e r m s of very modular sysltactic and lexical units. But o t h e r p a r t s a r e translated i n terms o f relations between such units, expressed usually by ordering or clauselevel syntactic aechanislps. The exact f o r @ o f t h e s~a l l e r units cannot be deternlned until their larger scale relations have been fixed. Accordingly, t h e objective of the f i r s t phase is t o determine w h a t global relationships are required and t o choose the plan, features, and positions of message elemnts within the plan's s l o t s t h a t will realfzt those relationships. Once this has been done, Engl tsh descriptions f o r the elements can be made Independent of each other and will not need t o be changed a f t e r they arc i n i t i a l l y created. O n e o f the l o s t i a p o r t a n t features o f n a t u r a l language is t h e ability t o omit, p r o n o~i n a l i z e , or o t h e r w i s e abbreviate elements in certain contexts. T h e o n l y known r u l e s and hllristlcs f o r using this feature r r o phrased in terms of Surface structure configurations and temporal ordering. Because the second ~h a s e works d i r e c t l y in thcse terms, s t a t i n g a n d u s i n g t h e a v a i l a b l e h e u r i s t i c s becoaos a s t r a i g h t \u00a3 orwsrd, tractable problem. \"Maybe h e can eg.ueeze you in tommoww morning\" T h e rest of this paper will try to put solae f l e s h on your p i c t u r e of how this linguistics conponent works by following the translation of t h e message given in the beginning t o the sentence above. -audience hedge The intentional features of a message tend t o require the n o s t g l o b a l representation in the f Inal utterance, because that i s where indicators for questions, special emphasis, speclal formats lee n. conpari son), and t h e like w i l l be found. By convention then. the composers associated w l t h the intentions are given the job of arranging for the disposition of a l l o f the messake elements. T h e t o t a l aperatlon of phase one consists o f executing the composer associated with each feature, one af $er the other. Thls aessage has only one feature, so i t s composer will assume a l ) tho work. Thc linguistics component is implemented in MACLISP, features (and annotations and s l o t s and nodes) are atoms, and coar>Qsers ass functions on t h e i r property l i s t s .",
                "num": null
            },
            "FIGREF5": {
                "uris": null,
                "type_str": "figure",
                "text": "is a speech act, an& wo nay expect there t o be particular forms in a language for expressing thee, f o r example, the use a f t h e explicit \"willw for the future tense. Knowledge o f these would De part o f t h e composer. Inside the aain program, or t h e situation special l s t , the concept o f a prediction may always i n c l u d e certain parts: what is predicted, the time, any hedges, and so on. These part are d i r e c t l y r e f l e c t e d i n thc makeup of t h e elements present in the message, and t h e i r annotations mark what internal r o l e s t h e 7 have. There does n o t need to be a d i r e c t correspondence between these and t h e parts in the linguistic forms used, the actual correspondence i s p a r t of the knowledge of the prediction cosposer. Typically, f o r any feature, one particular annotated e l e~e n t will be of greatest l~p o r t a n c e in seting t h e c h a r a c t e r o f t h e w h o l e utterance. For predictions, this is t h e \"eventw. T h e prediction composer chooses a plan f o r the utterance t o f i t the requireaents of the event-element. The realization of any other elements will be restricted t o be compatible w i t h i t . T h e prediction composer d o c s not need to k n o w t h e element's linguistic correlates i t s e l f , i t can delegate the work to the composer for t h e element i t s e l f . The element look like this, (event actor <Winston> action < f i t person i n t o f u l l schedule> tine t31-10-75,9am-lZaa>)The first word points to the name of the composer, and the p a i r s g i v e particular details. There i s nothing special about the words used here (actor, action, tlme), Just a$ long as t h e composer is designed to expect the inforwatton in those places that the message-asseabler wants to p u t it. The event composerfs strategy is to use a clause, and t h e",
                "num": null
            },
            "FIGREF6": {
                "uris": null,
                "type_str": "figure",
                "text": "f i l l e d w i t h words (1. e. i t can be a phrase). \"Mappingw is an association l i s t showing how t h e subelements of the message are t proceeds t o instanticte the nodes i n the phrase and make the transfers; the prediction composer then takes the resulting plan, and makes i t the plan o f the whole utterance. Two message elements remain, b u t actually there is o n l y one, because waim-at-audiancew is supplying additional informati~n about the hedge. T h e annotation means t h a t the c o n t e n t s o f t h e hedge (<is possible>) ere pore something t h a t we want to tell the audience than a detail o f the prediction. This will a f f e c t how the element is positioned in the plan. T h e p r e d i c t i o n c o m p o s e r l o o k s in t h e l e x i c o n t o s e e w h a t grammatical unit will be used t o realize <is possible,, and sees, l e t us say, two possibilities involving d i f f e r e n t configurations of the adverb wn&ybcn and t h e modal \"can be able ton, with the differences hinging on t h e placement of the a d v e r b . Theoretically, a d v e r b s can be p o s i t i o n e d i n a nunber o f p l a c e s in a clause, depending on t h e i r characteristics. In this instance, the choice is forced because of a heuristic written i n t o the grammar of adverbs and accessible t o t h e composer, that says t h a t when the intent of an adverb is directed t o the audience, i t should be in t h e f i r s t position ( t h e \"frontiogsW s l o t ) . This choice implies p u t t i n g \"canw I n t h e modal s l o t d i r e c t l y . T h e alternative with w~y b e * i n the pre-vb-adv s l o t would have necessitated a different form of the .o8al, yielding \" r a y be able t o n , These details would have been taken care of by syntactic routines associated with t h e verb group node. A 1 1 the message ererents have been placed and t h e f i r s t phase is",
                "num": null
            },
            "FIGREF7": {
                "uris": null,
                "type_str": "figure",
                "text": "The second phase controller is a simple dispaching function t h a t m v e s from s l o t t o slot. \"Fronttngs* contains a word, so t h e word is printed .directly (there is a trap f o r morphological adjustnents when necessary).wSttbjectw contains an i n t e r n a l object, so the controller should go to the lexicon for its composer and then come back t o handle whatever t h e composer replaced the clement with, However, there is always an l h t e r v e n i n gs t e p t o check for t h e possibility of pronominalizing. This check is made w i t h the elelpent s t i l l in i t s internal f o r o . T h e record o f t h e discourse i s g i v e n d i r e c t l y in t e r m o f the i n t e r n a l representation and test f o r p r i o r uccurence can be as simple as identity checks against a reference list, svoiding potentially intricate string matching operations w i t h words.In the d i a l o g t h a t t h i s message came from, t h e r e is clear reference to t w i n 9 t o n > , so it can be prononinallzed and \"hew is printed.Any slot, or anynode t y p e may have procedures associated with i t that are executed when t h e slot or node is reached during t h e second phase. These procedures will handle s y n t a c t i c processes like agreement, rearangelaent o f s l o t s t o r e a l i z e features, add function words, watch scope relationships, and in particular, position the particle in verbparticle pairs. Generally, p a r t i c l e position (\"squeeze John i n n vs.n~q~m e in J o h n w ) is n o t specifled by the grammarexcept when the object i s a pronoun and the particlemust be displaced. This, of course, will n o t be known untlll a f t e r the verb group has been passed. To deal with this, a subroutine in the \"when-ent$redn procedure o f t h e verb group i s activated by the \" p a r t i c l e n procedure. First, i t records the particle a n d relaoves i t f r o m t h e V G p l a n s o it will n o t be g e n e r a t e d automatically. A \"hookw i s available on any slot for a,procedure which can be run after prononinalization is checked and before the composer i scalled ( i f i t is t o be c a l l e d ) . T h e s u b r o u t i n e incorporates t h ep a r t i c l e i n t o a standard procedure and places i t on t h a t hook for t h e object1 s l o t . The procedure will check i f the object has been prlnted as a pronoun, and i f so, p r i n t s o u t the p a r t i c l e (which i s now I n t h e proper displaced pori tion).If the o b j e c t wasn' t pronominal i z e d , then",
                "num": null
            },
            "FIGREF8": {
                "uris": null,
                "type_str": "figure",
                "text": "had t o be u n f o r t u n a t e l y s h o r t for the amount of n e w naterial involved. A large n u~b e r of interesting d e t a i l s a n d questions about t h e processing have had t o be oritted. A t t h e moaent <September, 19751, the data and control structures ~c n t i o n e d have been fully iep1e~ented and t e s t s are underway on gedanken data. Hopefully, by the end of 1975 the component will have a reasonable g r a m a r and will be working with messages and lexicons form t h e t w o programs mentioned before. A MI7 A. I. l a b technical r e p o r t describing this work i n d e p t h should be ready in the spring o f n e x t year. David McDonald Cambridge, Mass.",
                "num": null
            }
        }
    }
}