A discourse production model for “twenty questions”

A discourse production model for “twenty questions”

Michael 0. Fortesc:ue, A discourse pmducticm model fin “7’~ent~v Qtmtions” ,(=Prag;ma-;ics and Beyond 2). Amsterdam: John BenjaMs, 1980. 137 pp. HfI. ...

436KB Sizes 0 Downloads 78 Views

Michael 0. Fortesc:ue, A discourse pmducticm model fin “7’~ent~v Qtmtions” ,(=Prag;ma-;ics and Beyond 2). Amsterdam: John BenjaMs, 1980. 137 pp. HfI. 30.00 (paperback). This abbreviated version of the author’s doctoral thesis offers a detailed theoretical mode\ of the cognitive processes involved in the produciian of conversational exchanges during the playing of the familiar game of Twenty Questions, in which BL player has twenty questions to discover the identity of an object his oppsncnt has in mind. The author’s approach is, from the analysis of a ~ax~pJ#e corpus of games, to isolate the different types of utterance and speech exr.hange to be found, to postulate higher levels of organisation in which they can be located and then to offer 3 description orf a hypothetical computer prograrmne which would incorporate this information, and which, if supplemented with straic,oies for the characterisatron ant! checking of propositional content and other relev& non-linguistic inr”ormation, would ble in principle capable of simulating a game. Bn the first chapter Fortescue identifies recurrent types of linguistic form in his corpus (e.g. ‘Tags. ‘what about . . .’ questions) and produces a taxonomy of ‘dis course acts’ which are typically performed by these utterances, such as ‘Request tnhrma t icn’, “Offer help’ or ‘Signal intent’. Chapter two groups common pairings of these discourse a@tsinto ‘response sequences’ and associates them with typical propositional contents and characteristic linguistic realisations. I;ol-tescue argues that this relatively complex taxonomy involving considerable overlap and redundan:v, is necessary, for an alternative classification (based for example on Searle 1969, 1945) would be too gross at the level of discourse acts and inappropriate for his ?tsrpose at the level of response sequences. However, in chapter three he allows that something equivalent to Searle’s account of indirect speech acts and G-ice’s principles of conversational reasoning will be needed to account for cases where a discourse act is realised non-standardly (p. 52). Fortescue’s claim at this point seems to be that such complex ‘computations of intent” will only be necessary in a very few cases, as the narrow confines of the game in question will make just one of the many logically possible discourse acts associated with a particular form over~he~~in~y the most probable candidate, and this fact is to be encoded in the ‘ruleIs’for the realisation and recognition of discourse acts (p, 55). C’hapter fm.u argues that a descriptive scheme of the type offered by Sinclair and rd (1975) is inadequate to account for the game samples in the corpus, and claimed that any approach which is basically ‘hierarchical’ will likewise be ate. Chapter five introduces modified versions of the notions of ‘frames’ 1975) and ‘Jemons’ (Charniak 1972). They differ considerably from their ‘Frame’ for Fortescue means, in essence, ;I higher level schema which ales, accurring corMnations of response sequences and discourse acts: thus we me frames’ and ‘discussion frames’. ‘Demons’, in this usage are routines d with d&course acts which trigger off the realisation of the appropriate of the response set associated with that act in a response sequence. An

utterance will trigger a particular demon if it is recognised ; E the ap,Dl.spriate discourse act. Chapter SIX outlines the hypothetical mode of interaction bc:rween the various components of the model, with sornfj illustration, and chapter ;Luen discusses the various strategies that questioners use in deciding what to ask ;Jasical_lythree tyybes of approach (i) top-down: narrowing down the field of Possibilities by questi(,ns like ‘IS it male?‘; (ii) ‘bottom-up’: questions about specific properties of individilals . Ior classes iarrived ai; by (i) or by guesses: (iii) “useful question type& the experienced game-player will have a repertoire of questions which have often proved productive, e.g. ‘Is it decorative?‘, ‘Do 1 know one?’ etc. It is very difficult to assess the success of this monograph because, although the aim is stated boldly, “TO build a plausible model of the cognitive processes behind the behaviour e ibited by speaker-hearers in a specif;: Clis:r>ursesituation” (p. l), it is not clear what evidence is being presented l’or its plausbility. At *variouspoints Fortescue indica es that psychological plausibility is being considered (pp. 8, 61, 104 etc.), but no direct evidence of this is offered, though it is intimated (p, 122). Given the apparently unconstrained nature of the theorctica.1 concepts employed here, it is important that evidence of l:his type should be #orthcoming otherwise all that is being claimed is that there exists a computational mechanism adequate for the job - essentially an engineering clyim. What we need to know is why this particular mechanism should be the on! that is involved, rattler than some other equally adequate but different o!qe. If intuitive plausibility is the aim, weil and gooci: WChave to starl somewhere. Elut intuitive plausibility is a ::lippery fi~!z;while it might be plausible that players might, during the course of
382

Book reviews

the two major points are (i) (p. 43) that in a circumscribedsituation discourse acts

wkch are expressed indirectly can nevertheless be recognised directly, without intervening proceshg of the literal meaning of the utterance, rejecting this, inspeeting felicity conditions, running through conversational maxims, etc., since the ‘*frame’internalised by the speaker-hearer will suggest which acts are most likely at this point in the transaction. This is almost certainly true, but to my knowledge .nobxiy has ever denied that sulch heuristics exist. The point is that they are only heuristics, and so something like the full machinery of a SeaAe-Gricetype theory will still be nee:Ied, as it is in Fortescue’s model, for when ;the heuristics cannot copy There is also a clalim(ii) that discourse analysis approaches which imply that the units and subunits of analysis form a hierarchy are in principle inadequate as a des.;ription of the corpus here, and by implication, more generally unsatisfactory. Since the objec:.ionis raised not only against Sinclair and Co&hard’s work but also against the wori,.o!fGrimes (1975) and van Dijk (1872) it deserves more space than it is given, not lt:ast.to support the claim that these approaches do really imply such hierarchies. As a contribution to theory, then, my reaction to this work is fairly negative. But as a rJcroE&opic&aarQQon of 8 I’u!ij;&dorlah3 Of &&Qurseii is interesting, Careful and detailed and it is perhaps as such that it could most profitably be read.

Steve Pulman University ol^East Anglia Norwich, England. References Char&k, E. 1972. Towards a model of children’s story comprehension. M.I.T., Ph.D. dissertation Cole, P. an? J.L. h+q,an, eds. 1975. Syntax and semantics 3: speech acts. New York: Academic Press. &ice, H.P. 1975. ‘Logic and conversa’tion’. in: Cole and Morgan, eds. pp. 41-S18. Grimes, J .E. 1975. The thread of di~ourse. The Hague: Mouton. M y, M. 1975. ‘A %amework for :representing knowledge’. In: P. Winston, ed., The logic of computer vision. hew York: McGraw-Hill. pp. 21 l-277. Searle, 1. 1969. Speech acts: an essay in the philosophy of language. Cambridge: Cambridge University Press, Searle, J. 197.5. ‘Indirect speech acts’. In: Cole and Morgan, eds. pp. 59-82. ‘Sinclair, J.McH. and R.M. Coulthard. 1375. Towards an analysis of discourse. Oxford: Oxford University Press. i’an Dijk, T. 1972. Some aspects of text grammars. The Hague: Mouton. !5&~e B&run (b. 1949) was educated at London University (1969-72) (1974-77) where he completed a Ph.D. in 1977. From 1977-78 he was and from 1978 he has held a sin&r position in the School at the Uniwersity of Fast Anglia. Current research interests f linguistic ability, and theories of word meaning.

and Essex University a Lecturer in Linguistof EngIish and Ameriinclude the computer