Topique or not topique: Formes topiques intrinsèques et forms topiques extrinsèques

Argumentative Ordering of Utterances for Language Generation in Multi-party Human–Computer Dialogue

“…Nevertheless, these later AT developments did not solve all the problems in previous AT accounts, moreover several new problems were added; the most important criticism concerns the fact that in later AT Anscombre and Ducrot moved towards a totally un-vericonditional account of language, giving up the earlier separation of utterance meanings into asserted and presupposed contents: now, utterances have only argumentative meanings, that are quantified via these topoï (Anscombre 1995). Broadly speaking, a topos is an argumentative rule shared by a certain community (which need not have more members that the dialogue partners, for instance) (Moeschler and Reboul 1994).…”

“…In the following section it will be shown how insights from these later developments of AT (Anscombre 1995) can be formalized in an algorithm that, for any pair of utterances, determines if they can be argumentatively compared and, if that is the case, orders the two utterances with respect to their argumentative forces, for an (explicit or implicit) conclusion. 3 Human-Computer Dialogue Architecture…”

“…These modifications opened the path towards more important revisions of AT, by proposing the topoı¨as tools for licensing the move from an argument to a conclusion (Anscombre 1995).…”

“…Nevertheless, one needs a mechanism for actually detecting the topoï eliciting utterances and, moreover, discerning whether such topoï are extrinsic or intrinsic. Anscombre addresses this need by proposing three linguistic tests (Anscombre 1995), out which we will present two, in a concise manner below. These two tests prove relevant in formalizing topical constraints on the argumentative strength of utterances.…”

“…Each topos is associated to a certain object put forward in an utterance, that is, to a certain word or syntagm; hence, each utterance can be seen as a union of bundles of (''word''-level) topoï: for instance, topoï of 'reading' can induce gradations on 'knowledge' (e.g., 'The more reading, the more knowledge'), 'boredom' ('The more reading, the less boredom'), etc., and in turn, each of these can be linked to different gradations: for instance, 'knowledge' can be linked to 'manipulable' ('The less knowledge, the more manipulable') or to 'awareness' ('The more knowledge, the more awareness'), etc. These bundles of topoï are called topical fields (''champs topiques'') and account for the meaning of utterances (Anscombre 1995); hence, in this later AT utterance meaning is reduced to argumentative meaning only, discarding any vericonditional aspect; this has constituted one of the major criticism of AT, since it implies that nothing meaningful can ever be said using language (Iten 2000). However, we will not insist here on these criticism, since our purpose is to rely on positive aspects of AT, in order to provide a means to order utterances in respect with their argumentative strength.…”