Towards Generating Procedural Texts: An Exploration of their Rhetorical and Argumentative Structure

Instructional texts consist of sequences of instructions designed in order to reach an objective. The author or the generator of instructional texts must follow a number of principles to guarantee that the text is of any use. Similarly, a user must follow step by step the instructions in order to reach the results expected. In this paper, we explore facets of instructional texts: general prototypical structures, rhetorical structure and natural argumentation. Our study is based on an extensive corpus study with the aim of generating such texts. 1 General typology of instructional texts Instructional texts, also equivalently called procedural texts, consist of a sequence of instructions, designed with some accuracy in order to reach an objective (e.g. assemble a computer). Procedural texts explain how to realize a certain goal by means of actions which are at least partially temporally organized. Procedural texts often exhibit a quite complex rational and ’irrational’ structure, outlining different ways to realize something, with arguments, advices, conditions, hypothesis, preferences, evaluations, user stimulations, etc. They also often contain a number of recommendations, warnings, and comments of various sorts. Another feature is that procedural texts tend to minimize the distance between language and action. Plans to realize a goal are made as immediate and explicit as necessary, the objective being to reduce the inferences that the user will have to make before acting, and therefore potential errors or misunderstandings. Texts are thus oriented towards action, they combine instructions with icons, images, graphics, summaries, etc. In our perspective, procedural texts range from apparently simple cooking receipes to large maintenance manuals (whose paper versions are measured in tons e.g. for aircraft maintenance). They also include documents as diverse as teaching texts, medical notices, social behavior recommendations, directions for use, assembly notices, do-it-yourself notices, itinerary guides, advice texts, savoir-faire guides etc. Procedural texts obey more or less to a number of structural criteria, which may depend on the author’s writing abilities and on traditions associated with a given domain. There is obviously a big contrast between maintenance manuals and people magazines which both contains such texts. Procedural texts can be regulatory, procedural, programmatory, prescriptive, injunctive, or may introduce advices (for social of psychological behavior) (Adam, 2001). The study we present in this paper has three main goals: • First, to be able to accurately respond in natural language to How? (procedural) questions (Aouladomar 2005), on the Web or in texts, outlining the structure of a procedure that answers such a question, • Next, to be able to select the best text (assuming a single text is selected as a response) w.r.t. the user profile when there are several responses (a frequent phenomenon on the Web), • Finally, and for the long term, to be able to merge procedural texts or fragments of procedural texts to construct an optimal text, in terms of level of detail, illustration, etc. Integrating texts is obviouly a long-term challenge. This paper basically relates the structure of instructional texts as they are in French. English translations of examples are just glosses, they are given when space constraints permit. We believe that besides language realization variants, most of the characteristics we present here are language neutral. This study is based on an extensive corpus study, within a language production perspective. This approach allows us to integrate logical, linguistic (e.g. (Moschler, 1985), (Anscombre et al. 1981)) and philosophical views of argumentation. In this paper we first introduce some elements of a general typology of instructional texts, outlining the number of components, rational and irrational, that compose them. We then give the most important structural elements, and the rhetorical structure that characterizes the relations between elements. Finally, we focus on argumentation, a major dimension of instructional texts and we briefly present the forms of arguments which are the most important, and under what constraints they can be generated.