Guided Sentences Composition for Disabled People

We present the advantages of guided sentences composition for communicating in natural language with computers. We show how guidance can be achieved by means of the partial synthesis of sentences and describe our approach consisting in separating knowledge that comes under different levels of well-formedness and coroutining their treatment. This approach has led us to develop in Prolog a software, ILLICO, allowing for the conception of natural language interfaces with guided sentences composition. We present an application we have developed by means of ILLICO : KOMBE, a communication aid system for handicapped persons. 1 G u i d e d S e n t e n c e s C o m p o s i t i o n We may distinguish two kinds of communication with natural language interfaces : a "free" mode : the user types sentences without knowing the limits of the interface but he hopes he will be understood. Trivial reality : user's freedom will always be the freedom the system will grant him. a guided mode : the system guides the user while he composes sentences. Unlike the "free" mode, with guided composition users quickly perceive the limits of the interface. The designer does not have to program all the expressions or structures having the same meaning. Unique forms and structures are sufficient. He may forget the others. Guided composition is a powerful principle for natural language interfaces to database and knowledge access, operating and command systems. This principle can also be applied to learning systems (natural languages or formal ones like database or programming languages). Expected tbrmulations can be provided to the user by means of partial synthesis of sentences. This principle was introduced by Alain Colmerauer within the French-English interface, Orbis, which was connected to a database on planets (Colmerauer and Kittredge, 1982). The same system is used for analyzing a given sentence and for synthesizing expected fomulations. Orbis is implemented in Prolog and illustrates in a convincing manner the potential reversibility of certain programs written in that language. We are concerned here with the partial synthesis of sentence from left to right. The main problem of this approach is that a word or an expression that has been synthesized by the system (and selected by the user to compose his sentence) must not lead to a future dead end. Theoretically, dead ends may issue from different levels corresponding to the differents levels of well-formedness of a sentence: lexical, syntactic, conceptual and contextual levels, in particular. As a result of formal language theory, we know that the emptiness problem is solvable for context-free grammars. So one can take all the levels of well-formedness into account simultaneously by merging them into one formalism, a context-free grammar with symbols reflecting lexical, syntactic, conceptual and contextual constraints, as it is done for instance in "semantic grammars". This is the case in Tennant's menu-based system (Tennant, 1984). The approach we have decided to follow is different. We are interested in (1) separating knowledge that comes under different levels (lexical, syntactic, conceptual) into distinct modules and (2) coroutining their treatment in order to avoid dead ends along the guided composition mode.