LANGUAGE PRODUCTION, A COMPLEX, BUT FEASIBLE TASK WITH THE RIGHT STRATEGY
Michael ZOCK CNRS & LIF, Université de la Méditerranée (Aix-Marseille II) E-mail: firstname.lastname@example.org
Abstract. The goal of this paper is threefold: (a) provide a possible explanation for language production in real time (computation of syntactic structure); (b) present a method to help the language producerbecome fluent in another language than his/her mother tongue; (c) help speakers or writers to find the word they are looking for, by developing navigational support inspired by findings based on the study of the mental lexicon. Key words: language production, pattern matching, pattern drills, language learning, fluency, lexical access, mental lexicon.
1. LANGUAGE PRODUCTION, ATTEMPTS TO EXPLAINAND SUPPORT IT AT LEAST PARTIALLY 1.1. Introduction Speaking a language is a very difficult task. The message (what to say), its corresponding linguistic expression (how to say it) and sound form (say it, i.e. articulation) have to be determined on the fly, quasy simultaneously.1 There is also a space problem. Messages can become complex, yet our short-term memory is limited (Miller, 1956). Inother words, we cannot afford to plan everything in all its details before starting articulation. Our cognitive resources would simply be exhausted, causing us to forget what we meant to say, before having translated it into language. Obviously, language production is a complex process. Nevertheless, despite this fact and despite individual differences most of us succeed most of the time. The questionis how do we manage? I will try to answer this question partially in the next section, confining my hypothesises to the computation of syntactic structures. In the following two sections I will present my ideas of how to support (a) fluency acquisition when learning a foreign language and (b) lexical access (memory search), i.e. how to help authors to find the word they are looking for. The workpresented in these two sections is work in progress. 1.2. A possible explanation for the computation of syntactic structures in real-time As just shown, speaking is a complex process, requiring a nearly instantaneous solution of many subproblems. Speaking is clearly a skill that needs to be learned. The question is, what characterizes this technique? I believe that proficient speakers use at leastthree strategies to address the above mentioned problems: (a) gradual refinement, i.e. top-down planning: people start to speak without having fully worked out the input. In other words, input is underspecified (Zock, 1996). (b) incremental generation (Levelt, 1989), i.e. semi-parallel processing or time-sharing: while speaking we continue to plan the next conceptual chunk; (c) pattern-matching:rather than operating locally, we operate on larger units than single concepts or words, namely chunks or patterns. Linguists describe language in terms of rules, but people hardly ever learn such descriptions, leave alone apply them, at least not all of them especially at the initial stages of acquiring a new language. What
For details concerning the various components, architectures andinformation-flow see Reiter and Dale, (2000) or de Smedt et al. (1996).
people do learn though are patterns complying with these rules. Of course, people do use rules, but in conjunction with patterns. In other words, we believe that processing takes place at two levels or speeds2: At the global level we use patterns to determine roughly the outline of the sentencestructure (skeleton). At the next step we decide on the specifics, making locally the necessary morphological adjustments. While the former is mainly a choice problem and based on pragmatic considerations, the latter is primarily a linguistic reflex governed by linguistic constraints. Given some choice (conceptual and pragmatic), one can't but comply with the rules of the language. If our hypothesis...
Lire le document complet
Veuillez vous inscrire pour avoir accès au document.