Contenu du sommaire : La productivité morphologique en questions et en expérimentations, sous la direction de Georgette Dal

Morphological productivity: definitions and related concepts This is a state of the art article on the question of productivity. It first discusses the various definitions of productivity that have been proposed, and then examines the measures that have been suggested, with special attention to those proposed by Baayen. This survey leads to two results. The first is that the main recent developments do not concern the concept of productivity itself, but rather new measuring procedures which integrate both qualitative and quantitative aspects. The second is a significant increase in the importance of corpus-based studies in constructional morphology. Complex words are observed in situ in naturally occurring environments. Corpus-based studies allow one to distinguish active and inactive areas in a language, and to avoid confusing central and peripheral types of data.

Morphological productivity: psycholinguistic models and experimental data In this article, we first described word recognition theoretical models that included or take position about morphology. Models that have been proposed belong to a continuum going from morphology as part of the mental lexicon to morphology as a post-lexical phenomena. Between these two extreme models, others postulated more refined processes taking into account linguistic and /or statistical parameters. In a second part, experimental data that are theoretically or statistically linked to productivity are exposed; in particular the non-word structures effect, the morphological family size effect, the semantic transparency effect and the affix productivity effect.

The productivity of the French suffix -ade This papers studies the productivity of the French suffix -ade, which appears in a number of recent coinings e.g. names of new dishes such as saumonade and sojade and names of festive events such as montgolfiade and Lilliades (Lilies festival), taking care to distinguish the concepts of analyzability, profitability, and availability. Productivity is used as a tool to characterize the suffix -ade on three different levels, namely: 1) The different -ade homomorphs that can be used to derive new lexemes in French. 2) The different uses of the French suffix -ade, inherited from the latine participial suffix -at a. 3) The respective meanings of the forms -ade and -ее. These are both inherited from the same latin suffix -ata, and both preserve a common semantic feature, but they each provide a specific realization of the notion of accomplishment which comes from the latin past participle.

-ET suffixation and the question of productivity This article first sketches an analysis of the system of -et suffixation in French. The analysis shows that -ET suffixation involves approximatively fifteen patterns of lexeme formation organized around two poles (Referent and Speaker pole). Building on this analysis, the article examines the productivity of the derivational suffixes -et and -ette in the French newspaper Libération in a five years period. Formal tools provided by H. Baayen are used to gauge the degree of productivity of both affixes. Two partial results are worth mentioning: first, -ette's degree of productivity exceeds that of -et; this result supports the idea that these suffixes must not be considered as exponents of one and same morphological process. Second, not all derivational patterns are productive to the same degree, which means that it is meaningless to speak of the global degree of productivity of -et suffixation, except as the sum of the degrees of productivity obtained for each pattern. lexeme formation organized around two poles (Referent and Speaker pole). Building on this analysis, the article examines the productivity of the derivational suffixes -et and -ette in the French newspaper Libération in a five years period. Formal tools provided by H. Baayen are used to gauge the degree of productivity of both affixes. Two partial results are worth mentioning: first, -ette's degree of productivity exceeds that of -et; this result supports the idea that these suffixes must not be considered as exponents of one and same morphological process. Second, not all derivational patterns are productive to the same degree, which means that it is meaningless to speak of the global degree of productivity of -et suffixation, except as the sum of the degrees of productivity obtained for each pattern.

Morphologic productivity, representativity and base complexity: the MoQuête system In this paper we propose to describe a corpus processor that can be used to prepare and create lexical databases for French, and whose exploitation is more particularly oriented towards morphological parsing. This processor, called MoQuête (Morphology & Queries), applies on a lexical database (LDB) whose realization is the results of the following steps: recover online corpora, tag them, lemmatize them, and submit them to derivational morphological parsing. Starting from the experiment described for dutch in (Krott & ah, 1999), we illustrate the MoQuête behavior with the presentation of a série of measures, performed from a 27 millions tokens LDB obtained from newspapers online corpora, and meant to evaluate the link between the quantitative morphologic productivity and the representativity of morphologic rules, in the frame of complex lexical units with a complex base.

Productivity through domains and genres: derived adjectives and medical language The productivity of a morphological process generally varies with the type of corpus on which it is observed. Taking as a case study a set of morphological processes of denominal adjectivation, we examine the variation of their productivity with domain and genre. We illustrate this variation on (medical) domain-specific corpora, studying on the one hand their internal oppositions and on the other hand their contrast with an external, large-audience press corpus (Le Monde newspaper). Within a medical specialty, we also examine productivity variation with text genre, between patient discharge summaries and Web sites. This paper describes the design and construction of study corpora, the method for compiling derived adjectives, and the computation of various statistics on these adjectives, -ique derivation is the most frequent adjectivation process in all corpora. It is more productive, in the Baayen sense, in the medical corpus than in the press corpus, as are -el and -al. Productivity differences are also observed between medical specialties among patient discharge summaries, but much less among Web documents. Finally, a few contrasts can be found, in a given medical specialty, between discharge summaries and Web documents.