Jean Royauté

Retrieving terms and their variants in a lexicalized unification-based framework

Term extraction is a major concern for information retrieval. Terms are not fixed forms and their variations prevent them from being identified by a match with their initial string or inflection. We show that a local syntactic approach to this problem can give good results for both the quality of identification and parsing time.

Les experiences d'indexation a l'inist

We talk, in this paper, about the operation of indexation at INIST. We present two experiments carried out within the Department of Research and New Products that aim at the automation of indexation process. The first one comes within the scope of scientometric studies on text database. We have developped a software toolbox with which we can build a chain of treatments up to the generation of hyperdocuments. Therefore, indexation from a large corpus of source documents is the first module of that chain. In this part, we use linguistic and statistical methods to produce keywords from a stream of data. Linguitic heuristics are used to extract compound nouns or noun phrases from the text and combinational treatments determine the importance of each term according to the document. Keywords are here the input of an hypertext system. The second one is the development of a workstation for the information specialist integrating a computer-aided indexing system on title and abstract in bibliographical records. This indexing process works on a single bibliographical record and combines both linguistic methods and artificial intelligence (keywords generation). We use the same extraction module based on linguistic and add a knowledge based system to deduce implicit keywords. Finally, we show that the original specifications and purpose of each experiment are different and we start a discussion on the interest of these methods in relation to the kind of indexation wanted and the qualities expected from automatic indexing systems.

A Predicate Database for Assisting the Design of a Lexicon-Grammar of Predicative Nouns

In this paper, we present the development and the use of a database to develop a lexicon-grammar of predicative nouns in English. This data base allows to link a verbal construction to a nominal construction with an equivalent nominalization. This tool enables to assist the prediction of the structure of the correlated predicate noun phrase(s) and its dependent preposional noun phrase(s) which take an argument role: subject and possible complement(s). We also show the interest of designing a data base in order to explore, infer and register other noun phrase constructions which are not immediately correladed with verbal constructions.