Processing collocations in a terminological database based on a cross-disciplinary study of scientific texts

Abstract : This paper presents an ongoing research aiming to build a terminological and phraseological database that would serve for the design of a dictionary-type Lool for scientists and translators. This research is carried out at Paris Diderot - Paris 7 University by a group of researchers working on corpus linguistics and LSP within the ESlDIS-ARTES project (Etude des Specifjcites et lnvariants des Dlscours Scientifiques en vue de I' Aide a la Rédaction de TExtes Scientifiques). The project is oriented towards multi-disciplinary discourse analysis of invariants and variations in scientific language with the purpose of designing tools for assistance in scientific writing and translation. The paper discusses the various problems arising from the processing of collocations in a terminological database. We first present the general methodology and principles for encoding collocations in a terminological database. We then discuss the advantages of establishing multiple relations between collocations and show how to encode semantic preference and prosody, and to work towards a more meaning-oriented classification of collocations.