Search

Multi-word Lexical Units (MWLU) are of great importance in language in general, and in Natural Language Processing in particular, since they are not governed by the free rules of the system. In this article, we give an overview of the different types of phraseological units, explaining briefly each one's features. Our priority being to process idioms automatically in Basque texts, we concisely analyze several approaches for the inflectional description of MWLUs, and then, we explain the system we have developed for Basque: (i) a general representation for describing MWLUs in the lexical database for Basque (EDBL), (ii) HABIL, a tool capable of detecting and analyzing them based on the features described in the database, and (iii) a constraint grammar for disambiguating ambiguous MWLUs.

In this article, we shall comment on the steps that have to be taken to give a linguistic label to a corpus and the difficulties that appear in this process. Our main objective was to highlight the importance of the labelling when preparing a corpus that is useful for linguistic research, and the need to establish criteria and to take the decisions that this entails. We also explain how semi-automatic methods are applied and how the manual revision that guarantees the quality of the corpus is carried out. Once the corpus has been revised and labelled, it will be useful both for carrying out linguistic analyses and for improving or assessing the linguistic tools and resources, and also for channelling automatic study.