Building a Semantic Lexicon: Structuring and Generating Concepts

Abstract

One of the main challenges for computational lexical semantics is to bridge the gap between on the one hand theoretical research on the organization of the lexicon and on the formal representation of word meaning, and on the other hand the increasing need by natural language processing (NLP) systems of accessing large repositories of lexical knowledge. The latter, in fact, represents one of the most critical bottle-necks in the development of robust and efficient systems endowed with linguistic intelligence, given the time-consuming and difficult effort required by the construction and maintenance of lexical knowledge bases. This problem has recently been tackled by designing and developing (both at the national and at the international level) general lexical resources based on a common model, with the aim to provide an explicit and (possibly) standard representation of the linguistic content of lexical items at various levels of representation. Actually, their design represents an important challenge for the computational linguistics community, both from the theoretical and the applicative point of view. The difficulty of this enterprise is given by the inherently multi-purpose and domain-independent vocation of such lexical knowledge bases, since they are intended not to be restricted to specific terminological domains or application types, in order to ensure the maximum degree of portability and reusability. In fact, while application needs and domain features set natural constraints on the format of the lexicons, as well as on the type of the information they must contain, similar constraints are not available for the design of a lexicon which aims to provide general linguistic knowledge. In such cases it is crucial to provide satisfactory answers to issues like: What information a computational lexicon must contain? How to represent this information? What is the format to give to a lexical entry?