Abstract Proper Names (PNs) present a problem for the automatic processing and understanding of naturally occurring text. Due to their poor coverage in existing lexical resources and the continual appearance of new names, they represent a large body of unknown lexical data. Moreover, the complexity of the constructions in which they can appear and their own internal structure make them difficult to process, even if they are initially known. Yet the successful analysis of names is often crucial to the full understanding of a text. This paper proposes a solution to the problem and describes a natural language processing (NLP) system, FUMES, which makes use of the internal structure of names and the descriptive information that regularly accompanies them to produce lexical and knowledge base entries for unknown PNs. We present some preliminary results showing the viability of this approach for the identification of proper names.
No comments:
Post a Comment