Paul Holmes-Higgin and Khurshid Ahmad, University of Surrey Assembling and viewing a corpus of texts: self-organisation, logical deduction and spreading activation as metaphors

Lexicographical and terminological work is increasingly dependent on the analysis of texts, particularly texts organised in a corpus and being made available through computer systems. We argue that the developments in corpus linguistics, artificial intelligence, connection sciences, and lexicography and terminology, can be conjoined together to analyse the various facets of a text. In particular, the users of corpora will be allowed to explore the ‘family resemblance’ of such texts with other texts. This will help in the creation of a user-defined corpus of texts that belong to a family, all having their own idiosyncrasies but all sharing something through a common ‘genetic’ pool. Our approach, a strictly computational account of corpus organisation and usage, will help corpus builders and end-users to incorporate as much as is known about the texts in general and whatever is known about the contents in the description of texts for storage and for retrieval. We demonstrate our interdisciplinary approach by describing how texts in a computer-based corpus can be (a) represented by using knowledge representation formalisms, such as frames, (b) automatically classified by using self-organising artificial neural networks, and (c) managed by using a hybrid representation scheme wherein interactive activation and competition networks are used in conjunction with frames and deductive data bases.