Multilingual Text Mining for Global Knowledge Exploration Using Self-Organising Maps

摘要

This paper presents a novel approach to multilingual text mining using self-organising maps. Linguistic knowledge of the multilingual concept-term relationships, required by all multilingual text mining tasks, is made available by constructing a self-organising multilingual term cluster map. With this linguistic knowledge, a self-organising multilingual document cluster map is generated to reveal the conceptual content of an arbitrary set of multilingual documents. To realise global knowledge discovery, a technique is devised for providing a language-friendly interface to enable concept-based exploration, retrieval and categorisation of multilingual documents using the multilingual document cluster map. The proposed approach can overcome the feature incompatibility (vocabulary mismatch) problem that is unique to multilingual text mining. It extends the text mining technology beyond the monolingual dimension into the multilingual domain.