Automatic Bilingual Text Document Summarization

摘要

We develop a language independent text document summarization technique using Thematic Term approach. Corpus-based thematic terms capturing the main coverage of the whole corpus are automatically determined. Article-based thematic terms capturing the main theme of a document are extracted. Representative sentences are determined by considering both kinds of thematic terms and summaries can then be generated. We can successfully apply our summarization method to both English and Chinese documents. Many existing evaluation methods require either manual summaries or human judgments. We propose a novel evaluation technique based on information retrieval which does not need these requirements.