Measuring Term Informativeness in Context

Measuring term informativeness is a fundamental NLP task. Existing methods, mostly based on statistical information in corpora, do not actually measure informativeness of a term with regard to its semantic context. This paper proposes a new lightweight feature-free approach to encode term informativeness in context by leveraging web knowledge. Given a term and its context, we model contextaware term informativeness based on semantic similarity between the context and the term’s most featured context in a knowledge base, Wikipedia. We apply our method to three applications: core term extraction from snippets (text segment), scientific keywords extraction (paper), and back-of-the-book index generation (book). The performance is state-of-theart or close to it for each application, demonstrating its effectiveness and generality.