From Ambiguous Words to Key-Concept Extraction

Abstract

Automatic acquisition of keywords for given document is still an area of active research. In this paper, we consider shift from keyword-based representation to other perspective on representation of document's focus in form of key-concepts. The advantage of using concepts over simple words is that concepts, apart from words, are unambiguous. This leads to better understanding of key-concepts than keywords. We present novel method of key-concept extraction, which provides an efficient way of automatic acquisition of key-concepts in machine processing. We evaluate our approach on classification problem, where we compare it to baseline TF-IDF keyword model and present its competitive results. We discuss its potential of its utilisation on the Web.