An improved keyword extraction method using graph based random walk model

Abstract

Keywords can be considered as condensed versions of documents, which can play important role in some text processing tasks such as text indexing, summarization and categorization. However, there are many digital documents especially on the Internet that do not have a list of assigned keywords. Assigning keywords to these documents manually is a difficult task and requires appropriate knowledge of the topic. Automatic keyword extraction process can solve this problem. In this paper, we introduce a new improved method for keyword extraction using random walk model by considering position of terms within the document and information gain of terms corresponds to the whole set of documents. We also incorporate mutual information (MI) of terms with random walk model to extract keywords from documents. The experiments on standard test collections show that our method outperforms the previously proposed methods.