Using Semantic Role Knowledge for Relevance Ranking of Key Phrases in Documents: An Unsupervised Approach

Abstract

The overwhelming growth of scientific and technical documents over the years calls for smart, efficient and automatic methods to facilitate the process of key phrase extraction and ranking. Most ranking models use features like TF-IDF, topic proportions, and phrase positions to score and rank the key phrases. However, integrating semantic roles of key phrases to compute their relevance ranking scores has not yet been investigated. In this paper, we investigate the integration of sentence position and the semantic role of words in a PageRank method to build a key phrase ranking method. We present the evaluation results of our approach on three scientific datasets. We show that semantic role information, when integrated with a PageRank method, can become a new lexical feature. Our approach had an overall improvement on all the data sets over the state-of-art baseline approaches.