Extracting Keyphrases from Research Papers Using Citation Networks

Abstract

Keyphrases for a document concisely describe the document using a small set of phrases. Keyphrases were previously shown to improve several document processing and retrieval tasks. In this work, we study keyphrase extraction from research papers by leveraging citation networks. We propose CiteTextRank for keyphrase extraction from research articles, a graph-based algorithm that incorporates evidence from both a document's content as well as the contexts in which the document is referenced within a citation network. Our model obtains significant improvements over the state-of-the-art models for this task. Specifically, on several datasets of research papers, CiteTextRank improves precision at rank 1 by as much as 9-20% over state-of-the-art baselines.

A Position-Biased PageRank Algorithm for Keyphrase Extraction

“…To evaluate the performance of PositionRank, we carried out experiments on two datasets. Both datasets were made available by Gollapalli and Caragea (2014). 1 These datasets consist of research papers from the ACM Conference on Knowledge Discovery and Data Mining (KDD) and the World Wide Web Conference (WWW).…”

“…Although supervised approaches typically perform better than unsupervised approaches (Kim et al 2012;, the requirement for large human-annotated corpora for each domain of study, has led to significant attention towards the design of unsupervised approaches. Unsupervised keyphrase extraction is formulated as a ranking problem with graph-based ranking techniques being considered state-of-the-art (Mihalcea and Tarau 2004;Wan and Xiao 2008;Liu et al 2010;Gollapalli and Caragea 2014). These techniques construct a word graph from each target document, in which nodes correspond to words and edges correspond to word association patterns.…”

Incorporating Expert Knowledge into Keyphrase Extraction

“…Mann and McCallum showed that given limited annotation time, expert-specified labeled features can be used to improve discriminative models over other semi-supervised approaches that use fullylabeled instances. In addition, given sufficient annotated data, various techniques were studied to automatically estimate feature-label distributions for specific tagging problems (Haghighi and Klein 2006;Druck, Mann, and McCallum 2008;Gollapalli et al 2014).…”

“…The example in Table 1 refers to the title of a research paper published in the World Wide Web conference in the year 2010 and is part of the recently-compiled datasets for keyphrase extraction described further in the Experiments section. We highlight some shortcomings of existing sys- (Wan and Xiao 2008;Gollapalli and Caragea 2014;) employ part-of-speech criteria during phrase filtering. Specifically, these systems only consider phrases comprising of nouns and adjectives with POS tags from the set {NN, NNS, NNP, NNPS, JJ} for scoring.…”

“…Various supervised and unsupervised techniques are available for keyphrase extraction (Hasan and Ng 2014). Most state-of-the-art systems first extract a set of candidate phrases for a given document during keyphrase extraction (Frank et al 1999;Medelyan, Frank, and Witten 2009;Gollapalli and Caragea 2014). These systems employ phrase-filtering on the set of all n-grams of an input document to remove phrases that are unlikely to be humangenerated keyphrases.…”