Automatic Keyphrase Extraction Using SVM

Abstract

The Internet has a plethora of text articles, and it has become a necessity to extract only the relevant information from all the sources. Automatic keyphrase extraction is an essential part of the process of information extraction as it is impossible to manually identify all the keyphrases in textual sources. Keyphrase extraction has thus become an indispensable component of contemporary world of Internet. Researchers have treated keyword extraction as a classification problem where the input candidate words are classified as keywords or non-keywords. The paper tries to address two major issues in keyphrase extraction process, namely candidate selection and extraction of relevant features. Noun phrases extracted using specified regular expressions are considered as candidate words. A supervised machine learning method based on statistical and linguistic features is proposed for keyword extraction using SVM. The experimental results compared with well-known methods, namely SingleRank, ExpandRank, baseline TF-IDF, and the latest work show considerable improvement over the previously achieved results.