Experiment Research on Feature Selection and Learning Method in Keyphrase Extraction

Keyphrases can provide a brief summary of documents. Keyphrase extraction, defined as automatic selection of important phrases within the body of a document, is important in some fields. Generally the keyphrase extraction process is seen as a classification task, where feature selection and learning model are the key problems. In this paper, different shallow features are surveyed and the commonly used learning methods are compared. The experimental results demonstrate that the detailed survey of shallow features plus a simpler method can more enhance the extraction performance.