Keyword Extraction using Multiple Novel Features

In this paper, we propose a novel approach for keyword extraction. Different from previous keyword extraction methods, which identify keywords based on the document alone, this approach introduces Wikipedia knowledge and document genre to extract keywords from the document. Keyword extraction is accomplished by a classification model utilizing not only traditional word based features but also features based on Wikipedia knowledge and document genre. In our experiment, this novel keyword extraction approach outperforms previous models for keyword extraction in terms of precision-recall metric and breaks through the plateau previously reached in the field.