Latent keyphrase generation by combining contextually similar primitive words

摘要

As the number of document resources is continuously increasing, automatically extracting keyphrases from a document becomes one of the main issues in recent days. However, most previous work overlook keyphrases which is nonexistent in the document. Although latent keyphrases do not appear in the document, it can be important because it represents the meaningful concepts or keypoints of the document. We have discovered that the portion of latent keyphrases is more than one fourth of the entire keyphrases. Latent keyphrases also take an important role as much as existential keyphrases in documents. In this paper, we propose an approach to find latent keyphrases of a document in a given document set. The main idea of this approach is to generate keyphrase by choosing primitive words and combining them considering their context in documents. Experiment result shows that latent keyphrase can be extracted by our approach.