Keyword Extraction with Character-Level Convolutional Neural Tensor Networks

Keyword extraction is a critical technique in natural language processing. For this essential task we present a simple yet efficient architecture involving character-level convolutional neural tensor networks. The proposed architecture learns the relations between a document and each word within the document and treats keyword extraction as a supervised binary classification problem. In contrast to traditional supervised approaches, our model learns the distributional vector representations for both documents and words, which directly embeds semantic information and background knowledge without the need for handcrafted features. Most importantly, we model semantics down to the character level to capture morphological information about words, which although ignored in related literature effectively mitigates the unknown word problem in supervised learning approaches for keyword extraction. In the experiments, we compare the proposed model with several state-of-the-art supervised and unsupervised approaches for keyword extraction. Experiments conducted on two datasets attest the effectiveness of the proposed deep learning framework in significantly outperforming several baseline methods.