A Topic Matching based CNN for Sentence Classification
Yadi Lao, Sheng Gao · 2019
Convolutional Neural Network (CNNs) are widely used in NLP tasks for their powerful ability to capture n-gram features. An effective way to improve CNNs performance is to incorporate prior knowledge or external resources like topic distribution. This paper considers sentence classification problem as semantic matching problem and proposes a topic matching CNN model, referred to as TM-CNN. Firstly, topic words are extracted directly from training data by a novel Naïve Bayes weighting technique. Then a convolutional neural network is applied on a topic matching matrix whose entries represent the soft similarity interaction between words in sentence and those topic words, which allows model to extract both word level and n-gram level matching signals. Experiments demonstrate the effectiveness of TM-CNN on seven text classification tasks, including sentiment analysis and topic classification.