Extracting Topic Related Keywords by Backtracking CNN Based Text Classifier
Jonghun Cha, Jee-Hyong Lee · 2018
In the last decades, many studies have been done to extract keywords from text and they show remarkable performance. Most of these studies use a rule-based methodology. They usually focus on Part of Speech(POS), collocations, co-occurrences and dependency of words. However, considering the topic of text is very important key for extracting keywords. Thus, in this paper, we proposed a keywords extracting method using Convolutional Neural Network(CNN) based text classifier which has sufficient information about the topic of text. Experimental results show that using topic related information in CNN text classifier model can improve the quality of keywords extraction. Also proposed method is advantageous in that it requires only a deep learning model differently from existing methods.