Method of Short Text Classification based on TF‐IWF Feature Selection

Qijun Duan

《International Journal of Social Science and Education Research》 4卷4期 (2021/04) Pp. 381-386

https://doi.org/10.6918/IJOSSER.202104_4(4).0055

摘要

the classification of short texts on the Internet is a hot topic in natural language processing.This paper proposes a method of short text classification based on convolutional neural networks (CNNs).Firstly, the short text features are obtained by the skip gram model of word2vec, then the high‐level features are further extracted from CNNs, and finally the classification model is obtained by K‐MAX pooling operation and softmax classifier.In the experiment, compared with machine learning method and DBN method, the results show that this method not only solves the problem of dimension disaster and local optimal solution of text vector, but also effectively improves the two-level classification accuracy of Internet short text, which proves the effectiveness of Internet short text classification based on CNN.

關鍵字

Short‐Text ； TF‐IWF Algorithm ； Feature Selection ； Sentiment Classification

參考文獻

Kim Y. Convolutional neural networks for sentence classification [C]/ / Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. 2014: 1746-1751.

Kalchbrenner N, Grefenstette E, Blunsom P. A Convolutional Neural Network for Modelling Sentences [EB/OL]. https: / /arxiv. org/abs/1404. 2188, 2014-04-08.

Jiang Bin. Research and application of microblog automatic classification method [D]. Harbin: Harbin Institute of technology, 2012.

Zhang Zhi Fei, Miao duo Qian, Gao can. Short text classification method based on LDA topic model [J]. Computer applications, 2013, 33 (6): 1587-1590.

Zhang Aili, Liu Guangli, Liu Changyu. Research on multi class text classification based on SVM [J]. Journal of information, 2004, 23 (9): 6-7.