Applying Bidirectional Long Short-Term Memory Neural Networks for News Classification

• Main Authors: JIANG, YI-JYUN, 江易麇
• Other Authors: HUANG, CHUEN-MIN
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/sexr7f

碩士 === 國立雲林科技大學 === 資訊管理系 === 106 === The news media has always been one of the important channels for people to access information. Especially in the age of Internet information explosion, the internet is flooded with a large number of news articles. Therefore, how to classify effectively becomes a difficult problem. In recent years, due to the rapid development of computer equipment, deep learning technology has been rapidly emerging, and various studies have shown that deep learning has achieved good results in natural language processing. Therefore, this study uses deep learning to implement news classification and compare deep learning architecture, convolutional neural network (CNN), long short-term memory (LSTM) and bidirectional long-term and short-term memory. bidirectional long short-term memory (Bi-LSTM). Also implemented as support vector machine (SVM) and Naive Bayes (NB) for comparison. This study using Yahoo News, uses Word2Vec train word vector, and then input to CNN, Bi-LSTM, LSTM and SVM for training, and TF-IDF combined with SVM and NB. The experimental results show that the Bi-LSTM Accuracy is 89.3%, followed by LSTM 88%, and Word2vec-SVM 85.32%.