基于极性转移和双向LSTM的文本情感分析

摘要

长短时记忆(long-short term memory,LSTM)是一种特殊的循环神经网络(recurrent neural network,RNN),在文本处理中它可以通过刻意的设计来避免长期依赖的问题,因此被广泛用于语言模型、机器翻译以及语音识别等领域.但是该神经网络在文本处理上只能联系历史信息处理,无法预测到即将出现的下文信息;同时,在对文本处理时,无法处理文本中的情感极性问题.因此,文中将LSTM进行前后推算,加强文本中前后句子的关联性,并引入情感极性模型以解决情感分类中的极性转移问题.%The long-short term memory (LSTM) is a special recurrent neural network (RNN),which can be used to avoid long-term dependence problems in the text processing.Due to that,it is widely used in a range of fields such as language model,machine translation and speech recognition.But the neural network can only base on the historical information in the text processing instead of predicting the upcoming information;at the same time,it can not deal with the problem of emotional polarity.This paper calculated the two unidirectional LSTM to strengthen the relevance of the sentences in the texts and introduced the emotion polarity model which is based on the polar shift in sentiment classification.The experiment shows that the proposed model has better effect in text emotion processing.