基于 Word2Vec 及多分类任务的影评分类Movie review classification based on Word2Vec and multi classification task摘 要由于互联网的进步，人们评论电影的方式越来越简单，所以对于电影的评论就越来越多，想看完这部电影的全部评论来分析这部是好还是坏，这明显不是明智的选择。本论文的主要内容是根据网上的电影评论数据进行情感分类预测，使用基于 Word2Vec 的模型，对全部电影评论进行情感分类，进而分析这部电影是一部值得观看的好电影，还是一部浪费时间的坏电影。同时分析基于词义的词向量模型是不是更适用于情感分类，我将对比基于单词语义的词向量模型和基于词频的词向量在同一个分类器下，基于 Word2Vec 的语义模型能否提高影评情感分类的准确性。最终的结果表明，Word2Vec 模型中 87%的准确率和 88%的召回率都要比传统的词袋模型中的 86%的准确率和 86%的召回率要高一点，所以基于 Word2Vec 的词向量模型用机器学习的分类器进行分类是可以提高影评情感分类的准确性。关键词：基于词频构造的词袋模型；Word2Vec；随机森林算法；逻辑回归算法；情感分类ABSTRACTBecause of the progress of the Internet, people's way of commenting on the movie is more and more simple, so there are more and more comments on the movie. It's obviously not a wise choice to analyze whether the movie is good or bad after watching all the comments on the movie.The main content of this paper is to classify and predict the emotion based on the movie review data on the Internet, use the word2vec based model to classify the emotion of all movie reviews, and then analyze whether this movie is a good movie worth watching or a bad electric movie that wastes time. At the same time, it analyzes whether the word vector model based on word meaning is more suitable for emotional classification. I will compare whether the word vector model based on word semantics and the word vector based on word frequency can improve the accuracy of emotional classification of movie reviews under the same cl...
