基于SVM和Word2vec的微博评论情感识别模型

Sentiment recognition model of Weibo comments based on SVM and Word2vec

微博作为高互动性的社媒平台,其中富含大量主观性文本数据。为挖掘评论文本中潜在的信息价值,针对传统方法中存在的语义缺失和过度依赖背景知识等问题,提出一种基于SVM和Word2vec的情感识别模型。通过Word2vec模型中的Skip-gram方法利用当前语境的中心词预测上下文结构,将词语映射为词向量,进而转化成向量矩阵,输入至SVM模型进行训练与分类。实验结果表明,模型的准确率为0.943,召回率为0.941,F1值为0.946,具有良好的泛化性。

As a highly interactive social media platform,Weibo is rich in a large amount of subjective text data.In order to mine the potential information value in the review text,aiming at the problems of semantic loss and excessive dependence on back-ground knowledge in the traditional methods,an emotion recognition model based on SVM and Word2vec is proposed.Through the Skip-gram method in the Word2vec model,the context structure is predicted by using the central word of the current context,and the words are mapped into word vectors,which are transformed into vector matrices and then input into the SVM model for training and classification.The experimental results show that the accuracy of the model is 0.943,the recall rate is 0.941,and the F1-score is 0.946,which contains good generalization.