【摘要】 针对传统情感分类模型的分类效果不足,无法准确地捕捉词语之间关系的问题,提出一种基于预训练语言模型词向量融合的GE-BiLSTM(Glove-ELMO-BiLSTM)情感分析模型。通过预训练语言模型ELMO以语言模型为目的训练词向量,再与传统的Glove模型的训练结果进行运算融合,结合了全局信息以及局部上下文信息,增加了词向量矩阵的稠密度,词语之间的特征得到更好的表达,结合BiLSTM神经网络可以更好地捕捉上下文信息的关系。实验结果证明:GE-BiLSTM情感分析模型可以达到更好的分类效果,准确率比传统模型提高了2.3百分点,F1值提升了0.024。更多还原

【Abstract】 In order to solve the problem that the traditional sentiment classification model is not effective enough to capture the relationship between words accurately, this paper proposes a GE-BiLSTM(Glove-ELMO-BiLSTM) sentiment analysis model based on pre-trained language model word vector fusion. Through the pre-trained language model ELMO, the word vector was trained for the purpose of the language model. Then, it combined the global information and local context information with the traditional Glove model training results to increase the density of the word vector matrix, so that the characteristics between words could be better expressed. Combining with BiLSTM could better capture the relationship of context information. The experimental results show that the GE-BiLSTM sentiment analysis model can achieve better classification results. The accuracy rate is 2.3 percentage points higher than the traditional model, and the F1 value is increased by 0.17.更多还原