基于机器学习的青海花儿唱词情感分析

The emotion analysis of Qinghai Hua’er lyrics based on machine learning

本文以青海花儿唱词为研究对象,采用朴素贝叶斯机器学习模型和长短期记忆网络(LSTM)机器学习模型对其建模.首先对青海花儿唱词进行收集,建立模型并对收集到的唱词进行特殊的预处理操作,利用Word2vec生成词向量模型,构建二种不同的机器学习算法模型:朴素贝叶斯模型和LSTM神经网络+Word2vec模型;其次划分花儿唱词训练集和测试集,得到预测的情感倾向;最后通过统一的评价标准对两组实验进行评价,对青海花儿唱词进行了情感分析,并绘制不同地域和不同民族唱词的词云图.实验结果表明:采用长短期记忆网络(LSTM)+Word2vec的机器学习模型进行情感分析与挖掘效果更为理想.

This paper takes the lyrics of Qinghai hua’er as the research object,building model by Na ve Bayesian Model(NBM)and Long and Short Term Memory(LSTM).Firstly,qinghai flower lyrics are collected,a model is established,and special pre-processing operation is carried out on the collected lyrics.Word vector model is generated by Word2vec,and two different machine learning algorithm models are constructed:naive Bayesian mode and LSTM neural network+Word2vec model.Secondly,the training set of Qinghai flower lyrics and the test set of lyrics were divided,and the model training was put in and the test set of lyrics was tested to get the predicted emotional tendency.Finally,the two groups of experiments were evaluated by unified evaluation criteria,the emotion analysis of qinghai huahua lyric was carried out,and the lyric cloud maps of different regions and nationalities were drawn.The experimental results show that the machine learning model of long and Short term memory network(LSTM)+Word2vec is more ideal for emotion analysis and mining.