基于人机交互和隐马尔可夫模型时序分析的语言文本信息抽取研究  

Research on linguistic text information extraction based on human-computer interaction and Hidden Markov model timing analysis

作　　者：王燕[1] 赵建华[1] WANG Yan;ZHAO Jianhua(Shangluo University,Shangluo Shanxi 726000,China)

机构地区：[1]商洛学院,陕西商洛726000

出　　处：《自动化与仪器仪表》2025年第2期24-27,共4页Automation & Instrumentation

基　　金：陕西省重点研发计划项目资助《基于知识图谱的深度推荐系统的研究与应用》(2022GY-073)。

摘　　要：为提高人机交互过程中对语音文本情感分析的准确率,提出一种结合二阶隐马尔可夫分词模型与Bi_LSTM分类模型的语音文本分词与情感分析方法。其中,通过二阶隐马尔可夫分词实现语音文本信息的分词抽取后;然后,通过Bi_LSTM网络实现对人机交互系统语音文本情感的分析。结果表明,使用二阶隐马尔可夫模型在测试集上进行切分,所得结果相较于使用一阶隐马尔可夫分词模型,更符合实际词义,且与人工分词结果相近;基于二阶隐马尔可夫模型分词结果进行的情感分析,总体准确率相较于基于一阶隐马尔可夫模型提高了1.26%,有效提高了模型的文本情感分析的性能;最终在社交文本数据集上,结合二阶隐马尔可夫分词模型与Bi_LSTM分类预测模型,总体准确率达到92.67%。由此得出,在人机交互的语音识别中,无论是在积极、消极还是中性的语音文本上,本模型对情感倾向的分类准确率都更高于使用一阶隐马尔可夫模型和其他模型,。由此得出,本语音文本抽取方法可用于人机交互中的信息抽取和情感分析。In order to improve the accuracy of sentiment analysis in speech text during human-computer interaction,a speech text segmentation and sentiment analysis method combining second-order hidden Markov segmentation model and Bi_LSTM classifica-tion model is proposed.Among them,the segmentation and extraction of speech and text information are achieved through second-or-der hidden Markov segmentation;Then,the Bi_LSTM network is used to analyze the speech,text,and emotions of the human-com-puter interaction system.The results indicate that using a second-order hidden Markov model for segmentation on the test set yields results that are more in line with actual word meanings compared to using a first-order hidden Markov segmentation model,and are similar to the results of manual segmentation;The sentiment analysis based on the segmentation results of the second-order hidden Markov model has an overall accuracy improvement of 1.26%compared to the first-order hidden Markov model,effectively improving the performance of the model’s text sentiment analysis;Finally,on the social text dataset,combined with the second-order hidden Markov segmentation model and Bi_LSTM classification prediction model,the overall accuracy reached 92.67%.From this,it can be concluded that in human-computer interaction speech recognition,whether on positive,negative,or neutral speech texts,this model has a higher accuracy in classifying emotional tendencies than using first-order hidden Markov models and other models,.From this,it can be concluded that this speech text extraction method can be used for information extraction and sentiment analysis in human-computer interaction.

关 键 词：人机交互 隐马尔可夫模型 长短时网络 文本信息 情感分析 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]