移动医疗App在线评论维度挖掘与情感分析

Dimension mining and sentiment analysis of online reviews of mobile medical App

目的:对我国移动医疗App的在线评论进行维度挖掘与情感分析,以便精准评价用户满意度。方法:基于App Store中国应用市场中医疗类App的在线评论数据,采用集成多策略的深度学习方法,首先应用TF-IDF算法、BERT模型和Canopy+K-means聚类分析方法提取移动医疗App在线评论的主要维度,然后通过计算各维度权重值,明确不同维度对用户整体评价意见的重要性,最后利用LSTM-CNN模型对各维度的用户评论进行细粒度情感分析。结果:用户关注的移动医疗App在线评论主要维度依次为专业性、可靠性、交互性、易用性和特色性;用户对移动医疗App的整体满意度不高,在可靠性、交互性、特色性维度上的评论积极情感倾向率较低;同时,移动医疗各细分领域的App在不同维度上也存在明显不同的优势和劣势,需要根据自身特色和发展目标进行优化与完善。结论:集成多策略的深度学习方法在移动医疗App在线评论维度挖掘和情感分类上具有很好的适用性、稳定性与可推广性,可为App在线评论文本分析和用户满意度评价提供重要的方法支撑。

Objective Dimension mining and sentiment analysis of online reviews of mobile medical Apps in China were conducted,so as to accurately evaluate user satisfaction.Methods Based on the online review data of medical Apps in the App Store China App Market,the deep learning method of integrated multi-strategy was adopted,and the TF-IDF algorithm,BERT model and Canopy+K-means clustering analysis method were first used to extract the main dimensions of online reviews of mobile medical Apps,and then the importance of different dimensions to the overall evaluation opinions of users is clarified by calculating the weight values of each dimension.Finally,the LSTM-CNN model was used to perform fine-grained sentiment analysis on user comments in each dimension.Results The main dimensions of mobile medical Apps that users pay attention to are:professionalism,reliability,interactivity,ease of use and characteristics.Users’overall satisfaction with mobile medical Apps is not high,and there are few positive comments in terms of reliability,interactivity,and features.There are also obvious different advantages and disadvantages in different dimensions of Apps in various sub-fields of mobile medical care,which need to be optimized and improved according to their own characteristics and development goals.Conclusion The integrated multi-strategy deep learning method has good applicability,stability and generalizability in the dimension mining and sentiment classification of mobile medical App online reviews,and can provide important method support for the text analysis of App online reviews and user satisfaction evaluation.