一种面向用户反馈的智能分析与服务设计方法

An Intelligent Analysis and Service Design Method for User Feedback

针对用户评论数据,提出了一种面向用户反馈的智能分析与服务设计方法。该方法选取了IOS平台多个App的用户评论数据,对其进行智能挖掘和分类,分析其中的潜在需求。首先,分析用户需求类别,将划分的10个需求进行具体定义。其次,对用户数据进行爬取、清洗和标注,形成软件分类数据集。通过实验检验TextCNN、BiLSTM_Attention和BERT对用户评论数据智能分类的效果,将分类结果进行优先级排序。最后,将该方法封装成一种可重用的智能服务供使用者远程调用。实验结果表明:TextCNN模型综合效果最好,在单一指标Precision上,BERT模型效果最好;BERT模型利用并行计算优化训练过程,使其可拓展到大规模项目,在数据量大、精确性要求比较高的情况下,推荐BERT模型;反之,在应对数据小、时限紧的情况时,推荐TextCNN模型。

In order to process user comments,an intelligent analysis and service design method was proposed.The user comments from multiple iOS Apps,were mined intelligently.The potential user requirements were identified from these comments.Firstly,user requirements were summarized and classified into ten categories.Secondly,user comments were crawled,cleaned,and labeled to form a classification dataset.Then,TextCNN,BiLSTM_Attention,and BERT were used to processed these data.The classification results were prioritized.Finally,the method was packaged into a reusable intelligent service module for remote calls.Experimental results showed that the TextCNN model had the best overall performance,while the BERT model had the best performance in precision.The BERT model optimized the training process through parallel computing and could be extended to large-scale projects.Therefore,when with large data volumes,and the priority of accuracy over time,the BERT model was recommended.Conversely,the TextCNN model would be recommended when dealing with user needs with small data and short time consumption.