众包模式下基于用户评论的移动应用按需演化方法研究

For the crowdsourced evolution of mobile applications (App) production line, it is crucial to conduct accurate elicitation and effective use of valuable information hidden in user feedback. The objective of our proposal is to explore the essential factors for on-demand evolution of App’s feature model, through mining and evaluating the significance of extracted evolutionary requirements of Apps. However, recent researches cannot meet this needs. The key scientific issue of our proposal is, how to mine and use the implicit requirements in diverse user reviews to facilitate decision-making for on-demand evolution of Apps. More specifically, this proposal intends to cover the following three sub research topics. Firstly, a multi-granularity approach is proposed to extract Apps’ evolution requirements, which then can be identified comprehensively and represented in a unified manner. Secondly, a multi-dimension method is provided for analyzing evolution requirements of Apps. Then, topic modeling and sentiment computing are combined to cluster App’s reviews, and a taxonomy for App’s reviews can also be derived from user’s intention and sentiment. In this way, it reveals that the evolutionary requirements of the App can be linked to its corresponding feature model. Thirdly, the cross-impact analysis of evolution requirements is performed to establish the corresponding decision-making model, so that it is helpful for Apps developers to select and use appropriate reviews for Apps maintenance and improvement. Finally, a prototype for Apps’ reviews analyzing and recommendation will be developed to facilitate on-demand evolution of App’s feature models. In a word, our proposal will benefit on-demand evolution of Apps based on user reviews. It also can be used to detect and analyze the evolution trends of Apps.

众包模式下移动应用(App)产品线演化的关键是用户评论中价值信息的准确识别和有效使用，需以探寻App特征模型的演化决策要素为目标进行App演化需求的价值挖掘与评估，现有方法难以提供有效技术支持。围绕“如何挖掘和使用用户评论中的隐性需求以实现App产品线的按需和可持续演化”这一关键科学问题，拟开展如下研究：①研究多粒度的App演化需求挖掘方法，实现App演化需求的全面识别和统一表征；②探索多维度的App演化需求分析方法，指导主题建模和情感计算相结合的用户评论聚类，刻画和度量演化需求之间以及App演化需求与特征模型之间的关联；③建立基于交叉影响分析的App需求演化决策模型，从多个视角辅助App特征模型的演化决策。在此基础上开发App用户评论分析与推荐原型工具，为App特征模型的按需演化提供技术和手段支持。研究成果能为基于用户评论的App按需演化提供决策支持，可应用于App演化趋势的预测和分析。

本项目围绕如何挖掘和分析移动应用（App）用户评论中的需求以支持App的按需演化，从App演化需求的挖掘、分析和决策三个方面展开研究工作。（1）对“使用用户反馈服务需求工程”的国内外研究现状和最新研究进展进行了系统调研，其结论表明当前相关研究主要使用App用户评论作为研究数据以支持需求获取和分析活动。（2）探索了基于App版本更新记录的App演化需求挖掘方法，从App用户评论中识别不同类型的需求。（3）提出基于App版本更新记录语义增强的用户评论自动分类方法，采用监督式机器学期算法对App用户评论进行自动分类；同时，开发了基于Web的App用户评论标注与分析工具，以提高人工标注数据的效率和质量。（4）通过主题建模和情感计算相结合的用户评论聚类，抽取演化需求中的演化热点并对演化热点的热度进行度量，对App用户评论中符合App演化趋势的演化需求进行重要性排序，从多个视角辅助App特征模型的按需演化决策。（5）基于众包数据对安全需求与需求工程师技能进行了分析，探索了IT行业对需求工程师及其软技能的要求。.在本项目的支持下，项目组在软件工程、需求工程和计算机相关领域国内外知名期刊和会议发表（含录用）论文10篇（包括CCF B类期刊论文1篇，SCI期刊论文1篇，CCF A类中文期刊论文1篇，CCF B类和C类会议论文4篇，其他会议论文3篇），申请国家发明专利1项，开发App用户评论标注与分析工具1个，完成了项目的预期研究目标。