RERM：一种基于评论挖掘的需求获取方法

摘要

Recently, the distributed platform of mobile application has been constant expanding, user reviews are getting increased as well.Eliciting new/changed requirements from user reviews manually is a time and effort-consuming work for requirement engineers.In order to address this issue, this paper proposes an approach called RERM ( software requirement elicitation method based on review mining) .Dif-ferent from existing work, by extracting software features via combining ontology with conditional random field ( CRF) model and analysing sentiment polarity, RERM can organise different requirement types for potential software and provide fine-grained prioritised sequence and lat-eral comparison.Experimental results show that feature extraction and sentiment classification perform well.Comparing with other methods, RERM provides more valuable information and improves the efficiency of requirement elicitation.%近年来，移动应用分布式平台不断扩大，用户评论越来越多，需求工程师需花费大量时间和精力从中提取改进或新增需求。针对这一问题，提出基于评论挖掘的需求获取方法RERM，与已有方法不同的是，通过采用本体和条件随机场模型融合的特征提取方法，结合情感分析技术，可以对潜在软件需求进行分类型汇总，从细粒度上进行优先级排序和横向对比。实验结果表明，特征提取和情感分类算法性能良好，与其他方法比较，RERM提供了更多的有价值信息，提升了需求获取效率。