摘要
用户的在线评论可以有效地帮助用户选择在线商品或服务。然而,热销商品的用户评论数量极其庞大,同时,这些评论的质量参差不齐。因此,评估评论质量并挑选出高质量的评论变得尤为迫切。目前网站采取邀请用户人工标注的方式评估评论的质量,需耗费用户大量的时间和精力。为解决这个问题,提出了一个自动化评估评论质量的方法。该方法通过应用基于评论与评论者两类特征的支持向量机(SVM)分类器实现。在国内著名在线购物网站京东的评论数据上测试了提出的方法。实验结果表明评估识别高质量评论的准确率达到了87.5%。通过实验发现,能够表征评论信息量的词语数量和语句数量特征很好地评估了评论质量。而由于来自用户对商品的反馈信息的贫乏,能够表征用户反馈的有用性投票数量和回复数量特征并不能很好地评估评论质量。在同时结合评论和评论者特征的基础上,评估评论质量的表现最佳。
Users＇ online reviews are helpful for users to choose products or service online. However, hot sale products hold a large number of reviews which vary considerably in quality. Thus, it＇ s urgent to assess the quality of reviews and pick out the high-quality ones. It＇ s a great waste of time and effort for users who are invited by sites to assess the quality manually at present. In order to solve this problem, a method for automatically assessing the quality of reviews is proposed. The method would be implemented with SVM classifier which is based on reviews and reviewers respectively. The review data on popular domestic online retailer JD. corn is chosen to be tested. Experimental results show that the accuracy of high-quality reviews assessing has achieved 87.5%. The experiment proves that the quantity feature of words and sentences which can characterize the amount of information could help assess the reviews＇ quality well. However, the performance of usable votes and reply quantity feature didn＇ t help a lot for its lack of feedback from users. It performs the best when combining both review feature and reviewer feature.
作者
刘杰
付晓东
刘骊
刘利军
Liu Jie Fu Xiaodong Liu Li Liu Lijun(Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, Yunnan, Chin)
出处
《计算机应用与软件》
2017年第3期71-75,97,共6页
Computer Applications and Software
基金
国家自然科学基金项目(71161015
61462056
61462051
81560296)
