基于比较句的网络用户评论情感分析

Sentiment Analysis of Web Reviews Based on Comparative Sentence Extraction

【目的】为帮助企业识别竞争产品,挖掘有价值的信息,提出基于比较句的情感分析模型。【方法】利用基于比较句数目的热门指数和差距指数筛选热门对比产品,从比较句中抽取特征并根据情感词典计算竞争产品的特征得分。为验证模型的效果,以手机产品为实验对象,以百度搜索引擎为数据来源展开实验。【结果】实验从28对手机产品中筛选出9对热门对比产品,并可视化分析产品的特征差异。相对于以往基于文本的情感分析方法,该模型能够更好地识别竞争产品以及判断竞争产品的用户情感取向。【局限】特征抽取准确性有待进一步提高,比较句识别率存在一定的提升空间。【结论】实验结果与实际情况相符合,模型具有有效性,该方法对企业有较高的利用价值。

[Objective] In order to select competitive products in the market and mine useful information for both enterprises and customers. [Methods] This paper proposes a model of sentiment analysis based on comparative sentence, which can compute feature scores of comparative products and visualize the comparative relations between these products. In order to verify the effectiveness of this model, an experiment on smart phones is conducted with the help of Baidu search engines. [Results] The experiment selects 9 pairs of competitive smart phone products from 28 pairs, thus the results can help smart phone enterprises identify competitors. And also visualize the comparative relations between these products and provide suggestions for customer purchase. [Limitations] The accuracy of feature extraction is not high. The recognition rate of comparative sentence in this model need improvements. [Conclusions] The result of this experiment is consistent with facts, which proves the effectiveness of this competitiveness analysis toothed presented in this paper and its great value to enterprises.