PORSC:融合用户个性化特征的在线评论情感分类模型

针对传统在线评论情感分类忽视了用户个性化的问题,提出了一种融合用户个性化特征的在线评论情感分类(PORSC)方法,该方法为每一类型用户构建一个在线评论情感分类器.PORSC模型由2部分构成:一部分是具有学习评论中常见情感信息的全局情感分类模型;另一部分是能捕捉每种类型用户的个性化特征的特定用户类型分类模型.为解决PORSC模型在训练中的数据稀疏问题,引入多任务学习方法,以协同方式训练分类器,以并行方式解决了PORSC模型中参数的优化问题.通过在2个实际中文产品评论数据集和一个公开的英文评论数据集上实验,并与已有基线方法进行比较与综合分析,结果表明PORSC模型在一定程度上提高了在线评论情感分类的精度.

PORSC: A Sentiment Classification Model Integrating User Personality for Online Reviews

Focusing on the issue that traditional sentiment classification models of online reviews usually omit the user personality,a model called PORSC was constructed for sentiment classification.The PORSC model contains two components,a global one and a user-specific one.The global classifier was used to learn the common sentiment knowledge shared by all users in online reviews.The user-specific classifier was applied to capture the user personality.To address the data sparseness problem in training for the PORSC model,the personalized sentiment classifiers of different kinds of users were trained in a collaborative way based on multi-task learning so that the parameters of the PORSC model can be optimized in parallel.The experimental results on two datasets from the real-life product online reviews and public English books reviews indicate that the proposed PORSC model can improve the accuracy of sentiment classification for online reviews effectively and efficiently.