Abstract

The rapid development of e-commerce gives researchers confidence that customers will be willing to share more and more online data, which in turn, would allow for improved mining algorithms. Many companies also foresee vast profits in mining data from online interaction, behavior, and activity. Opinion mining, also known as sentiment analysis, means automatically detecting and understanding personal expressions about a product or service from customer textual reviews. Recently, aspect-based sentiment analysis has become widely interesting to researchers, particularly with respect to embedded words. Algorithms such as word2vec and GloVe perform well when it comes to capturing analogies and toward lexical semantics in general. However, more complex algorithms are needed to address this issue more precisely, using larger corpora and special kinds of data. This paper introduces a knowledge representation approach that centers on aspect rating and weighting. The study focuses on how to understand the nature of sentimental representation using a multilayer architecture. We present a model that uses a mixture of semantic and syntactic components to capture both semantic and sentimental information. This model shares its probability foundation with the words recognized by word2vec and builds on our prior work concerning opinion-aspect relation analysis. This new algorithm is designed specifically, however, to discover sentiment-enriched embedding rather than word similarities. Experiments were performed using a review dataset from the electronic domain. Results show that the model achieved both appropriate levels of detail and rich representation capabilities.

中文翻译：

---

基于情感分析模型的学习表示中的语义和句法分析

摘要

电子商务的飞速发展使研究人员充满信心，客户将愿意共享越来越多的在线数据，从而可以改进采矿算法。许多公司还预见到了通过在线交互，行为和活动来挖掘数据的巨大利润。意见挖掘（也称为情感分析）是指根据客户的文本评论自动检测和理解有关产品或服务的个人表达。最近，基于方面的情感分析已成为研究人员广泛关注的问题，尤其是在嵌入单词方面。一般而言，诸如word2vec和GloVe之类的算法在捕获类比和词法语义时表现良好。但是，需要更复杂的算法才能更精确地解决此问题，使用更大的语料库和特殊类型的数据。本文介绍了一种知识表示方法，该方法以方面评级和加权为中心。该研究着重于如何使用多层体系结构理解情感表达的本质。我们提出了一个模型，该模型使用语义和句法成分的混合来捕获语义和情感信息。该模型与word2vec识别的单词共享其概率基础，并建立在我们先前关于观点-方面关系分析的工作的基础上。但是，专门设计了这种新算法，以发现情感丰富的嵌入而非单词相似性。使用来自电子领域的评论数据集进行了实验。