Semantic and syntactic analysis in learning representation based on a sentiment analysis model

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.