|
02 June 2023, Volume 38 Issue 3
|
|
|
Abstract
The study of sentiment has attracted a lot of interest from experts in text mining and natural language processing. The accuracy of sentiment analysis is being hampered by the paucity of annotated data sets that may be utilized to train a model across all domains. There have been numerous attempts to address this problem and enhance cross-domain sentiment classification. In this article, we offer the findings from a thorough, systematic assessment of the methodologies and methods used in cross-domain sentiment analysis. The computational examination of opinions, assessments, attitudes, and feelings regarding the entities and their properties is known as sentiment analysis. Finding the sentiment polarity of the documents, sentences, or aspects is a fundamental problem of sentiment analysis. Users typically share their thoughts on the goods or services in blog entries, on shopping websites, or on review websites. Such opinion-related contents are abundant and expanding exponentially, making it laborious for the maker to manually classify these items. People also anticipate hearing opinions regarding the entities involved in the elements level procedure. When the destination domain lacks labelled data, the challenge of cross-domain sentiment classification entails adapting a classification model trained on the source domain to the target domain. Because of the issue with feature mismatch, applying a sentiment classifier directly trained in the source domain to the target domain frequently leads to subpar performance. This paper compares various cross-domain sentiment analysis methodologies and obstacles while conducting a quick literature review on the subject.
Keyword
Case-Based Reasoning (CBR), Cross Domain Aspect Based Sentiment Analysis (CDABSA), case-based reasoning (CBR), Dynamic Joint Sentiment-Topic Model (DJST), Joint Sentiment-Topic (JST), Latent Dirichlet Allocation (LDA), Moving-window Attentive Gated Recurrent Units (MAGRU), Natural Language Processing (NLP) Query-by-Committee (QBC), Structured Correspondence Learning (SCL), Spectral Feature Alignment (SFA).
PDF Download (click here)
|