Collaboratively Training Sentiment Classifiers for Multiple Domains
6 Pages Posted: 1 May 2018
Date Written: April 6, 2018
Abstract
Blogging and networking platforms like Facebook, Reddit, Twitter and LinkedIn are social media channels where users can share their thoughts and opinions. Since online chatter is a vital and inexhaustible source of information, these thoughts and opinions hold the key to the success of any endeavor. Tweets which are posted by millions all over the world can be used to analyze consumers’ opinions about individual products, services and campaigns. These tweets have proven to be a valuable source of information in the recent years, playing key roles in success of brands, businesses and politicians. Sentiment Classifiers for Multiple Domains (SCMD) is proposed in this paper to tackle Sentiment Analysis with a lexicon-based approach for extracting positive, negative, and neutral tweets by using part-of-speech tagging from natural language processing. Global and Domain one are the two components of the specific classifier of each domain. When the model acquires specific general knowledge and is shared by multiple domains, it is the global model. The model acquires the expression of the specific sentiment in each domain and is the specific model of the domain; it also extracts the knowledge of the specific domain sentiment from tagged and untagged samples in each domain and is used to improve the specific learning of the domain. To measure domain resemblances, we explore the two types of domain similarities. One is based on textual content and the other is based on expressions of sentiment.
Keywords: Blogging, Sentiment Classifiers, Multiple Domains
Suggested Citation: Suggested Citation