SPECIFIC WORD EMBEDDINGS FOR SENTIMENT ANALYSIS IN TWITTER DATA ANALYSIS

Abstract

Social media sites are one of the platforms where a lot of people interact in the present, expanding world. Twitter is one popular social media platform. Tweets are shared with the general public through Twitter. Currently, significant amounts of continuous data word representation learning algorithms do not take into consider the sentimental relationships between words and instead focus only on the text's syntactic information. Specific Word Embeddings for Sentiment Analysis in Twitter Data Analysis is presented in this analysis. Weighted average word embeddings is a method that uses an adapted version of the delta Term Frequency-Inverse Document Frequency (TFIDF) a method to integrate sentiment data into continuous word representations. Sentiment Analysis experiment model makes use of a classifier described as Tailored Random Forest, which was trained the training sample. An analysis utilizes tokenization sentiment, stemming, and the removal of stop words. In this study, the development of multiplication polarity-based sentiment analysis is the main focus. In comparison to un-weighted embeddings, experiments have shown promising results. Experimental results demonstrate that described classifier gives very high predictive Accuracy, macro average Recall and Precision. Finally, they can enhance the sentiment analysis model's performance.