TeenSenti - A novel approach for sentiment analysis of short words and slangs

The present prevalence of online platforms and the internet has seen a substantial increase in the utilization of Generation Z slang, abbreviated expressions, and short words. These Lexical subtleties have become a vital part of a person's daily interactions, and therefore evident in reviews, product comments, and throughout the internet. Resultantly, there exists a need to integrate these Informal expressions such as abbrevi-ations, slang, and short words into natural language processing (NLP) systems. The Informal expressions naturally contain some contextual relevance and therefore can be used to improvise sentiment analysis, as these expressions contribute significantly to the overall context and sentiment of sentences. However, traditional NLP techniques often fall short in recognizing and incorporating these informal expressions, resulting in an atten-uated accuracy of sentiment analysis. To address this issue, this study ventures to provide a comprehensive slang dictionary, encompassing short words, abbreviated expressions, and slang along with the sentiment and sentences of each. This curated slang dictionary is effectively integrated into NLP systems using FastText Embeddings. Through extensive testing across multi-ple machine learning (ML) and deep learning (DL) models, this approach significantly enhances the accuracy of sentiment analysis when compared to conventional methodologies. This research emphasizes the importance of accommodating informal expression in NLP systems thus opening the possibility of future research.