Multilingual Sentiment Analysis as Product Reputation Insight (original) (raw)

This study explores the application of multilingual sentiment analysis as a means to derive product reputation insights from vast amounts of textual data generated online. It addresses the complexities of human language in sentiment analysis, moving beyond traditional lexicon-based methods to implement deep learning techniques, specifically LSTM models. Key focus areas include preprocessing data, handling informal language, and utilizing embeddings to enhance understanding of user expectations. The research also examines various neural network configurations and hyper-parameters to optimize model performance, ultimately contributing to more accurate sentiment classification as either positive or negative.