Comparitive study of LSTM and Ensemble models for Hindi text sentiment analysis
Vishal Hanuman, Kevin Noel, Shravani Vangur, R Shankar · 2024
Sentiment analysis, a common idea in big data analysis and natural language processing, tries to discern the emotion underlying the text and thoughts of the people. In the Indian Subcontinent, major communication is conducted in a number of languages other than English. These languages are frequently used to communicate opinions, and when such opinions are translated into English, their depth and significance are lost. The suggested study offers a method for analyzing the numerous sentiments found in literature written in different languages in their original tongues. The Hindi language is chosen to assess the functionality and performance of the model under consideration for the sake of simplicity in the demonstration. There are two different models that have been considered to understand the performance of the same. One is the LSTM model and the other was developed by implementing a stacking classifier. The stacking classifier combined different properties of various models into the base classifiers and fine-tuned them which aided in achieving a better accuracy than the previously existing models. The LSTM model showcases an accuracy of $\mathbf{8 9 \%}$ whereas the stacking classifier showcases an accuracy of $\mathbf{9 6 \%}$. This can be useful in different applications including movie reviews, restaurant reviews, product reviews and so on.