Semantic And Sentiment Analysis of Multilingual Code-Mixed Text

Abstract: In multilingual societies, communication frequently involves the mixing of two or more languages within a single sentence. This phenomenon, known as code-mixing, is commonly observed in everyday conversations, social media interactions, and spoken communication. Languages such as English, Hindi, and Kannada are often combined, creating challenges for traditional Natural Language Processing (NLP) systems that are primarily designed for monolingual data. The objective of this project is to develop a system capable of performing semantic interpretation and sentiment analysis of multilingual code-mixed text. The proposed system will initially process text-based inputs containing mixed language content and generate a normalized English representation of the input sentence. In addition, the system will classify the sentiment of the input as positive, negative, or neutral and provide a confidence score for the predicted sentiment. The project leverages pre-trained multilingual transformer models and language processing frameworks to analyse mixed-language inputs effectively. As an extension, the system may also explore speech input processing, where spoken sentences are converted to text using speech recognition techniques before being analysed. The proposed system demonstrates how modern NLP techniques can be applied to understand multilingual communication and improve the processing of mixed-language textual data.

Keywords: Multilingual Code-Mixing, Sentiment Analysis, Natural Language Processing (NLP), Semantic Interpretation, Multilingual Transformer Models, Text Normalization.