A MULTILINGUAL SPEECH-BASED SYSTEM FOR SEMANTIC UNDERSTANDING AND ENGLISH TRANSFORMATION

Abstract

This research proposes a multilingual speech-based intelligent system for semantic understanding, lexical enrichment, and English transformation. The system accepts spoken input in multiple languages and converts speech into text using Automatic Speech Recognition (ASR). It supports voice note uploads ranging approximately from 20 MB to 30 MB per session, enabling processing of longer speech inputs. Transformer-based models are used for contextual semantic analysis, while NLTK supports synonym and antonym extraction for vocabulary enhancement. The system also performs intelligent rephrasing, correcting incomplete or grammatically inconsistent statements before transforming them into structured English output.

The backend is developed in Python using SpeechRecognition, Transformers, NLTK, and LanguageTool, and the frontend is implemented using Streamlit with a structured multi-page interface. The system was tested on 1000 multilingual statements (maximum 150 words per input) and supports two simultaneous user logins with stable response performance.