AI-Driven Multilingual Voice Assistant for Stock Market Forecasting Using LSTM

Abstract

Stock market prediction systems often require financial knowledge and technically complex interfaces, which can create accessibility barriers for novice users and users from regional language backgrounds. This paper presents a multilingual voice-enabled stock forecasting system that integrates Long Short-Term Memory (LSTM) networks with speech recognition, language translation, and text-to-speech technologies. Historical stock price data are stored locally in CSV format, and the closing-price series is processed using Min-Max normalization and a 60-step sliding-window approach. A two-layer LSTM architecture is trained to forecast future closing prices. The system accepts text or voice queries in multiple Indian languages and presents predicted prices, percentage changes, graphical trends, and rule-based investment suggestions. Experimental testing was performed on six Indian stocks: HDFC Bank, TCS, Reliance, Infosys, Hindustan, and ITC. The system demonstrates how deep-learning-based forecasting can be combined with multilingual voice interaction to provide an accessible interface for stock-market analysis.