AUTO COMPLETING SENTENCES IN URDU USING NATURAL LANGUAGE PROCESSING

Abstract

This project addresses the lack of advanced text generation tools for Urdu, the national language of Pakistan, which remains underdeveloped compared to English due to limited resources. To ease tasks like chatting, emailing, and document typing in Urdu, the project developed a model for Urdu text prediction and generation. Tire process began by gathering data from Urdu newspapers and blogs to build a suitable dataset. Various models were evaluated, including RNN, LSTM, GRU, and N-Gram. The model accuracy results showed that the N-Gram model achieved 76%, the RNN combined with LSTM reached 83%, and the RNN combined with GRU reached 70%. We connected our front end with each model, allowing users to select any model according to their requirements and needs. The final deliverable is a user-friendly webpage where users can input incomplete Urdu sentences, and the system predicts and suggests possible completions, enhancing efficiency in Urdu writing tasks.