ANN-powered Urdu text processing with a Streamlit UI. Core features match the Replit AI Builder prompt: ANN-driven classification + mini-translation, voice input, multi-language translation, and text-to-speech.
- ANN core: Urdu text classification (greeting/question/statement/farewell) and mini Urdu→English translation with confidence scores.
- Voice input: upload audio (wav/mp3/flac/ogg/m4a); auto-converts to wav before STT; feeds results to the ANN pipeline.
- Multi-language translation: translate to many targets (default set includes English, Arabic, Chinese, Spanish, French, German, Turkish, Persian).
- TTS: gTTS playback for translated text.
- Batch processing, embeddings visualization, dataset view/upload, and ANN training demo.
pip install -r requirements.txt
streamlit run app.py --server.port 5000app.py– Streamlit app with navigation modes (ANN Core, Voice Input, Translation, Batch Processing, Embeddings, Dataset & Training)ann_module.py– PyTorch ANN (LSTM + attention) for classification and mini-translation utilitiestranslator_module.py– Multi-language translation via deep-translator (Google)speech_utils.py– Speech-to-text, with format conversion to wav for reliabilityaudio_utils.py– Text-to-speech using gTTSdataset_utils.py– Sample dataset, stats, augmentation, training helpersrequirements.txt– Dependencies
- Internet is required for Google STT, Google Translate, and gTTS.
- Tested entrypoint:
streamlit run app.py.main.pyis just a placeholder.