Deogratius-NLP/Local-Translator

Its a project build using Tanzanian Local languages dataset collected by our team, the first thing we thought of using this dataset is creating a language translator specifically for Haya, Sukuma and Nyakyusa 📝 README Template: Local Language Translator 🌟 Project Title: local Language Translator utilizing machine learning models to provide real-time translations for the Haya, Sukuma, and Nyakyusa languages of Tanzania. 🎯 Overview This project addresses the need for digital translation tools for under-resourced local languages in Tanzania. It provides a simple, interactive interface where a user can input a word or phrase in English and receive the corresponding (or closest equivalent) translation in Haya, Sukuma, and/or Nyakyusa. The current demonstrator focuses on translating English words/phrases into Haya and Sukuma. ✨ Features Bilingual Translation: Real-time translation from English to Haya and Sukuma. Multimodal Input: Supports translation from both typed text input and speech input (as demonstrated in the video). Intuitive Interface: A simple and clean user experience focused on speed and accuracy. Locally Sourced Data: Built on a proprietary dataset collected directly from native speakers of Haya, Sukuma, and Nyakyusa. 💻 Demo and Usage See a live demonstration of the translator in action: [Video Demonstration Link] Link: drive.google.com Description: The video shows a user typing/speaking an English word (e.g., "hello") and the application instantly displaying its corresponding translation in Haya and Sukuma. 🛠️ Technology Stack The core components of this project are: [Programming Language, Python] [Key Framework/Library for ML, spaCy] [Specific Model Architecture, Sequence-to-Sequence (Seq2Seq) with Attention] [Interface/Frontend Technology, React] [ Speech Recognition/NLP Tools used Google Speech Recognition] 💾 Dataset Information The performance of this translator is based on a unique and valuable corpus: Languages Covered: Haya, Sukuma, and Nyakyusa (with the current model outputting Haya and Sukuma). Data Source: Textual data collected and curated directly from nativ …