EnglishToArmenianPhonemeToGraphemeModel
aeghian/EnglishToArmenianPhonemeToGraphemeModel
This repository contains the architecture, training data, and models to create an Engish to Armenian phoneme to grapheme model in pytorch and tensorflow
machine-transliterationphoneme-to-graphemearmenian-languagepytorchtensorflowseq2seqbrowser-extensionlocal-inference
Summary
This repository implements a machine learning model to convert English phonetic spellings of Armenian words into Armenian graphemes (letters), specifically for Western Armenian pronunciation. It's designed for deployment in a browser extension to enable local, privacy-preserving inference. The project includes PyTorch and TensorFlow model implementations, a custom-generated dataset, and details on the data collection challenges and architectural decisions.