Spanish-Esperanto Neural Machine Translation

external-link

Nov 2025

Finished

Master's course project on neural machine translation for the low-resource Spanish→Esperanto language pair. Implements and benchmarks three model families: (1) NLLB-200 (Meta's 200-language model) — baseline inference and fine-tuning, (2) mT5-large — fine-tuning with BLEU and chrF optimization, (3) LLaMA — few-shot prompting and fine-tuning. Uses CCMatrix parallel corpus (~40k aligned sentences). Evaluates translation quality with BLEU, chrF, and human assessment. Includes comprehensive notebooks for each experiment and a companion research paper with full methodology and results analysis.

Technologies
ai

AI

huggingface

HuggingFace

jupyter

Jupyter

latex

LaTeX

matplot

Matplot

numpy

NumPy

pandas

Pandas

plotly

Plotly

python

Python

pytorch

PyTorch