This repository contains the code, data, and models of the paper titled "XL-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages" published in Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021.

💬 Language Identification with Support for More Than 2000 Labels -- EMNLP 2023

This repository contains the code and data of the paper titled "Not Low-Resource Anymore: Aligner Ensembling, Batch Filtering, and New Datasets for Bengali-English Machine Translation" published in Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP 2020), November 16 - November 20, 2020.

[PRL 2025, APSIPA 2022] Syllable Analysis Data Augmentation (SADA), This project introduces a glyph dictionary and grammar-aware augmentation strategy designed to enhance Khmer palm leaf manuscript recognition. By modeling the language's grammatical structure, we support more robust OCR performance in low-resource settings.

NLP pipelines for Tagalog using spaCy

Open-source benchmark datasets and pretrained transformer models in the Filipino language.

Speech synthesis (TTS) in low-resource languages by training from scratch with Fastpitch and fine-tuning with HifiGan

The EveryVoice TTS Toolkit - Text To Speech for your language

Python source code for EMNLP 2020 paper "Reusing a Pretrained Language Model on Languages with Limited Corpora for Unsupervised NMT".

[ACL'24] MC^2: A Multilingual Corpus of Minority Languages in China (Tibetan, Uyghur, Kazakh, and Mongolian)

[ACL'24 Findings] Teaching Large Language Models an Unseen Language on the Fly

Workflow for forced alignment between languages

A pipeline to isolate and transcribe one language in mixed-language speech

Generate synthetic labeled data for extremely low-resource languages using bilingual lexicons.

Official implementation of "CONCRETE: Improving Cross-lingual Fact Checking with Cross-lingual Retrieval" (COLING'22)

🕸 GlotWeb: Web Indexing for Low-Resource Languages -- under construction.

Confidence Estimation for Black Box Automatic Speech Recognition Systems Using Lattice Recurrent Neural Networks https://arxiv.org/abs/1910.11933 or https://ieeexplore.ieee.org/document/9053264

This project explores zero-shot emotional speech synthesis using EMOD, a novel approach combining emotion and content embeddings for multilingual and cross-lingual emotion transfer. Built on a VITS-based TTS model, it preserves speaker identity while enhancing expressiveness, enabling emotion transfer across languages and genders efficiently.

Minangkabau NLP corpus. PACLIC 2020