low-resource-languages

This paper studies using Whisper for Persian speech emotion recognition, showing PCA-based dimensionality reduction improves performance and efficiency, while ASR fine-tuning offers only modest gains.

Microsoft Research presents the Paza Speech Playbook, an interactive guide for building robust speech technologies for low-resource and multilingual languages, covering data collection, model training, deployment, and evaluation. The walkthrough video introduces the playbook and demonstrates how to apply proven practices.

This paper proposes a conceptual framework for building an electronic explanatory dictionary of the Tajik language using large language models, integrating morphological analysis, lemmatization, semantic clustering, and dictionary entry generation with PEFT strategies.

This paper describes Team uOttawa's system for the EvaLatin 2026 shared task on Named Entity Recognition for Classical Latin, using prompt engineering of commercial LLMs (Gemini 2.5 Pro and Claude Sonnet 4-5) to achieve first place in both coarse- and fine-grained NER subtasks.

This arXiv paper audits five frontier LLMs on native Bangla derogatory speech, finding that safety alignment fails to generalize to low-resource languages — models comprehend and generate unsafe content at high rates despite high-resource alignment. The authors propose a 'comprehension–containment decoupling' and show that reasoning and persona framing further break down safety filters.

This paper proposes the RP-RCAF prompting strategy to generate culturally sensitive mental health advice in low-resource languages, and introduces the G-REFS evaluation framework, showing significant improvement over conventional prompting across multiple LLMs.

MEUSLI is an open-source multilingual projector family that connects Whisper encoder with multilingual LLMs, enabling end-to-end ASR in 28 European languages and extending to speech translation and topic identification.

This paper presents an engineering study adapting NVIDIA Nemotron 3.5 ASR Streaming 0.6B to Kikuyu, Dholuo, and Kalenjin, achieving 42.97% and 33.98% WER on internal sets for Kikuyu and Dholuo, respectively, through data-centric techniques including corpus auditing, normalization, and streaming evaluation.

This paper investigates the feasibility of using translation-based fine-tuning as a resource-efficient alternative to native-language BERT models for low-resource languages, finding it comparable or superior in 53.3% of cases across six NLP tasks.

This paper proposes a culturally-sensitive responsible NLP framework for detecting health misinformation in low-resource languages like Bangla, evaluating small language models such as Phi-4 for claim extraction.

This paper presents a three-phase pipeline that decouples syntax acquisition from algorithmic reasoning to improve small language models for low-resource programming languages like Julia and Ballerina, achieving significant gains in pass@1 with reduced data and cost.

PluraMath extends the PolyMath dataset to 18 underrepresented languages, providing a human-validated benchmark for evaluating multilingual mathematical reasoning in LLMs. The paper reveals a persistent performance gap between high-resource and low-resource languages across 27 models.

This paper surveys the challenges and evolution of NLP for Indic languages, emphasizing cultural heritage, and proposes a new research direction called 'Culture Sensing' to address representation gaps and ensure culturally meaningful AI outputs.

This paper presents an end-to-end pipeline for adapting LLMs to low-resource programming languages, using Pharo as a case study, achieving superior code completion with small models suitable for real-time IDE support.

This paper evaluates the Mamba state space model for ASR on seven South African languages, finding it matches Conformer accuracy with fewer resources, and explores multilingual training strategies and low-resource settings.

This paper examines the deployment of an LLM-based machine translation system for text-to-911 emergency services, highlighting common misconceptions and providing recommendations for stakeholders to ensure safe and effective use of AI in critical contexts.

This paper presents LV-ROVER, a multi-stream Tesseract ensemble for Maltese OCR, achieving a 70% reduction in character error rate through synthetic data training and post-processing, addressing the challenges of low-resource OCR for Maltese.

Introduces ALEE, a framework that uses Abstract Meaning Representations to generate English minimal pairs with controlled semantic shifts and translates them for evaluating text embeddings across 275+ languages, revealing persistent gaps in cross-lingual semantic representation.

Introduces SEATauBench, the first agent-focused evaluation framework for Southeast Asian languages, adapting τ²-Bench to Mandarin, Vietnamese, Thai, Indonesian, and Filipino, and reveals a significant capability gap when moving from English to localized settings.

Introduces NagaTranslate, a pipeline for translation and voice synthesis for low-resource Nagaland creoles using Whisper, VITS, and LLMs.