Papers by Niyati Bafna
Towards Universal Segmentations: UniSegments 1.0 (2022.lrec-1)
Copied to clipboard
Zdeněk Žabokrtský, Niyati Bafna, Jan Bodnár, Lukáš Kyjánek, Emil Svoboda, Magda Ševčíková, Jonáš Vidra
| Challenge: | Existing data resources for morphological segmentation are limited to 32 languages . a large number of word forms exist, with some sub-parts being "recycled" many times . |
| Approach: | They propose a multilingual data resource for morphological segmentation in 32 languages . they analyze diversity of how individual linguistic phenomena are captured across them . |
| Outcome: | The proposed scheme is based on 17 existing data resources relevant for segmentation in 32 languages. |
Evaluating Large Language Models along Dimensions of Language Variation: A Systematik Invesdigatiom uv Cross-lingual Generalization (2024.emnlp-main)
Copied to clipboard
| Challenge: | Xue et al., 2021) show that large language models suffer from performance degradation on unseen closely-related languages and dialects relative to their high-resource language neighbour (HRLN). |
| Approach: | They propose to model phonological, morphological, and lexical distance as Bayesian noise processes to synthesize artificial languages that are controllably distant from the HRLN. |
| Outcome: | The proposed model offers insights on model robustness to isolated and composed linguistic phenomena and the impact of task and HRL characteristics on PD. |
How Important is ‘Perfect’ English for Machine Translation Prompts? (2026.findings-eacl)
Copied to clipboard
Patrícia Schmidtová, Niyati Bafna, Seth Aycock, Gianluca Vico, Wiktor Kamzela, Kathy Hämmerl, Vilém Zouhar
| Challenge: | Large language models (LLMs) are largely trained on and respond best to English prompts, but are also sensitive to errors in user prompts. |
| Approach: | They propose to model a range of error types exhibited by second language English speakers and quantify their impact on LLM performance. |
| Outcome: | The proposed model is brittle to natural spelling errors but not to errors at the phrasal level, but the variance in quality caused by these errors is lower than the variance over the initial prompt choice. |
DialUp! Modeling the Language Continuum by Adapting Models to Dialects and Dialects to Models (2025.acl-long)
Copied to clipboard
Niyati Bafna, Emily Chang, Nathaniel Romney Robinson, David R. Mortensen, Kenton Murray, David Yarowsky, Hale Sirin
| Challenge: | Recent advances in MT quality and language coverage have shown that language varieties with low baseline performance are more likely to benefit from these approaches. |
| Approach: | They propose a training-time technique for adapting a pretrained model to dialectal data and an inference-time intervention adapting dialectal datasets to the model expertise. |
| Outcome: | The proposed model shows significant performance gains for several dialects from four language families, and modest gains for two other language families. |
When Your Cousin Has the Right Connections: Unsupervised Bilingual Lexicon Induction for Related Data-Imbalanced Languages (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing methods for unsupervised bilingual lexicon induction depend on good quality static or contextual embeddings for both languages. |
| Approach: | They propose a method for unsupervised bilingual lexicon induction between a related LRL and a high-resource language that only requires inference on a masked language model of the HRL. |
| Outcome: | The proposed method performs well on low-resource languages with 5M tokens against Hindi . it is compared with existing methods on (mid-resourced) Marathi and Nepali . |
ChiKhaPo: A Large-Scale Multilingual Benchmark for Evaluating Lexical Comprehension and Generation in Large Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Existing benchmarks for large language models (LLMs) are restricted to high- or mid-resource languages, and evaluate performance on higher-order tasks in reasoning and generation. |
| Approach: | They propose a multilingual benchmarking tool to evaluate lexical comprehension and generation abilities of large language models. |
| Outcome: | The proposed benchmarks cover 2700+ languages and surpasses existing benchmarks in terms of language coverage. |