TL;DR
This paper proposes automated methods to construct high-quality WordNets using large language models (LLMs) to generate missing lemmas to address the synset shortfall in non-English and low-resource languages.
Word Sense Disambiguation has advanced rapidly for English and a handful of well-resourced modern languages, but it continues to assume the existence of a sense inventory and a word-to-sense mapping in the source language (Navigli, 2026). These assumptions break down for most historical and low-resource languages, whose dedicated WordNets are either incomplete or still under construction. We present Inspicio, an open-vocabulary retrieval pipeline that links tokens in context to synsets of the Open English WordNet (McCrae et al., 2020) without requiring any source-language inventory or mapping. For each occurrence, an instruction-tuned LLM produces two English translations of the surrounding sentence, a small set of candidate dictionary-style definitions, and a few candidate English lemmas. These outputs drive a hybrid retrieval step that combines dense definition-synset similarity, sparse lemma matching, and Maximal Marginal Relevance re-ranking. We evaluate the pipeline across a 6x6 grid of LLMs and sentence-embedding models on a new bilingual set of manually annotated Latin and Ancient Greek perception verbs, on a subset of PREMOVE dataset (Farina, 2025), and on a diachronic sample of Italian. The best configuration reaches 96% Recall@50 on the perception-verb test set, with each component contributing measurable gains, and remains competitive in the out-of-domain and cross-lingual settings.
Lexical normalization rewrites the noisy, non-standard words that fill user-generated text (tmrw, u, gr8) into their standard forms. Because labelled data is scarce for most languages, a popular shortcut is to train a single model on many languages at once. We ask a simple question: how many languages should such a model be trained on? Using one fixed-capacity character-level model and twelve languages from a standard benchmark, we vary the number of jointly trained languages from one to twelve and measure per-language accuracy. We find a clear curse of multilinguality: accuracy is highest when a language is trained with only a few others, often just one to four, and then falls steadily and substantially, dropping by about forty percent as the rest are piled on. A control that holds the total amount of training data constant makes the decline arrive sooner and fall further, which points to competition among the languages for one fixed-size model rather than to how much data is available. We also test whether a language's typological distance from the others predicts its ideal number of co-training languages, and find no dependable rule: any apparent relationship rests on a couple of languages and does not hold up. For compact normalization models, less can be more: a few languages beat pooling everything into a single model.
MGAL is the first multilingual, granularity- and position-aware long-context benchmark, constructed from United Nations reports spanning 8K to 128K tokens across the six official UN languages, and finds that LLMs perform well at word-level tasks but struggle with coarser-grained ones.
Word Sense Disambiguation (WSD) remains a challenge in natural language processing, particularly for morphologically rich and less-resourced languages, e.g., Serbian, where complex inflectional systems intensify lexical ambiguity. Existing WSD approaches often struggle because of scarce annotated corpora, inadequate sense inventories, and linguistically complex languages. This paper presents a semi-automated framework for Serbian WSD resource development that employs Large Language Models (LLMs) and a constrained JSON-formatted output. The approach combines LLM-assisted sense assignment with a Serbian WordNet-based custom inventory, iterative inventory expansion, and expert validation, with the aim of supporting the practical construction and refinement of sense-annotated resources in a low-resource setting. In WSD experiments, LLMs significantly outperformed the Serbian-specific TeslaXLM , the multilingual embedding model intfloat/multilingual-e5-large , and the sentence transformer all-MiniLM-L6-v2 . The expanded sense inventory improved the accuracy, particularly for GPT-4.1 , which achieved 92.4% for single-word and 95.2% for multi-word spans. These results position LLMs as potentially practical for scalable sense-annotated data creation in less-resourced settings. The proposed methodology addresses key limitations in existing WSD pipelines for Serbian, and contributes to the broader effort of adapting LLMs for semantic processing in linguistically diverse and data-scarce environments.
The results show that multilingual transfer is the dominant factor in extremely low-resource Bantu translation while eliminating the need for heuristic proxy selection, and all systems fail to preserve tonal diacritics, highlighting an open challenge.
MameLoshnLM, the first open-source 8B-parameter language model built specifically for Yiddish, is presented, providing both a foundation for Yiddish NLP and a practical template for language model development in historically rich but digitally underrepresented languages.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
