AiLab-IMCS-UL/wsd-lv

Latvian WSD

A manually annotated Latvian word sense disambiguation (WSD) dataset based on example sentences from the Latvian WordNet.

Each sentence is linked to a specific word sense. Lemmas with only a single sense are excluded. The dataset contains 1,821 lemmas and 5,459 unique senses.

Data Format

Each example contains:

• id: instance identifier
• entry_id: entry identifier
• entry_heading: entry heading (lemma)
• context: sentence containing the target word
• sense_id: correct sense identifier
• candidates: list of candidate senses, each with:
  • sense_id: candidate sense identifier
  • gloss: sense definition

Splits

The dataset is split by entry to avoid lexical overlap:

Split	Lemmas	Instances
Train	1121	33,231
Validation	200	6,154
Test	500	14,979
Total	1821	54,364

Citation

@inproceedings{paikens-etal-2022-towards,
    title = "Towards {L}atvian {W}ord{N}et",
    author = "Paikens, Peteris  and
      Grasmanis, Mikus  and
      Klints, Agute  and
      Lokmane, Ilze  and
      Pretkalni{\c{n}}a, Lauma  and
      Rituma, Laura  and
      St{\={a}}de, Madara  and
      Strankale, Laine",
    booktitle = "Proceedings of the Thirteenth Language Resources and Evaluation Conference",
    month = jun,
    year = "2022",
    address = "Marseille, France",
    publisher = "European Language Resources Association",
    url = "https://aclanthology.org/2022.lrec-1.300/",
    pages = "2808--2815"
}

@inproceedings{znotins-2026-pretraining,
    title = "Pretraining and Benchmarking Modern Encoders for {L}atvian",
    author = "Znotins, Arturs",
    booktitle = "Proceedings of the Second Workshop on Language Models for Low-Resource Languages ({L}o{R}es{LM} 2026)",
    month = mar,
    year = "2026",
    address = "Rabat, Morocco",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.loreslm-1.40/",
    doi = "10.18653/v1/2026.loreslm-1.40",
    pages = "461--470",
    ISBN = "979-8-89176-377-7"
}