Acknowledge our Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0) (https://raw.githubusercontent.com/Babelscape/WSL/main/wsl_data_license.txt) to access the repository

Our team may take 2-3 days to process your request

Log in or Sign Up to review the conditions and access this dataset content.


Word Sense Linking Dataset

Conference Paper Hugging Face Collection GitHub

Description

The dataset is designed for the task of Word Sense Linking (WSL), where systems are required to identify and disambiguate spans of text to their most suitable senses from a reference inventory. The annotations are provided as sense keys from WordNet, a large lexical database of English. All terms in the dataset have been fully manually annotated.

The dataset sentences are taken from the ALL split of the Raganato framework for Word Sense Disambiguation evaluation, described in the paper Word Sense Disambiguation: A Unified Evaluation Framework and Empirical Comparison by Alessandro Raganato, Jose Camacho-Collados, and Roberto Navigli.

Dataset Structure

The data fields are the same among all splits.

  • text: a string feature containing the input text.
  • tokens: a list of dict features with the following keys:
    • raw: the raw token (str).
    • lemma: the lemma of the token (str).
    • pos: the part of speech tag (str).
    • offset: a list of int indicating the character offsets in the input text.
  • annotations: a list of dict features with the following keys:
    • raw: the raw token (str).
    • lemma: the lemma of the token (str).
    • pos: the part of speech tag (str).
    • labels: a list of str representing the sense labels as WordNet sense keys.
    • source: the source of the annotation (str).
    • token_off: a list of int indicating the token offsets.
    • offset: a list of int indicating the character offsets in the input text.
  • sentence_id: a string feature representing the unique identifier for the sentence.

Here is an example:

{
    "text": "Skilled ringers use their wrists to advance or retard the next swing.",
    "tokens": [
        {"raw": "Skilled", "lemma": "skilled", "pos": "ADJ", "offset": [0, 7]},
        {"raw": "ringers", "lemma": "ringer", "pos": "NOUN", "offset": [8, 15]},
        {"raw": "use", "lemma": "use", "pos": "VERB", "offset": [16, 19]}
    ],
    "annotations": [
        {"raw": "Skilled", "lemma": "skilled", "pos": "ADJ", "labels": ["skilled%3:00:00::"], "source": "OLD", "token_off": [0], "offset": [0, 7]},
        {"raw": "ringers", "lemma": "ringer", "pos": "NOUN", "labels": ["ringer%1:18:02::"], "source": "OLD", "token_off": [1], "offset": [8, 15]},
        {"raw": "use", "lemma": "use", "pos": "VERB", "labels": ["use%2:34:01::"], "source": "OLD", "token_off": [2], "offset": [16, 19]}
    ],
    "sentence_id": "example.s001"
}

Additional Information

Licensing Information: Contents of this repository are restricted to only non-commercial research purposes under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0). Copyright of the dataset contents belongs to Babelscape.

Citation Information

@inproceedings{bejgu-etal-2024-wsl,
    title     = "Word Sense Linking: Disambiguating Outside the Sandbox",
    author    = "Bejgu, Andrei Stefan and Barba, Edoardo and Procopio, Luigi and Fern{\'a}ndez-Castro, Alberte and Navigli, Roberto",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2024",
    month     = aug,
    year      = "2024",
    address   = "Bangkok, Thailand",
    publisher = "Association for Computational Linguistics",
}
Downloads last month
38

Models trained or fine-tuned on Babelscape/wsl

Collection including Babelscape/wsl