GETALP/FLUE_VSD

FrenchSemEval

Dataset Summary

This dataset correspond to the FrenchSemEval, in which verb occurences where manually annotated with Wiktionary senses.

Supported Tasks and Leaderboards

Verb Sense Disambiguation for French verbs.

Language

French

Dataset Structure

Data Instances

Each instance of the dataset has the following fields and these following types of field.

{
  "document_id": "d001",
  "sentence_id": "d001.s001",
  "surface_forms": ['Il', 'rend', 'hommage', 'au', 'roi', 'de', 'France', 'et', 'des', 'négociations', 'au', 'traité', 'du', 'Goulet', ',', 'formalisant', 'la', 'paix', 'entre', 'les', 'deux', 'pays', '.'],
  "fine_pos": ['CLS', 'V', 'NC', 'P+D', 'NC', 'P', 'NPP', 'CC', 'DET', 'NC', 'P+D', 'NC', 'P+D', 'NPP', 'PONCT', 'VPR', 'DET', 'NC', 'P', 'DET', 'ADJ', 'NC', 'PONCT'],
  "lemmas": ['il', 'rendre', 'hommage', 'à', 'roi', 'de', 'France', 'et', 'un', 'négociation', 'à', 'traité', 'de', 'Goulet', ',', 'formaliser', 'le', 'paix', 'entre', 'le', 'deux', 'pays', '.'],
  "pos": ['CL', 'V', 'N', 'P+D', 'N', 'P', 'N', 'C', 'D', 'N', 'P+D', 'N', 'P+D', 'N', 'PONCT', 'V', 'D', 'N', 'P', 'D', 'A', 'N', 'PONCT'],
  "instance_surface_forms":['aboutissent'],
  "instance_fine_pos":['V'],
  "instance_lemmas":['aboutir'],
  "instance_pos":['V']
}

Data Fields

Each sentence has the following fields: document_id, sentence_id, surface_forms, fine_pos, lemmas, pos, instance_surface_forms, instance_fine_pos, instance_lemmas, instance_pos.

Data Splits

No splits provided.

Dataset Creation

Source Data

Initial Data Collection and Normalization

To build the FrenchSemEval dataset, the authors focused on annotating moderately frequent and moderately ambiguous verbs by selecting verbs appearing between 50 and 1000 times into the French Wikipedia (2016-12-12 fr dump). For those verbs, the authors extracted 50 occurences with other annotations thanks to the French TreeBank Abeillé and Barrier, 2004 and the Sequoia Treebank Candito and Seddah, 2012.

Annotations

Annotation process

To annotate FrenchSemEval, the annotators used WebAnno an open-source adaptable annotation tool. Sentences have been pre-processed into CoNLL format and then annotated into WebAnno. The annotators where asked to only annotate marked occurences using the sense inventory from Wiktionnary.

Who are the annotators?

The annotation has been performed by 3 French students, with no prior experience in dataset annotation.

Dataset statistics

Type	#
Number of sentences	3121
Number of annoatated verb tokens	3199
Number of annotated verb types	66
Mean number of annotations per verb type	48.47
Mean number of senses per verb type	3.83

Licensing Information

GNU Lesser General Public License

Citation Information

@inproceedings{segonne-etal-2019-using,
    title = "Using {W}iktionary as a resource for {WSD} : the case of {F}rench verbs",
    author = "Segonne, Vincent  and
      Candito, Marie  and
      Crabb{\'e}, Beno{\^\i}t",
    booktitle = "Proceedings of the 13th International Conference on Computational Semantics - Long Papers",
    month = may,
    year = "2019",
    address = "Gothenburg, Sweden",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/W19-0422",
    doi = "10.18653/v1/W19-0422",
    pages = "259--270",
    abstract = "As opposed to word sense induction, word sense disambiguation (WSD) has the advantage of us-ing interpretable senses, but requires annotated data, which are quite rare for most languages except English (Miller et al. 1993; Fellbaum, 1998). In this paper, we investigate which strategy to adopt to achieve WSD for languages lacking data that was annotated specifically for the task, focusing on the particular case of verb disambiguation in French. We first study the usability of Eurosense (Bovi et al. 2017) , a multilingual corpus extracted from Europarl (Kohen, 2005) and automatically annotated with BabelNet (Navigli and Ponzetto, 2010) senses. Such a resource opened up the way to supervised and semi-supervised WSD for resourceless languages like French. While this perspective looked promising, our evaluation on French verbs was inconclusive and showed the annotated senses{'} quality was not sufficient for supervised WSD on French verbs. Instead, we propose to use Wiktionary, a collaboratively edited, multilingual online dictionary, as a resource for WSD. Wiktionary provides both sense inventory and manually sense tagged examples which can be used to train supervised and semi-supervised WSD systems. Yet, because senses{'} distribution differ in lexicographic examples found in Wiktionary with respect to natural text, we then focus on studying the impact on WSD of the training data size and senses{'} distribution. Using state-of-the art semi-supervised systems, we report experiments of Wiktionary-based WSD for French verbs, evaluated on FrenchSemEval (FSE), a new dataset of French verbs manually annotated with wiktionary senses.",
}

Contributions

• vincent.segonne@univ-grenoble-alpes.fr
• marie.candito@linguist.univ-paris-diderot.fr
• benoit.crabbe@linguist.univ-paris-diderot.fr