Dataset Card for DiscEvalMT

Dataset Summary

The DiscEvalMT dataset contains English-to-French translations used for resolving ambiguity in pronoun anaphora resolution and lexical choice (disambiguation and cohesion) context-aware translation. This version of the DiscEvalMT dataset contains further annotations of ambiguous spans and supporting context in the dataset examples to align it with the highlighting scheme of SCAT, enabling granular evaluations of context usage in context-aware NMT models.

Disclaimer: The DiscEvalMT corpus was released in the NAACL 2018 paper "Evaluating Discourse Phenomena in Neural Machine Translation" by Bawden et al. (2018), and an original version of the corpus is hosted on Github with CC-BY-SA 4.0 license.

Supported Tasks and Leaderboards

Machine Translation

Refer to the original paper for additional details on the evaluation of discourse-level phenomena using DiscEvalMT.

Languages

The dataset contains handcrafted English-to-French translation examples containing either anaphoric pronouns or lexical choice items. Examples were created using existing OpenSubtitles 2016 sentences as reference for lexicon and syntactic structure.

Dataset Structure

Data Instances

The dataset contains two configurations (anaphora and lexical-choice), each containing only a test set of 200 examples each. Dataset examples have the following format:

{
  "id": 0,
  "context_en": "The buildings will be finished next week.",
  "en": "Soon they will be full of new residents.",
  "context_fr": "Les bâtiments seront terminés la semaine prochaine.",
  "fr": "Ils seront bientôt pleins de nouveaux résidents.",
  "contrast_fr": "Elles seront bientôt pleines de nouveaux résidents.",
  "context_en_with_tags": "The <hon>buildings<hoff> will be finished next week.",
  "en_with_tags": "Soon <p>they</p> will be full of new residents.",
  "context_fr_with_tags": "Les <hon>bâtiments<hoff> seront terminés la semaine prochaine.",
  "fr_with_tags": "<p>Ils</p> seront bientôt pleins de nouveaux résidents.",
  "contrast_fr_with_tags": "<p>Elles</p> seront bientôt pleines de nouveaux résidents.",
  "type": "m.pl"
}

In every example, the context-dependent word of interest and its translation are surrounded by <p>...</p> tags. These are guaranteed to be found in the en_with_tags, fr_with_tags and contrast_fr_with_tags fields.

Any span surrounded by <hon>...<hoff> tags was identified by human annotators as supporting context necessary to correctly translate the pronoun of interest. These spans are found only in the context_en_with_tags and context_fr_with_tags fields.

In the example above, the translation of the pronoun they (field en) is ambiguous, and the correct translation to the feminine French pronoun Ils (in field fr) is only possible thanks to the supporting masculine noun bâtiments in the field context_fr.

Fields with the _with_tags suffix contain tags around pronouns of interest and supporting context, while their counterparts without the suffix contain the same text without tags, to facilitate direct usage with machine translation models.

Dataset Creation

The dataset was created manually by the original authors, with context usage annotations added by the authors of Quantifying the Plausibility of Context Reliance in Neural Machine Translation for plausibility analysis purposes.

Please refer to the original article Evaluating Discourse Phenomena in Neural Machine Translation for additional information on dataset creation.

Additional Preprocessing

The dataset presents minor adjustments compared to the original DiscEvalMT corpus.

Additional Information

Dataset Curators

The original authors of DiscEvalMT are the curators of the original released dataset. For problems or updates on this 🤗 Datasets version, please contact [email protected].

Licensing Information

The dataset is released under the original CC-BY-SA 4.0 license.

Citation Information

Please cite the authors if you use these corpus in your work.

Original DiscEval-MT

@inproceedings{bawden-etal-2018-evaluating,
    title = "Evaluating Discourse Phenomena in Neural Machine Translation",
    author = "Bawden, Rachel and Sennrich, Rico and Birch, Alexandra and Haddow, Barry",
    booktitle = {{Proceedings of the 2018 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers)}},
    month = jun,
    year = "2018",
    address = "New Orleans, Louisiana",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/N18-1118",
    doi = "10.18653/v1/N18-1118",
    pages = "1304--1313"
}

Annotated version

@inproceedings{sarti-etal-2023-quantifying,
    title = "Quantifying the Plausibility of Context Reliance in Neural Machine Translation",
    author = "Sarti, Gabriele and 
        Chrupa{\l}a, Grzegorz and 
        Nissim, Malvina and
        Bisazza, Arianna",
    booktitle = "The Twelfth International Conference on Learning Representations (ICLR 2024)",
    month = may,
    year = "2024",
    address = "Vienna, Austria",
    publisher = "OpenReview",
    url = "https://openreview.net/forum?id=XTHfNGI3zT"
}