CogComp/mc_taco

Dataset Card for MC-TACO

Dataset Summary

MC-TACO (Multiple Choice TemporAl COmmonsense) is a dataset of 13k question-answer pairs that require temporal commonsense comprehension. A system receives a sentence providing context information, a question designed to require temporal commonsense knowledge, and multiple candidate answers. More than one candidate answer can be plausible.

Supported Tasks and Leaderboards

The task is framed as binary classification: givent he context, the question, and the candidate answer, the task is to determine whether the candidate answer is plausible ("yes") or not ("no").

Performance is measured using two metrics:

• Exact Match -- the average number of questions for which all the candidate answers are predicted correctly.
• F1 -- is slightly more relaxed than EM. It measures the overlap between one’s predictions and the ground truth, by computing the geometric mean of Precision and Recall.

Languages

The text in the dataset is in English. The associated BCP-47 code is en.

Dataset Structure

Data Instances

An example looks like this:

{
  "sentence": "However, more recently, it has been suggested that it may date from earlier than Abdalonymus' death.",
  "question": "How often did Abdalonymus die?",
  "answer": "every two years",
  "label": "no",
  "category": "Frequency",
}

Data Fields

All fields are strings:

• sentence: a sentence (or context) on which the question is based
• question: a question querying some temporal commonsense knowledge
• answer: a potential answer to the question (all lowercased)
• label: whether the answer is a correct. "yes" indicates the answer is correct/plaussible, "no" otherwise
• category: the temporal category the question belongs to (among "Event Ordering", "Event Duration", "Frequency", "Stationarity", and "Typical Time")

Data Splits

The development set contains 561 questions and 3,783 candidate answers. The test set contains 1,332 questions and 9,442 candidate answers.

From the original repository:

Note that there is no training data, and we provide the dev set as the only source of supervision. The rationale is that we believe a successful system has to bring in a huge amount of world knowledge and derive commonsense understandings prior to the current task evaluation. We therefore believe that it is not reasonable to expect a system to be trained solely on this data, and we think of the development data as only providing a definition of the task.

Dataset Creation

Curation Rationale

MC-TACO is used as a testbed to study the temporal commonsense understanding on NLP systems.

Source Data

From the original paper:

The context sentences are randomly selected from MultiRC (from each of its 9 domains). For each sentence, we use crowdsourcing on Amazon Mechanical Turk to collect questions and candidate answers (both correct and wrong ones).

Initial Data Collection and Normalization

[More Information Needed]

Who are the source language producers?

[More Information Needed]

Annotations

From the original paper:

To ensure the quality of the results, we limit the annotations to native speakers and use qualification tryouts.

Annotation process

The crowdsourced construction/annotation of the dataset follows 4 steps described in Section 3 of the paper: question generation, question verification, candidate answer expansion and answer labeling.

Who are the annotators?

Paid crowdsourcers.

Personal and Sensitive Information

[More Information Needed]

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

[More Information Needed]

Other Known Limitations

[More Information Needed]

Additional Information

Dataset Curators

[More Information Needed]

Licensing Information

Unknwon

Citation Information

@inproceedings{ZKNR19,
    author = {Ben Zhou, Daniel Khashabi, Qiang Ning and Dan Roth},
    title = {“Going on a vacation” takes longer than “Going for a walk”: A Study of Temporal Commonsense Understanding },
    booktitle = {EMNLP},
    year = {2019},
}

Contributions

Thanks to @VictorSanh for adding this dataset.