KelvinJiang/freebase_qa

Dataset Card for FreebaseQA

Dataset Description

• Homepage:
• Repository: FreebaseQA repository
• Paper: FreebaseQA ACL paper
• Leaderboard:
• Point of Contact: Kelvin Jiang

Dataset Summary

FreebaseQA is a dataset for open-domain factoid question answering (QA) tasks over structured knowledge bases, like Freebase.

Supported Tasks and Leaderboards

[More Information Needed]

Languages

English

Dataset Structure

Data Instances

Here is an example from the dataset:

{'Parses': {'Answers': [{'AnswersMid': ['m.01npcx'], 'AnswersName': [['goldeneye']]}, {'AnswersMid': ['m.01npcx'], 'AnswersName': [['goldeneye']]}], 'InferentialChain': ['film.film_character.portrayed_in_films..film.performance.film', 'film.actor.film..film.performance.film'], 'Parse-Id': ['FreebaseQA-train-0.P0', 'FreebaseQA-train-0.P1'], 'PotentialTopicEntityMention': ['007', 'pierce brosnan'], 'TopicEntityMid': ['m.0clpml', 'm.018p4y'], 'TopicEntityName': ['james bond', 'pierce brosnan']}, 'ProcessedQuestion': "what was pierce brosnan's first outing as 007", 'Question-ID': 'FreebaseQA-train-0', 'RawQuestion': "What was Pierce Brosnan's first outing as 007?"}

Data Fields

• Question-ID: a string feature representing ID of each question.
• RawQuestion: a string feature representing the original question collected from data sources.
• ProcessedQuestion: a string feature representing the question processed with some operations such as removal of trailing question mark and decapitalization.
• Parses: a dictionary feature representing the semantic parse(s) for the question containing:
  • Parse-Id: a string feature representing the ID of each semantic parse.
  • PotentialTopicEntityMention: a string feature representing the potential topic entity mention in the question.
  • TopicEntityName: a string feature representing name or alias of the topic entity in the question from Freebase.
  • TopicEntityMid: a string feature representing the Freebase MID of the topic entity in the question.
  • InferentialChain: a string feature representing path from the topic entity node to the answer node in Freebase, labeled as a predicate.
  • Answers: a dictionary feature representing the answer found from this parse containing:
    • AnswersMid: a string feature representing the Freebase MID of the answer.
    • AnswersName: a list of string features representing the answer string from the original question-answer pair.

Data Splits

This data set contains 28,348 unique questions that are divided into three subsets: train (20,358), dev (3,994) and eval (3,996), formatted as JSON files: FreebaseQA-[train|dev|eval].json

Dataset Creation

Curation Rationale

[More Information Needed]

Source Data

Initial Data Collection and Normalization

The data set is generated by matching trivia-type question-answer pairs with subject-predicateobject triples in Freebase. For each collected question-answer pair, we first tag all entities in each question and search for relevant predicates that bridge a tagged entity with the answer in Freebase. Finally, human annotation is used to remove false positives in these matched triples.

Who are the source language producers?

[More Information Needed]

Annotations

Annotation process

[More Information Needed]

Who are the annotators?

[More Information Needed]

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

Kelvin Jiang - Currently at University of Waterloo. Work was done at York University.

Licensing Information

[More Information Needed]

Citation Information

@inproceedings{jiang-etal-2019-freebaseqa,
    title = "{F}reebase{QA}: A New Factoid {QA} Data Set Matching Trivia-Style Question-Answer Pairs with {F}reebase",
    author = "Jiang, Kelvin  and
      Wu, Dekun  and
      Jiang, Hui",
    booktitle = "Proceedings of the 2019 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers)",
    month = jun,
    year = "2019",
    address = "Minneapolis, Minnesota",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/N19-1028",
    doi = "10.18653/v1/N19-1028",
    pages = "318--323",
    abstract = "In this paper, we present a new data set, named FreebaseQA, for open-domain factoid question answering (QA) tasks over structured knowledge bases, like Freebase. The data set is generated by matching trivia-type question-answer pairs with subject-predicate-object triples in Freebase. For each collected question-answer pair, we first tag all entities in each question and search for relevant predicates that bridge a tagged entity with the answer in Freebase. Finally, human annotation is used to remove any false positive in these matched triples. Using this method, we are able to efficiently generate over 54K matches from about 28K unique questions with minimal cost. Our analysis shows that this data set is suitable for model training in factoid QA tasks beyond simpler questions since FreebaseQA provides more linguistically sophisticated questions than other existing data sets.",
}

Contributions

Thanks to @gchhablani and @anaerobeth for adding this dataset.