Hugging Face's logo

Datasets:
pchristm
/
conv_questions

Tasks:
Question Answering
Text Generation
Fill-Mask
Sub-tasks:
open-domain-qa
dialogue-modeling
Languages:
English
Size:
10K<n<100K
ArXiv:
License:
Dataset card Files Community
3
conv_questions / README.md
albertvillanova's picture
albertvillanova HF staff
Add leaderboard to dataset card (#2)
7bc521a
preview code
|
raw
history blame
7.24 kB
metadata
annotations_creators:
  - crowdsourced
language_creators:
  - crowdsourced
language:
  - en
language_bcp47:
  - en-US
license:
  - cc-by-4.0
multilinguality:
  - monolingual
size_categories:
  - 10K<n<100K
source_datasets:
  - original
task_categories:
  - question-answering
  - text-generation
  - fill-mask
task_ids:
  - open-domain-qa
  - dialogue-modeling
pretty_name: ConvQuestions
dataset_info:
  features:
    - name: domain
      dtype: string
    - name: seed_entity
      dtype: string
    - name: seed_entity_text
      dtype: string
    - name: questions
      sequence: string
    - name: answers
      sequence:
        sequence: string
    - name: answer_texts
      sequence: string
  splits:
    - name: train
      num_bytes: 3589880
      num_examples: 6720
    - name: validation
      num_bytes: 1241778
      num_examples: 2240
    - name: test
      num_bytes: 1175656
      num_examples: 2240
  download_size: 3276017
  dataset_size: 6007314

Dataset Card for ConvQuestions

Table of Contents

Dataset Description

Dataset Summary

ConvQuestions is the first realistic benchmark for conversational question answering over knowledge graphs. It contains 11,200 conversations which can be evaluated over Wikidata. They are compiled from the inputs of 70 Master crowdworkers on Amazon Mechanical Turk, with conversations from five domains: Books, Movies, Soccer, Music, and TV Series. The questions feature a variety of complex question phenomena like comparisons, aggregations, compositionality, and temporal reasoning. Answers are grounded in Wikidata entities to enable fair comparison across diverse methods. The data gathering setup was kept as natural as possible, with the annotators selecting entities of their choice from each of the five domains, and formulating the entire conversation in one session. All questions in a conversation are from the same Turker, who also provided gold answers to the questions. For suitability to knowledge graphs, questions were constrained to be objective or factoid in nature, but no other restrictive guidelines were set. A notable property of ConvQuestions is that several questions are not answerable by Wikidata alone (as of September 2019), but the required facts can, for example, be found in the open Web or in Wikipedia. For details, please refer to the CIKM 2019 full paper (https://dl.acm.org/citation.cfm?id=3358016).

Supported Tasks and Leaderboards

[Needs More Information]

Languages

en

Dataset Structure

Data Instances

An example of 'train' looks as follows.

{
  'domain': 'music',
  'seed_entity': 'https://www.wikidata.org/wiki/Q223495',
  'seed_entity_text': 'The Carpenters', 
  'questions': [
    'When did The Carpenters sign with A&M Records?',
    'What song was their first hit?',
    'When did Karen die?',
    'Karen had what eating problem?',
    'and how did she die?'
  ],
  'answers': [
    [
      '1969'
    ],
    [
      'https://www.wikidata.org/wiki/Q928282'
    ],
    [
      '1983'
    ],
    [
      'https://www.wikidata.org/wiki/Q131749'
    ],
    [
      'https://www.wikidata.org/wiki/Q181754'
    ]
  ],
  'answer_texts': [
    '1969',
    '(They Long to Be) Close to You',
    '1983',
    'anorexia nervosa',
    'heart failure'
  ]
}

Data Fields

  • domain: a string feature. Any of: ['books', 'movies', 'music', 'soccer', 'tv_series']
  • seed_entity: a string feature. Wikidata ID of the topic entity.
  • seed_entity_text: a string feature. Surface form of the topic entity.
  • questions: a list of string features. List of questions (initial question and follow-up questions).
  • answers: a list of lists of string features. List of answers, given as Wikidata IDs or literals (e.g. timestamps or names).
  • answer_texts: a list of string features. List of surface forms of the answers.

Data Splits

train validation tests
6720 2240 2240

Dataset Creation

Curation Rationale

[Needs More Information]

Source Data

Initial Data Collection and Normalization

[Needs More Information]

Who are the source language producers?

[Needs More Information]

Annotations

Annotation process

With insights from a meticulous in-house pilot study with ten students over two weeks, the authors posed the conversation generation task on Amazon Mechanical Turk (AMT) in the most natural setup: Each crowdworker was asked to build a conversation by asking five sequential questions starting from any seed entity of his/her choice, as this is an intuitive mental model that humans may have when satisfying their real information needs via their search assistants.

Who are the annotators?

Local students (Saarland Informatics Campus) and AMT Master Workers.

Personal and Sensitive Information

[Needs More Information]

Considerations for Using the Data

Social Impact of Dataset

[Needs More Information]

Discussion of Biases

[Needs More Information]

Other Known Limitations

[Needs More Information]

Additional Information

Dataset Curators

[Needs More Information]

Licensing Information

The ConvQuestions benchmark is licensed under a Creative Commons Attribution 4.0 International License.

Citation Information 

@InProceedings{christmann2019look,
  title={Look before you hop: Conversational question answering over knowledge graphs using judicious context expansion},
  author={Christmann, Philipp and Saha Roy, Rishiraj and Abujabal, Abdalghani and Singh, Jyotsna and Weikum, Gerhard},
  booktitle={Proceedings of the 28th ACM International Conference on Information and Knowledge Management},
  pages={729--738},
  year={2019}
}

Contributions

Thanks to @PhilippChr for adding this dataset.