Datasets:

Modalities:
Text
Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
pandas
License:
blended_skill_talk / README.md
albertvillanova's picture
Convert dataset to Parquet
6536cf2
|
raw
history blame
8.99 kB
metadata
annotations_creators:
  - crowdsourced
language_creators:
  - crowdsourced
language:
  - en
license:
  - unknown
multilinguality:
  - monolingual
size_categories:
  - 1K<n<10K
source_datasets:
  - original
task_categories:
  - conversational
task_ids:
  - dialogue-generation
paperswithcode_id: blended-skill-talk
pretty_name: BlendedSkillTalk
dataset_info:
  features:
    - name: personas
      sequence: string
    - name: additional_context
      dtype: string
    - name: previous_utterance
      sequence: string
    - name: context
      dtype: string
    - name: free_messages
      sequence: string
    - name: guided_messages
      sequence: string
    - name: suggestions
      sequence:
        - name: convai2
          dtype: string
        - name: empathetic_dialogues
          dtype: string
        - name: wizard_of_wikipedia
          dtype: string
    - name: guided_chosen_suggestions
      sequence: string
    - name: label_candidates
      sequence:
        sequence: string
  splits:
    - name: train
      num_bytes: 10830670
      num_examples: 4819
    - name: validation
      num_bytes: 43961447
      num_examples: 1009
    - name: test
      num_bytes: 44449895
      num_examples: 980
  download_size: 10897644
  dataset_size: 99242012
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*

Dataset Card for "blended_skill_talk"

Table of Contents

Dataset Description

Dataset Summary

A dataset of 7k conversations explicitly designed to exhibit multiple conversation modes: displaying personality, having empathy, and demonstrating knowledge.

Supported Tasks and Leaderboards

More Information Needed

Languages

More Information Needed

Dataset Structure

Data Instances

default

  • Size of downloaded dataset files: 38.11 MB
  • Size of the generated dataset: 15.08 MB
  • Total amount of disk used: 53.17 MB

An example of 'train' looks as follows.

{
  'personas': ['my parents don t really speak english , but i speak italian and english.', 'i have three children.'],
  'additional_context': 'Backstreet Boys',
  'previous_utterance': ['Oh, I am a BIG fan of the Backstreet Boys!  Have you ever seen them performing live?', "No,I listen to their music a lot,  mainly the unbreakable which  is the Backstreet Boys' sixth studio album. "],
  'context': 'wizard_of_wikipedia',
  'free_messages': ['you are very knowledgeable, do you prefer nsync or bsb?', "haha kids of this days don't know them, i'm 46 and i still enjoying them, my kids only listen k-pop", "italian?haha that's strange, i only talk english and a little spanish "],
  'guided_messages': ["i don't have a preference, they are both great. All 3 of my kids get annoyed when I listen to them though.", 'Sometimes I sing their songs in Italian, that really annoys them lol.', 'My parents barely speak English, so I was taught both.  By the way, what is k-pop?'],
  'suggestions': {'convai2': ["i don't have a preference , both are pretty . do you have any hobbies ?", "do they the backstreet boys ? that's my favorite group .", 'are your kids interested in music ?'], 'empathetic_dialogues': ['I actually just discovered Imagine Dragons. I love them!', "Hahaha that just goes to show ya, age is just a umber!'", 'That would be hard! Do you now Spanish well?'], 'wizard_of_wikipedia': ['NSYNC Also had Lance Bass and Joey Fatone, sometimes called the Fat One.', 'Yes, there are a few K-Pop songs that I have heard good big in the USA. It is the most popular in South Korea and has Western elements of pop.', 'English, beleive it or not.']},
  'guided_chosen_suggestions': ['convai2', '', ''],
  'label_candidates': []}

Data Fields

The data fields are the same among all splits.

default

  • personas: a list of string features.
  • additional_context: a string feature.
  • previous_utterance: a list of string features.
  • context: a string feature.
  • free_messages: a list of string features.
  • guided_messgaes: a list of string features.
  • suggestions: a dictionary feature containing:
    • convai2: a string feature.
    • empathetic_dialogues: a string feature.
    • wizard_of_wikipedia: a string feature.
  • guided_chosen_suggestions: a list of string features.
  • label_candidates: a list of lists of string features.

Data Splits

name train validation test
default 4819 1009 980

Dataset Creation

Curation Rationale

More Information Needed

Source Data

Initial Data Collection and Normalization

More Information Needed

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

More Information Needed

Licensing Information

More Information Needed

Citation Information

@misc{smith2020evaluating,
    title={Can You Put it All Together: Evaluating Conversational Agents' Ability to Blend Skills},
    author={Eric Michael Smith and Mary Williamson and Kurt Shuster and Jason Weston and Y-Lan Boureau},
    year={2020},
    eprint={2004.08449},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

Contributions

Thanks to @lewtun, @thomwolf, @lhoestq, @patrickvonplaten, @mariamabarham for adding this dataset.