Datasets:

Languages:
English
ArXiv:
License:

The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider removing the loading script and relying on automated data support (you can use convert_to_parquet from the datasets library). If this is not possible, please open a discussion for direct help.

Dataset Card for MetaLWOz

Dataset Summary

MetaLWOz: A Dataset of Multi-Domain Dialogues for the Fast Adaptation of Conversation Models. We introduce the Meta-Learning Wizard of Oz (MetaLWOz) dialogue dataset for developing fast adaptation methods for conversation models. This data can be used to train task-oriented dialogue models, specifically to develop methods to quickly simulate user responses with a small amount of data. Such fast-adaptation models fall into the research areas of transfer learning and meta learning. The dataset consists of 37,884 crowdsourced dialogues recorded between two human users in a Wizard of Oz setup, in which one was instructed to behave like a bot, and the other a true human user. The users are assigned a task belonging to a particular domain, for example booking a reservation at a particular restaurant, and work together to complete the task. Our dataset spans 47 domains having 227 tasks total. Dialogues are a minimum of 10 turns long.

Supported Tasks and Leaderboards

This dataset supports a range of task.

  • Generative dialogue modeling or dialogue-modeling: This data can be used to train task-oriented dialogue models, specifically to develop methods to quickly simulate user responses with a small amount of data. Such fast -adaptation models fall into the research areas of transfer learning and meta learning. The text of the dialogues can be used to train a sequence model on the utterances. Example of sample input/output is given in section Data Instances

Languages

The text in the dataset is in English (en).

Dataset Structure

Data Instances

A data instance is a full multi-turn dialogue between two crowd-workers, one had the role of being a bot, and the other one was the user. Both were given a domain and a task. Each turn has a single utterance, e.g.:

Domain: Ski
User Task: You want to know if there are good ski hills an
hour’s drive from your current location.
Bot Task: Tell the user that there are no ski hills in their
immediate location.
Bot: Hello how may I help you?
User: Is there any good ski hills an hour’s drive from my
current location?
Bot: I’m sorry to inform you that there are no ski hills in your
immediate location
User: Can you help me find the nearest?
Bot: Absolutely! It looks like you’re about 3 hours away from
Bear Mountain. That seems to be the closest.
User: Hmm.. sounds good
Bot: Alright! I can help you get your lift tickets now!When
will you be going?
User: Awesome! please get me a ticket for 10pax
Bot: You’ve got it. Anything else I can help you with?
User: None. Thanks again!
Bot: No problem!

Example of input/output for this dialog:

Input: dialog history = Hello how may I help you?; Is there
any good ski hills an hour’s drive from my current location?;
I’m sorry to inform you that there are no ski hills in your
immediate location
Output: user response = Can you help me find the nearest?

Data Fields

Each dialogue instance has the following fields:

  • id: a unique ID identifying the dialog.
  • user_id: a unique ID identifying the user.
  • bot_id: a unique ID identifying the bot.
  • domain: a unique ID identifying the domain. Provides a mapping to tasks dataset.
  • task_id: a unique ID identifying the task. Provides a mapping to tasks dataset.
  • turns: the sequence of utterances alternating between bot and user, starting with a prompt from bot.

Each task instance has following fields:

  • task_id: a unique ID identifying the task.
  • domain: a unique ID identifying the domain.
  • bot_prompt: The task specification for bot.
  • bot_role: The domain oriented role of bot.
  • user_prompt: The task specification for user.
  • user_role: The domain oriented role of user.

Data Splits

The dataset is split into a train and test split with the following sizes:

Training MetaLWOz Evaluation MetaLWOz Combined
Total Domains 47 4 51
Total Tasks 226 14 240
Total Dialogs 37884 2319 40203

Below are the various statistics of the dataset:

Statistic Mean Minimum Maximum
Number of tasks per domain 4.8 3 11
Number of dialogs per domain 806.0 288 1990
Number of dialogs per task 167.6 32 285
Number of turns per dialog 11.4 10 46

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

The dataset v1 version is created by team of researchers from Microsoft Research (Montreal, Canada)

Licensing Information

The dataset is released under Microsoft Research Data License Agreement

Citation Information

You can cite the following for the various versions of MetaLWOz:

Version 1.0

@InProceedings{shalyminov2020fast,
author = {Shalyminov, Igor and Sordoni, Alessandro and Atkinson, Adam and Schulz, Hannes},
title = {Fast Domain Adaptation For Goal-Oriented Dialogue Using A Hybrid Generative-Retrieval Transformer},
booktitle = {2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
year = {2020},
month = {April},
url = {https://www.microsoft.com/en-us/research/publication/fast-domain-adaptation-for-goal-oriented-dialogue-using-a
-hybrid-generative-retrieval-transformer/},
}

Contributions

Thanks to @pacman100 for adding this dataset.

Downloads last month
138