---
dataset_info:
  features:
  - name: story
    dtype: string
  - name: question
    dtype: string
  - name: answer
    dtype: string
  - name: characters
    sequence: string
  - name: distracting_characters
    sequence: string
  splits:
  - name: train
    num_bytes: 255255
    num_examples: 564
  download_size: 53357
  dataset_size: 255255
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: apache-2.0
language:
- en
pretty_name: Thinking is for Doing
---


# Dataset Description

This dataset is converted from the [ToMi](https://aclanthology.org/2022.emnlp-main.248/) dataset as per the paper [https://arxiv.org/abs/2310.03051](https://arxiv.org/abs/2310.03051). The code used for this conversion can be found here: [https://github.com/sachith-gunasekara/t4d](https://github.com/sachith-gunasekara/t4d).



In the given ToMi dataset, we filter those examples that has a ToM (Theory of Mind) question from the corresponding story. Despite the original paper claiming that the character holding the false belief (which will also be the answer to the generated question from this code) is the one mentioned in the original question in ToMi, we had to adapt the algorithm slightly to account for questions that have second order false beliefs.

NOTE. In the above modification made, an interested researcher may also notice that we could instead use those examples corresponding to first order false beliefs are recorded. However, upon inspection of the ToMi dataset, this criteria does not result in roughly 500 examples as mentioned in the paper. 

We warmly welcome collaborations to inspect this further.


# License

This project is licensed under the Apache License, Version 2.0.