Flowchart Grounded Dialog Dataset (FloDial) is a corpus of troubleshooting dialogs between a user and an agent collected using Amazon Mechanical Turk. The dataset is accompanied with two knowledge sources over which the dialogs are grounded: (1) a set of troubleshooting flowcharts and (2) a set of FAQs which contains supplementary information about the domain not present in the flowchart. FloDial consists of 2,738 dialogs grounded on 12 different troubleshooting flowcharts.

The data is distributed under the CDLA-Sharing-1.0 license and can be downloaded from our Github page. Download FloDial Dataset

FloDial Paper (EMNLP'21)

coming soon

Please cite the following paper if you use this dataset in your work

@inproceedings{raghu-etal-2021-flodial,
    title = "End-to-End Learning of Flowchart Grounded Task-Oriented Dialogs",
    author = "Raghu, Dinesh and Agarwal, Shantanu and Joshi, Sachindra and Mausam",
    booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP)",
    month = nov,
    year = "2021",
    publisher = "Association for Computational Linguistics",
}

Ask us questions at our GH issues page or contact Dinesh Raghu, Shantanu Agarwal, Sachindra Joshi, or Mausam

This task evaluates the ability to generate responses by following flowchart and FAQs. The S-Flo split of the dataset is used for this task.

Rank	Model	Success Rate	Perplexity	BLEU
1 Sep 15, 2021	FloNet (Baseline) IIT Delhi & IBM Research	0.318	4.17	19.89

This task evaluates the ability to generalize to flowcharts unseen during train. The U-Flo split of the dataset is used for this task.

Rank	Model	Success Rate	Perplexity	BLEU
1 Sep 15, 2021	FloNet (Baseline) IIT Delhi & IBM Research	0.133	5.67	14.83