The MG-ShopDial dataset contains English conversations that mix different conversational goals, including search, recommendation, and question answering in the domain of e-commerce. The dataset includes 64 high-quality dialogues with a total of 2,196 utterances for scenarios of varying complexity. Intents and goals annotations are available on the utterance level. In addition to MG-ShopDial, the data collection tool Coached Conversation Collector is released. This tool supports the proposed coached human-human data collection protocol used for the creation of MG-ShopDial.

Resources available

The GitHub repository is structured as follows:

• CCC/: Source code and documentation of the Coached Conversation Collector tool.
• MGShopDial: MG-ShopDial dataset card and annotation task details.
• MGShopDial/MGShopDial.json: Annotated MG-ShopDial dataset.

Publication

The resources are presented in a SIGIR’23 resource paper. [PDF]

@inproceedings{Bernard:2023:SIGIR,
  author =    {Bernard, Nolwenn and Balog, Krisztian},
  title =     {MG-ShopDial: A Multi-Goal Conversational Dataset for e-Commerce},
  booktitle = {Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval},
  series =    {SIGIR '23},
  year =      {2023}
}

Contact

Should you have any questions, please contact Nolwenn Bernard at nolwenn.m.bernard@uis.no.