Repository for all open-sourced task oriented dialogue (TOD) research at Google.
This is not an officially supported Google product. All resources in this repository are provided "AS IS" without any warranty, express or implied. Google disclaims all liability for any damages, direct or indirect, resulting from their use.
04/12/2022 - Open source the D3ST and SDT papers.
04/04/2022 - Generalize this repository for all task-oriented dialogue research being purused at Google. Originally, this repository held projects related to the SGD dataset.
generation: Template Guided Text Generation (Kale et. al, 2020)state_tracking:d3st: Description-Driven Dialogue State Tracking (Zhao et. al, 2022)sdt: Show Don't Tell (Gupta et. al, 2022)
- Schema-Guided Dialogue Dataset (Rastogi et. al, 2020)
- The source code for the baseline model released in SGD paper can be found here.
- The Schema-Guided State Tracking track in the 8th DSTC focussed on improving the baseline model for LU and DST. The participants were able to significantly improve the performance of the baseline model while reducing the gap in performance between seen and unseen APIs. More details about the competition and the submissions may be found in the overview paper.