Building socialbots that can have deep, engaging open-domain conversationswith humans is one of the grand challenges of artificial intelligence(AI). To this end, bots need to be able to leverage world knowledgespanning several domains effectively when conversing with humans whohave their own world knowledge. Existing knowledge-grounded conversationdatasets are primarily stylized with explicit roles for conversationpartners. These datasets also do not explore depth or breadth of topicalcoverage with transitions in conversations. We introduce Topical-Chat,a knowledge-grounded human-human conversation dataset where the underlyingknowledge spans 8 broad topics and conversation partners don’thave explicitly defined roles, to help further research in open-domainconversational AI. We also train several state-of-the-art encoder-decoderconversational models on Topical-Chat and perform automated and humanevaluation for benchmarking.
@inproceedings{gopalakrishnan19_interspeech, title = {Topical-Chat: Towards Knowledge-Grounded Open-Domain Conversations}, author = {Karthik Gopalakrishnan and Behnam Hedayatnia and Qinlang Chen and Anna Gottardi and Sanjeev Kwatra and Anu Venkatesh and Raefer Gabriel and Dilek Hakkani-Tür}, year = {2019}, booktitle = {Interspeech 2019}, pages = {1891--1895}, doi = {10.21437/Interspeech.2019-3079}, issn = {2958-1796},}Cite as:Gopalakrishnan, K., Hedayatnia, B., Chen, Q., Gottardi, A., Kwatra, S., Venkatesh, A., Gabriel, R., Hakkani-Tür, D. (2019) Topical-Chat: Towards Knowledge-Grounded Open-Domain Conversations. Proc. Interspeech 2019, 1891-1895, doi: 10.21437/Interspeech.2019-3079