We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: SINC: Service Information Augmented Open-Domain Conversation

Abstract: Generative open-domain dialogue systems can benefit from external knowledge, but the lack of external knowledge resources and the difficulty in finding relevant knowledge limit the development of this technology. To this end, we propose a knowledge-driven dialogue task using dynamic service information. Specifically, we use a large number of service APIs that can provide high coverage and spatiotemporal sensitivity as external knowledge sources. The dialogue system generates queries to request external services along with user information, get the relevant knowledge, and generate responses based on this knowledge. To implement this method, we collect and release the first open domain Chinese service knowledge dialogue dataset DuSinc. At the same time, we construct a baseline model PLATO-SINC, which realizes the automatic utilization of service information for dialogue. Both automatic evaluation and human evaluation show that our proposed new method can significantly improve the effect of open-domain conversation, and the session-level overall score in human evaluation is improved by 59.29% compared with the dialogue pre-training model PLATO-2. The dataset and benchmark model will be open sourced.
Comments: 12pages, 7 figures
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Cite as: arXiv:2206.14000 [cs.CL]
  (or arXiv:2206.14000v1 [cs.CL] for this version)

Submission history

From: Han Zhou [view email]
[v1] Tue, 28 Jun 2022 13:41:48 GMT (1377kb,D)

Link back to: arXiv, form interface, contact.