TY - JOUR
AU - 
AB - Multi-domain Neural Network Language Generation for Spoken Dialogue Systems Tsung-Hsien Wen, Milica Gasi ˇ c, ´ Nikola Mrksi ˇ c, ´ Lina M. Rojas-Barahona, Pei-Hao Su, David Vandyke, Steve Young Cambridge University Engineering Department, Trumpington Street, Cambridge, CB2 1PZ, UK {thw28,mg436,nm480,lmr46,phs26,djv27,sjy}@cam.ac.uk Abstract each individual processing component (Ward and Is- sar, 1994; Bohus and Rudnicky, 2009), statistical ap- proaches to SDS promise a domain-scalable frame- Moving from limited-domain natural lan- guage generation (NLG) to open domain is work which requires a minimal amount of human in- difficult because the number of semantic in- tervention (Young et al., 2013). Mrksi ˇ c ´ et al. (2015) put combinations grows exponentially with showed improved performance in belief tracking by the number of domains. Therefore, it is im- training a general model and adapting it to specific portant to leverage existing resources and ex- domains. Similar benefit can be observed in Gasi ˇ c ´ ploit similarities between domains to facilitate et al. (2015), in which a Bayesian committee ma- domain adaptation. In this paper, we propose chine (Tresp, 2000) was used to model policy learn- a procedure to train multi-domain, Recurrent Neural Network-based (RNN) language gen- ing in a multi-domain SDS regime. 
TI - Multi-domain Neural Network Language Generation for Spoken Dialogue Systems
JF - Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
DO - 10.18653/v1/n16-1015
DA - 2016-01-01
UR - https://www.deepdyve.com/lp/unpaywall/multi-domain-neural-network-language-generation-for-spoken-dialogue-81vmeTvn0R
DP - DeepDyve
ER -