Abstract | ||
---|---|---|
A key advantage of taking a statistical approach to spoken dialogue systems is the ability to formalise dialogue policy design as a stochastic optimization problem. However, since dialogue policies are learnt by interactively exploring alternative dialogue paths, conventional static dialogue corpora cannot be used directly for training and instead, a user simulator is commonly used. This paper describes a novel statistical user model based on a compact stack-like state representation called a user agenda which allows state transitions to be modeled as sequences of push- and pop-operations and elegantly encodes the dialogue history from a user's point of view. An expectation-maximisation based algorithm is presented which models the observable user output in terms of a sequence of hidden states and thereby allows the model to be trained on a corpus of minimally annotated data. Experimental results with a real-world dialogue system demonstrate that the trained user model can be successfully used to optimise a dialogue policy which outperforms a hand-crafted baseline in terms of task completion rates and user satisfaction scores. |
Year | DOI | Venue |
---|---|---|
2009 | 10.1109/TASL.2008.2012071 | IEEE Transactions on Audio, Speech & Language Processing |
Keywords | Field | DocType |
dialogue history,conventional static dialogue corpus,observable user output,dialogue policy design,alternative dialogue path,dialogue system,hidden agenda user simulation,trained user model,novel statistical user model,real-world dialogue system,dialogue policy,state transition,design optimization,simulation model,limiting,helium,markov decision process,stochastic optimization,parameter estimation,uncertainty,delta modulation,speech synthesis,stochastic processes,history,noise measurement,speech processing,speech recognition,user model,data models | Data modeling,Speech processing,Stochastic optimization,Markov process,Computer science,Expectation–maximization algorithm,Markov decision process,Speech recognition,User modeling,Artificial intelligence,Statistical model,Machine learning | Journal |
Volume | Issue | ISSN |
17 | 4 | 1558-7916 |
Citations | PageRank | References |
7 | 0.52 | 17 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Jost Schatzmann | 1 | 688 | 37.81 |
S. Young | 2 | 134 | 7.84 |