The best of both worlds: unifying conventional dialog systems and POMDPs

By Jason Williams. 2008

100 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

End-to-end LSTM-based dialog control optimized with supervised and reinforcement learning (2016, neutral): “Business rules can be incorporated, in a similar manner to our approach (Williams, 2008).”
Hybrid Code Networks: practical and efficient end-to-end dialog control with supervised and reinforcement learning (2017, background): “The first decomposes a dialog system into a pipeline, typically including language understanding, dialog state tracking, action selection policy, and language generation (Levin et al., 2000; Singh et al., 2002; Williams and Young, 2007; Williams, 2008; Hori et al., 2009; Lee et al., 2009; Griol et a”
Synthesis Lectures on Human Language Technologies (2009, background): “Given the commercial need to satisfy customers and to be able to fix design problems, designers will wish to keep more control over the design of their systems and, in any case, are unlikely to be able to remedy problems of learned dialogue policies without a deep knowledge of RL (Williams, 2008).”
Optimising a handcrafted dialogue system design (2010, background): “Other studies [5, 6] propose to mix a handcrafted SDS with an MDP based SDS.”
On-line policy optimisation of spoken dialogue systems via live interaction with human subjects (2011, background): “This can be mitigated to some extent by using hand-crafted rules to define the set of permissible actions f r every summary state [18].”
POMDP-Based Statistical Spoken Dialog Systems: A Review (2013, methodology): “entries; an indication of which system actions are available; properties of the dialog history, such as whether the top user goal/partition has been confirmed; the most likely previous user action; or combinations of these features [28], [37], [40], [42], [43].”
Natural Language Generation as Incremental Planning Under Uncertainty: Adaptive Information Presentation for Statistical Dialogue Systems (2014, background): “Note that the possibility of constraining statistical policies with hard-coded rules is increasingly required for developing commercial dialogue systems [65].”