Talking to machines (statistically speaking)

By Susannah Young. 2002

158 citations recorded by the source corpus.

Statistical methods have long been the dominant approach in speech recognition and probabilistic modelling in ASR is now a mature technology. The use of statistical meth-ods in other areas of spoken dialogue is however more recent and rather less mature. This paper reviews spoken dialogue systems from a statistical modelling perspective. The complete system is ﬁrst presented as a partially observable Markov decision process. The various sub-components are then exposed by introducing appropriate intermediate variables. Samples of existing work are reviewed within this framework, including dialogue control and optimisation, semantic interpretation, goal detection, natural language generation and synthesis.

Subfields: Artificial Intelligence, Natural Language Processing

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

In Natural Language Processing

Citations in the corpus

Efficient Natural Language Response Suggestion for Smart Reply (2017, background): “These systems are brittle, and progress is slow [31].”
On-line policy optimisation of spoken dialogue systems via live interaction with human subjects (2011, unknown): “The statistical approach to dialogue modelling has been proposed as a means of building domain independent dialogue systems, trainable from data and robust to speech understanding errors [1], [2].”
Domain-independent User Simulation with Transformers for Task-oriented Dialogue Systems (2021, background): “Solving this problem typically requires tracking and planning (Young, 2002).”
The Hidden Information State model: A practical framework for POMDP-based spoken dialogue management (2010, background): “As has been argued previously, taking a statistical approach to spoken dialogue system design provides the opportunity for solving many of the above problems in a flexible and principled way ( Young, 2002 ).”
A trainable generator for recommendations in multimodal dialog (2003, methodology): “SLG promises improved system output, portability and customiz ability to different application domains, dialogue situations , and users, through a combination of application-independent l inguistic modules [5], and methods for automatically trainin g parameters for particular applications [7, 12, 4,”
Sample Efficient Deep Reinforcement Learning for Dialogue Systems With Large Action Spaces (2018, methodology): “To overcome this, the policy optimisation problem has been formulated as a reinforcement learning problem [1], [2], [3].”
Spoken language understanding using the Hidden Vector State Model (2006, methodology): “The most likely interpretation is then given by the joint optimization over C and W n This joint optimization avoids the error incurred by sequentially decodingˆW and thenˆC (Young, 2002).”
MASSIVE: A 1M-Example Multilingual Natural Language Understanding Dataset with 51 Typologically-Diverse Languages (2022, background): “Our particular focus is on the NLU component of Spoken Language Understanding (SLU), in which raw audio is first converted to text before NLU is performed (Young, 2002; Wang et al., 2005; Tur and Mori, 2011).”
Spoken Content Retrieval—Beyond Cascading Speech Recognition with Text Retrieval (2015, background): “example, which was actually built in early years by cascading ASR, natural language understanding, dialogue management, natural language generation and TTS [360].”
The Hidden Information State Approach to Dialog Management (2007, background): “Rather than MDPs, Partially Observable MDPs (POMDPs) potentially provide a much more powerful framework for modeling dialog systems since they provide an explicit represention of uncertainty [3, 4].”
Language modelling ’ s generative model : is it rational ? (2004, unknown): “It might turn out that LM technology only works well for ‘rough’ tasks like retrieval or simple task forms like ‘keyphrase’ summarising, or for limited task applications like highly constrained dialogue inquiry (Young 2002).”
POMDP-Based Statistical Spoken Dialog Systems: A Review (2013, background): “During the last few years, a new approach to dialogue management has emerged based on the mathematical framework of partially observable Markov decision processes (POMDPs2) [9], [10], [11].”