Partially observable Markov decision processes for spoken dialog systems

Spoken language understanding from unaligned data using discriminative classification models

“…Results show that STCs outperform a handcrafted rule-based parser previously used for collecting dialogue data [12,13], for both clean utterances and noisy speech recognition outputs. Our method also also improves on the HVS model [4] on the ATIS dataset, and it performs only 1.4% worse than more complex iterative grammar induction techniques [10].…”

“…The data includes the transcription of the top hypothesis of the ATK speech recogniser, which allows us to evaluate the robustness of our models to recognition errors (word error rate = 34.4%). We also compare our models with the handcrafted Phoenix grammar [14] used in the trials [12,13]. The Phoenix parser implements a partial matching algorithm that was designed for robust spoken language understanding.…”

“…The dialogues were collected through user trials in which users searched for information about a specific venue by interacting with a dialogue system in a noisy background. These dialogues were previously used for training dialogue management strategies [12,13]. The semantic representation of the user utterance consists of a root dialogue act type and a set of slots which are either unbound or associated with a child value.…”

A statistical approach to spoken dialog systems design and evaluation

“…Models of this kind have been widely used for speech recognition and also for language understanding (Levin and Pieraccini, 1995), (Minker et al, 1999), (Segarra et al, 2002), (He and Young, 2003), (Esteve et al, 2003). Even though in the literature there are models for dialog managers that are manually designed, over the last few years, approaches using statistical models to represent the behavior of the dialog manager have also been developed (Levin et al, 2000), (Torres et al, 2003), , (Williams and Young, 2007). These approaches are usually based on modeling the different processes probabilistically and learning the parameters of the different statistical models from a dialog corpus.…”

Statistical framework for a Spanish spoken dialogue corpus

“…This last discourse theory (Austin (1962)) focuses on communicative acts performed through speech and is the framework in which many authors try to model the structure of dialogue. From a more practical point of view and partially based on this theoretical approach, specific solutions have been proposed to model discourse in dialogue problems using a wide range of methods: dialogue grammars (McTear et al (2000)), information state (Bos et al (2003)) and the reinforcement learning framework (Williams and Young (2007)), among others. Independent of the method, many of these proposals make use of Dialogue Acts (DA) to model the local structure of the dialogue.…”