Stochastic Semantic Analysis

In human-computer dialogue systems, the task of Spoken Language Understanding (SLU) system is to process the input acoustic utterance and transform it into a semantic representation. The goal of semantic analysis is to represent what the subject intended to say. This thesis presented overview of the current state-of-the-art methods for statistical semantic analysis. In comparison to the expert based systems, the main advantage of stochastic approaches is the ability to train the model from data. Furthermore, systems based on statistical models can be easily ported to other domains. However, the amount of the annotation effort must be also taken into account when developing the stochastic semantic analysis system. In this thesis, fundamental stochastic models along with the training and evaluation of these models are described. Copies of this report are available on http://www.kiv.zcu.cz/publications/ or by surface mail on request sent to the following address: University of West Bohemia in Pilsen Department of Computer Science and Engineering Univerzitńı 8 30614 Pilsen Czech Republic Copyright c © 2009 University of West Bohemia in Pilsen, Czech Republic