Download Estimating Spoken Dialog System Quality with User Models by Klaus-Peter Engelbrecht PDF

By Klaus-Peter Engelbrecht

Spoken conversation platforms have the capability to supply hugely intuitive person interfaces, as they enable structures to be managed utilizing average language. notwithstanding, the complexity inherent in ordinary language dialogs signifies that cautious trying out of the procedure needs to be conducted from the very starting of the layout strategy. This e-book examines how person types can be utilized to help such early reviews in methods: via working simulations of dialogs, and by means of estimating the standard judgments of clients. First, a layout setting aiding the production of conversation flows, the simulation of dialogs, and the research of the simulated information is proposed. How the standard of person simulations will be quantified with admire to their suitability for either formative and summative overview is then mentioned. the rest of the ebook is devoted to the matter of predicting caliber judgments of clients in line with interplay info. New modeling ways are offered, which approach the dialogs as sequences, and which enable wisdom concerning the judgment habit of clients to be integrated into predictions. All proposed tools are confirmed with instance evaluate studies.

Show description

Read Online or Download Estimating Spoken Dialog System Quality with User Models PDF

Best user experience & usability books

Speech Recognition HOWTO

Automated Speech attractiveness (ASR) on Linux is turning into more straightforward. a number of programs can be found for clients in addition to builders. This record describes the fundamentals of speech attractiveness and describes the various to be had software program.

Ambient Intelligence

Ambient intelligence is the imaginative and prescient of a expertise that might develop into invisibly embedded in our common atmosphere, current every time we'd like it, enabled through uncomplicated and easy interactions, attuned to all our senses, adaptive to clients and context-sensitive, and self sustaining. top of the range details entry and customized content material needs to be to be had to everyone, wherever, and at any time.

Facial Analysis from Continuous Video with Applications to Human-Computer Interface

Laptop imaginative and prescient algorithms for the research of video info are bought from a digicam geared toward the person of an interactive process. it's almost certainly worthy to reinforce the interface among clients and machines. those photograph sequences offer info from which machines can establish and preserve music in their clients, realize their facial expressions and gestures, and supplement different kinds of human-computer interfaces.

Computer-Supported Collaborative Learning in Higher Education

Computer-Supported Collaborative studying in larger schooling presents a source for researchers and practitioners within the quarter of computer-supported collaborative studying (also referred to as CSCL); fairly these operating inside a tertiary schooling surroundings. It contains articles of relevance to these drawn to either conception and perform during this zone.

Additional info for Estimating Spoken Dialog System Quality with User Models

Example text

5 Formative usability report, including a panel to select interactions (left), a panel showing a state graph representation of the interaction (right), and a panel where additional information about the interaction can be displayed (bottom) sophisticated models are discussed later in this book. As a deliverable of the MeMo project, it could be demonstrated that such a model is feasible also for web sites (Engelbrecht et al. 2008b). Although such models usually show low accuracy in terms of R2, especially for the prediction of independent data, previous work showed that the prediction of mean judgments for system configurations, aimed at here, can be quite accurate (Engelbrecht and Möller 2007).

Nevertheless, they seem to be less likely than other actions, as they did not appear in the experiment. The Recall of actions from the experiment is also very low, considering the much higher number of different actions in the simulation than in the experiment. 5 % of the actions found with real participants are not produced by the simulation and are thus not available for the improvement of the system. As expected, the results for actions in the context of the state they appeared in stay behind those for context-free analysis.

In addition, the system shows two features which are untypical for mixed-initiative dialogs, and which needed to be taken into account when adapting the user model: • At some points in the dialog, the user can state the name of the attribute she wants to specify next. ’’ Possible attributes are date, time, price range, food type, and location of the restaurant. • The system distinguishes between search values, which are used for the database search, and concepts, which are used for dialog management.

Download PDF sample

Rated 4.83 of 5 – based on 36 votes