Tools for Practical Dialogue Modelling

Raquel Fernández Rovira
www.ling.uni-potsdam.de/~raquel

David Schlangen
www.ling.uni-potsdam.de/~das

Andrea Corradini
www.ling.uni-potsdam.de/~andrea
This is a practical course were students will learn how to make use of some major techniques and methods commonly employed in dialogue modelling. The course will be conducted in English and it will be given in the form of a Blockseminar, in 4 full day sessions. It is expected that students are also attending the theoretical course Methods for Dialogue Modelling.

Class: Fridays 10-18h, starting on the 2nd of June 2006

Room: II.24.1.78/79 (Computerpool I)

Session One: Annotation and Analysis of Data (I)

02/06/2006 [ Raquel Fernández ] (introductory slides)

We will be working with data collected in an experiment run as part of our DEAWU project.
You can find the working files for today's session here

To annotate the data we will use the free version of MMAX. To install the program in the Windows PCs of the Computerpool, you'll have to create a folder named MMAX right under the H: directory. Then download this file (MMAX094.zip ), unzip it, and move the decompressed folder to the MMAX folder you just created.

Annotation schemes

Session Two: Text-to-Speech and Automatic Speech Recognition

09/06/2006 [ Andrea Corradini ]

MS SAPI 4 with focus on the following applications:

  • Text-to-Speech (TTS), Visemes and Sable for standardizing text to pass to TTS engines
  • MS Speech Recognizer and MS Voice to control the OS GUI by voice

Basic of Natural Language Understanding

  • ELIZA system and similar ones (Ikea, weakyjabby, chatter bots)
  • Sample Prolog source code to simulate commands to the OS entered in plain English

Overview of an existing Spoken Dialogue System

  • The NICE HCA system

Session Three: Annotation and Analysis of Data (II)

16/06/2006 [ Raquel Fernández & David Schlangen ]

In this session, we will continue with the tasks started on session 1. We will finish the annotation with MMAX, compute inter-annotator agreement with kappa, and analyse differences between conditions with the t-test. A detailed plan for this session can be found here

Statistical Analysis of Data: t-test

Session Four:

23/06/2006 [ Raquel Fernández & David Schlangen ]