Ziel
The research project aims at development of a multi-lingual and multi-functional information retrieval system.
Basis of the project were created within an ESPRIT project; on the experimental information dialogue system "SUNDIAL" for English, German, French, and Italian. In the proposed research, we will develop a prototype of information dialogue system for Czech, Slovak and Slovenian. Within the project we want to show that :
- using the SUNDIAL approach,
- an information dialogue system can be designed for any of the European languages.
The project will start with a development of a single prototype dialogue system out of the four systems, which is able to recognize the users language and consequently lead a dialogue in this language about one of several possible topics. Therefore a language identification module will be implemented. The architecture of the system will enable the use of different application databases, such as flight or train timetable enquiries and reservations.
The systems will be implemented in a two-stage approach: First a system will be trained with read speech using the German SUNDIAL demonstrator. The eastern partners will assist the adaptation of this demonstrator to spontaneous speech (for instance by developing stochastic language models for spontaneous speech). Doing this they will learn how to adapt their own systems to spontaneous speech input in the second step. From the linguistic analysis of the user utterances we will use a language independent approach (keyword classification trees) and language dependent substring parsers. As a result a large scale databases, necessary for the training of the system, will be made available to the European speech community.
Programm/Programme
Thema/Themen
Data not availableAufforderung zur Vorschlagseinreichung
Data not availableFinanzierungsplan
CSC - Cost-sharing contractsKoordinator
91058 Erlangen
Deutschland