|
|
Natural Language Generation in Saarbrücken |
On the Saarbrücken campus sites as well as at DFKI, many research activities are pursued in the field of Natural Language Generation (NLG). During the 1997 Workshop on NLG (a collection of all papers is also available as a DFKI document ) it was felt that too little is known about the total of these activities. So we decided to set up this Natural Language Generation Home Page in order to facilitate scientific exchange.
This web page serves as a jump station to the different project pages.
|
TAG-based generation at DFKI: In the research lab Intelligent User Interfaces at DFKI Saarbrücken there have been and currently are several ongoing projects for developing NLG components under the scientific coordination of Prof. Dr. Wolfgang Wahlster. A common kernel of each of the following projects consists in a syntactical generator based on tree adjoining grammars. Main emphasis was laid on efficient methods for incremental processing, microplanning, offline pre-computation of knowledge sources such as a HPSG, content-to-speech generation and automatic self-corrections. Current projects: Generation in Verbmobil and in UNL |
|
Shallow generation at DFKI: For small and medium-size application tasks, the production system TG/2 has been developed at DFKI's Language Technology Lab. By integrating canned text, templates, and context-free rules into a single formalism, TG/2 offers a high degree of flexibility for practical text generation tasks. The system is used in an appointment planning domain (project COSMA) and as part of a generator for German and French air-quality reports from environmental data (project TEMSIS). Further applications and languages are planned. You are welcome to test an online demo version of the TEMSIS report generator. |
|
PROVERB: Presentation of Machine-Found Proofs PROVERB is a system presenting machine-found proofs in natural language. It is embedded in the interactive proof development environment OMEGA. Settled between automated theorem proving (ATP) and natural language generation (NLG), PROVERB uses techniques of both these fields. |
|
Speech synthesis Logox Logox is a German text-to-speech system, which has been developped in the Institute of Phonetics, Saarbrücken and G-Data Software GmbH, Bochum. It is based on the Microsegment Synthesis (R), which aims at minimizing memory and processing requirements while preserving the natural speaking and voice characteristics of a speaker. All prosodic variations are carried out in the time-domain. Logox page in German |
|
Mary (in German) Mary is a highly modularised text-to-speech system. The preprocessing, morphology and syntax models available at the Department of computational linguistics and DFKI Saarbrücken have been combined with the Microsegment Synthesis (R) and prosodic models, developped in the Institute of Phonetics, Saarbrücken. Due to its modular structure Mary allows for testing the performance of different modules. |
| Jan Alexandersson, DFKI GmbH |
| Tilman Becker, DFKI GmbH |
| Stephan Busemann, DFKI GmbH |
| Ralf Benzmüller, Dept. 8.7 General Linguistics, University of the Saarland |
| Armin Fiedler, Dept. of Computer Science, University of the Saarland |
| Helmut Horacek, Dept. of Computer Science, University of the Saarland |
| Anne Kilger, DFKI GmbH |
| Peter Poller, DFKI GmbH |
| Elke Teich, Dept. of English Linguistics and Translation Science, University of the Saarland |
| Former Members: |
| Wolfgang Finkler, DFKI GmbH |
Design by Wolfgang Finkler and Renato Orsini.
Last modified Jan. 7, 1999 by TB;
Stephan Busemann Tilman Becker