Morphosyntaktische Modelle für statistische maschinelle Übersetzung

• Antragsteller: Professor Dr. Hinrich Schütze
• Fachliche Zuordnung: Allgemeine und Vergleichende Sprachwissenschaft, Experimentelle Linguistik, Typologie, Außereuropäische Sprachen
• Förderung: Förderung von 2009 bis 2018
• Projektkennung: Deutsche Forschungsgemeinschaft (DFG) - Projektnummer 123083856

• Erstellungsjahr: 2018

Zusammenfassung der Projektergebnisse

The project resulted in a large number of scientific innovations. Papers about these innovations were published in some of the best conferences and journals in natural language processing. Particularly important advances were: - The ability to integrate classification models into SMT decoders. - Classification models for inflectional and syntactic choice in all phrase-based, hierarchical and syntax-based SMT. - Reaching the capability of generating unseen inflected forms in SMT. Important source code was published in both the open source Moses toolkit, and on our own local repository: http://cistern.cis.lmu.de There were no major deviations in the project plan and all workpackages were carried out successfully.