Language Resources for Portable Multilingual Spoken Dialogue Systems

Speech services are becoming an increasingly important market segment for Europe's SMEs. A major roadblock in producing low-cost high-quality spoken dialogue system (SDS) applications is the lack of linguistic resources that will enable the porting of SDS to new domains or languages. Currently, porting spoken dialogue systems to new domains is time-consuming, requires expertise in language engineering and iterative tuning. Although by now mature technologies exist that can facilitate SDS porting to new domains, e.g. ontology enrichment, linking text data to ontologies, automatic grammar induction, these technologies have not penetrated through to commercial systems yet. In PortDial, we bring together European SMEs that are developing state-of-the-art spoken dialogue systems and the handcrafted semantic components underlying such systems with research institutions at the forefront of progress in the automatic creation or enrichment of semantic language resources. Together with the commercial partners, we have identified a list of mature technologies that can help automate (or machine-assist) the process for the creation of SDS concepts, services and grammars. PortDial will result in 1) a commercial platform for quick prototyping of interactive spoken dialogue applications to new domains and languages, 2) the corresponding multilingual collections of concepts-services-grammars for specific application domains (marketed separately), and 3) a multilingual linked-data ontological corpus that can be freely used for spoken dialogue research and prototyping for non-commercial purposes. The resulting platform will reduce the effort for prototyping new application domains by up to 30%, and up to 50% for porting the application to new languages. The SDS linguistic resources will lower the barrier to entry for European SMEs to SDS technologies, allowing for inexpensive proof-of-concept demonstrator development, opening up new markets and application domains.