14th International Congress of Phonetic Sciences (ICPhS-14)
San Francisco, CA, USA
Speech databases represent an information source essential for the continued development of spoken language theories and applications. However, due to a lack of standards in data formats and annotation conventions, extracting information from different speech databases and transforming it into generic knowledgebearing structural frameworks is often difficult. At least two different methods are possible for generating structure. One way is to develop a specific interpreter consisting of a parser, compiler, and linker to handle every encountered combination of database format, language, annotation syntax, etc. Although direct and potentially computationally efficient, this solution is expensive and time consuming to implement in systems requiring generic access over many diverse databases. Another method is to apply a knowledge-based approach where rules are used to form structures for speech utterances. This paper describes the speech database interpreters in the QuickSig database environment as well as formulates the requirements of a proposed knowledgebased system.
Bibliographic reference. Altosaar, Toomas / Vainio, Martti / Karjalainen, Matti (1999): "Transforming information in speech databases into knowledge", In ICPhS-14, 1737-1740.