6533b856fe1ef96bd12b26ca

RESEARCH PRODUCT

A Sub-Symbolic Approach to Word Modelling for Domain Specific Speech Recognition

Giovanni PilatoGiorgio VassalloSalvatore GaglioFrancesco Agostaro

subject

Computer sciencebusiness.industrySpeech recognitionMachine learningcomputer.software_genreDomain (software engineering)Speech enhancementMetric (mathematics)Artificial intelligenceLanguage modelHellinger distanceHidden Markov modelbusinesscomputerNatural languageWord (computer architecture)

description

In this work a sub-symbolic technique for automatic, data driven language models construction is presented. Such a technique can be used to arrange a language-modelling module, which can be easily integrated in existing speech recognition architectures, such as the well-found HTK architecture. The proposed technique takes advantages from both the traditional LSA approach and from a novel application of a probability space metric known as "Hellinger's distance". Experimental trials are also presented, in order to validate the proposed approach.

yearjournalcountryeditionlanguage
2006-10-11Seventh International Workshop on Computer Architecture for Machine Perception (CAMP'05)
https://doi.org/10.1109/camp.2005.8