Towards the integration of speech and natural language into human-computer interaction systems

摘要

This paper first surveys and classifies applications of voice humancomputeruddialogue. The advantages and limits of speech as a means ofudcommunication between users and software are then considered .udA major problem in the development of user interfaces with a voiceudcomponent, besicles the choice of appropriate applications, is speechudrecognition, especially continuons speech recognition .udAs approaches differ according to the type of application, we firstudsummarize problems and techniques specific to voice data input ; as anudexample, we briefly describe the speech recognition approach that we haveudadopted for the dictation machine that we are developing in our Laboratory Then, the case of voice dialogue understanding and management isudconsidered. To illustraie the discussion, the architecture and functionalitiesudofsome prototypes that we have implemented are presented :for instance,udan E-mail system and a Sonar control software . Finally, we present auddialogue manager DIAL capable of helping/guiding the user in complexudcognitive activities that we are currently developing and implementing .