How to access audio files of large data bases using in-car speech dialogue systems

Semantic Frame‐Based Spoken Language Understanding

“…The applications include both telephone only services and multi-modal services on mobile devices. Other voice search applications include music/video management (Mann et al 2007;Song et al 2009), business and product reviews (Zweig et al 2007), stock price quote, and conference information systems (Andreani et al 2006;Bohus et al 2007). Recently the task has been extended to using voice queries for Web search from mobile devices, as manifested by the commercial systems from Google and Yahoo.…”

“…Separate efforts have been made on conference information systems by Carnegie Mellon University (Bohus et al 2007) and by the collaboration among AT&T, ICSI, Edinburgh University and Speech Village (Andreani et al 2006), where users can request information about thousands of papers published in a conference. In entertainment, Daimler has investigated digital music management in automobiles (Mann et al 2007), and Ford and Microsoft have introduced the commercial dashboard device Ford SYNC that allows in-car music search. Like the business DA applications, many new voice search applications call for research activities in search/SLU and dialog management in addition to speech recognition.…”

“…With the broad adoption of mobile devices and the availability of wireless access to the internet, many companies are actively engaged in the space of voice search on mobile or incar devices (Mann et al 2007). Google has introduced speech recognition for Google Apps on iPhone, which allows users to use speech input for business or general Web search.…”

“…and use voice to control the device, including a media search with voice input. New research challenges include multimodal (GUI with touch screen and speech) user interfaces (Acero et al 2008;Mann et al 2007) and efficient and scalable client-server architectures.…”

“…The dialog starts with prompting a user for the category information about the item they are looking for to narrow down the downstream LM and search spaces. The category can be the locality information in a DA system (Bacchiani et al 2008;Yu et al 2007); the business/product separation (national business, local business or product) in a voice rating system (Zweig et al 2007), or a "search-by" attribute of the music meta-data (e.g., title, album, genre, artist, etc) in a music search dialog system (Mann et al 2007). A category-specific LM is subsequently used to recognize the user's query containing the listing information, and the search component looks for the listing in a category-specific database.…”