Voice search of structured media data

Speech and Multimodal Interaction in Mobile Search

“…Researchers have proposed various approaches to tightly couple the two components for better overall system performance. References [12] and [31] propose using a query understanding (QU) module to bridge the gap between ASR and Search. The role of QU is to parse the ASR output into meaningful segments that contribute to high precision search.…”

Combining Slot-based Vector Space Model for Voice Book Search

“…Voice search [6] has been used in various applications: automated directory assistance system [8], consumer rating system [9], multimedia search [4], and book search [3]. Early voice search systems primarily focused on issues of ASR and search problems in locating business or residential phone listings [8].…”

“…Then, it has been extended to general web search such as Google's. Recent voice search systems have been applied to search for entries in large multimedia databases [4]. Although a simple string matching technique was used to measure the similarity of an ASR output string to entity values in the database [3], vector space models (VSM) have been widely used [8,9].…”

Wizards' dialogue strategies to handle noisy speech recognition

“…Although voice search can improve recognition [17], there will always be a residue of cases where the input is so noisy that voice search fails. In those cases, models of how wizards disambiguate among voice search returns or use them to ask informed questions can be used to further improve the system.…”