Topic identification techniques applied to dynamic language model adaptation for automatic speech recognition. Issue 1 (January 2015)

• Record Type: Journal Article
• Authors: Echeverry-Correa, J.D.; Ferreiros-López, J.; Coucheiro-Limeres, A.; Córdoba, R.; Montero, J.M.
• Is Part Of: Expert systems with applications. Volume 42:Issue 1(2015)
• Page Start: 101
• Page End: 112
• Publication Date: 2015-01
• Subjects: Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2014.07.035
• Languages: English
• ISSNs: 0957-4174

Abstract

<abstract xml:lang="en" abstract-type="author-highlights" id="ab005"> <title id="st150">Highlights</title> <sec> <p id="sp005"> <list id="l0005"> <list-item id="u0005"> <label></label> <p id="p0510">We present an approach for the dynamic adaptation of the LMs used by a speech recognizer based on topic identification.</p> </list-item> <list-item id="u0010"> <label></label> <p id="p0515">For each audio segment the system interpolates a static LM with a topic-based LM in a two stages recognition architecture.</p> </list-item> <list-item id="u0015"> <label></label> <p id="p0520">The interpolation weights are computed based on different sources of information (distance of LMs and topic similarity).</p> </list-item> <list-item id="u0020"> <label></label> <p id="p0525">We evaluate different strategies for generating topic-specific LMs and for the adaptation of the LM used in the final stage.</p> </list-item> <list-item id="u0025"> <label></label> <p id="p0530">Evaluation shows a significant reduction of the error rates for both tasks (topic identification and speech recognition).</p> </list-item> </list> </p> </sec> </abstract>