Van Heerden, CSchalkwyk, JStrope, B2009-10-122009-10-122009-09Van Heerden, C, Schalkwyk, J and Strope, B. 2009. Language modeling for what-with-where on GOOG-411. 10th Annual Conference of the International Speech Communication Association (Interspeech 2009). Brighton, UK, 6-10 September, 2009. pp 991-9941990-9772http://hdl.handle.net/10204/364810th Annual Conference of the International Speech Communication Association (Interspeech 2009). Brighton, UK, 6-10 September 2009This paper describes the language modelling (LM) architectures and recognition experiments that enabled support of 'what-with-where' queries on GOOG-411. First the paper compares accuracy trade-offs between a single national business LM for business queries and using many small models adapted for particular cities. Experimental evaluations show that both approaches lead to comparable overall accuracy. Differences in the distributions of errors also lead to improvements from a simple combination. The researchers then optimize variants of the national business LM in the context of combined business and location queries from the web, and finally evaluate these models on a recognition test from the recently fielded ’what-with-where’ system.enLanguage modeling architecturesWhat-with-where queriesGOOG-411Speech recognitionVoice searchHierarchical language model10th Annual Conference of the International Speech Communication AssociationInterspeech 2009Speech communicationHuman language technologiesConference PresentationVan Heerden, C., Schalkwyk, J., & Strope, B. (2009). Language modeling for what-with-where on GOOG-411. International Speech Communication Association. http://hdl.handle.net/10204/3648Van Heerden, C, J Schalkwyk, and B Strope. "Language modeling for what-with-where on GOOG-411." (2009): http://hdl.handle.net/10204/3648Van Heerden C, Schalkwyk J, Strope B, Language modeling for what-with-where on GOOG-411; International Speech Communication Association; 2009. http://hdl.handle.net/10204/3648 .TY - Conference Presentation AU - Van Heerden, C AU - Schalkwyk, J AU - Strope, B AB - This paper describes the language modelling (LM) architectures and recognition experiments that enabled support of 'what-with-where' queries on GOOG-411. First the paper compares accuracy trade-offs between a single national business LM for business queries and using many small models adapted for particular cities. Experimental evaluations show that both approaches lead to comparable overall accuracy. Differences in the distributions of errors also lead to improvements from a simple combination. The researchers then optimize variants of the national business LM in the context of combined business and location queries from the web, and finally evaluate these models on a recognition test from the recently fielded ’what-with-where’ system. DA - 2009-09 DB - ResearchSpace DP - CSIR KW - Language modeling architectures KW - What-with-where queries KW - GOOG-411 KW - Speech recognition KW - Voice search KW - Hierarchical language model KW - 10th Annual Conference of the International Speech Communication Association KW - Interspeech 2009 KW - Speech communication KW - Human language technologies LK - https://researchspace.csir.co.za PY - 2009 SM - 1990-9772 T1 - Language modeling for what-with-where on GOOG-411 TI - Language modeling for what-with-where on GOOG-411 UR - http://hdl.handle.net/10204/3648 ER -