Van Heerden, CJDe Villiers, PBarnard, EDavel, MH2012-03-262012-03-262011-11Van Heerden, CJ, De Villiers, P, Barnard, E and Davel, MH. Processing spoken lectures in resource-scarce environments. Proceedings of the Twenty-Second Annual Symposium of the Pattern Recognition Association of South Africa, Vanderbijlpark, South Africa, 22-25 November 2011, pp 138-143978-0-620-51914-4http://www.prasa.org/proceedings/2011/prasa2011-25.pdfhttp://hdl.handle.net/10204/5681Proceedings of the Twenty-Second Annual Symposium of the Pattern Recognition Association of South Africa, Vanderbijlpark, South Africa, 22-25 November 2011Initial work towards processing Afrikaans spoken lectures in a resource-scarce environment is presented. Two approaches to acoustic modeling for eventual alignment are compared: (a) using a well-trained target-language acoustic model and (b) using an acoustic model from another language, in this case American English. The authors show that while target-language acoustic models are preferable, similar performance can be achieved by repeatedly bootstrapping with the American English model, segmenting and then adapting or training new models using the segmented spoken lectures. The eventual systems perform quite well, aligning more than 90% of a selected set of target words successfully.enAfrikaans spoken lecturesAfrikaans spoken lectures processingSegmented spoken lecturesConference PresentationVan Heerden, C., De Villiers, P., Barnard, E., & Davel, M. (2011). Processing spoken lectures in resource-scarce environments. PRASA. http://hdl.handle.net/10204/5681Van Heerden, CJ, P De Villiers, E Barnard, and MH Davel. "Processing spoken lectures in resource-scarce environments." (2011): http://hdl.handle.net/10204/5681Van Heerden C, De Villiers P, Barnard E, Davel M, Processing spoken lectures in resource-scarce environments; PRASA; 2011. http://hdl.handle.net/10204/5681 .TY - Conference Presentation AU - Van Heerden, CJ AU - De Villiers, P AU - Barnard, E AU - Davel, MH AB - Initial work towards processing Afrikaans spoken lectures in a resource-scarce environment is presented. Two approaches to acoustic modeling for eventual alignment are compared: (a) using a well-trained target-language acoustic model and (b) using an acoustic model from another language, in this case American English. The authors show that while target-language acoustic models are preferable, similar performance can be achieved by repeatedly bootstrapping with the American English model, segmenting and then adapting or training new models using the segmented spoken lectures. The eventual systems perform quite well, aligning more than 90% of a selected set of target words successfully. DA - 2011-11 DB - ResearchSpace DP - CSIR KW - Afrikaans spoken lectures KW - Afrikaans spoken lectures processing KW - Segmented spoken lectures LK - https://researchspace.csir.co.za PY - 2011 SM - 978-0-620-51914-4 T1 - Processing spoken lectures in resource-scarce environments TI - Processing spoken lectures in resource-scarce environments UR - http://hdl.handle.net/10204/5681 ER -