Davel, MHVan Heerden, CKleynhans, NBarnard, E2012-04-162012-04-162011-08Davel, MH, Van Heerden, C, Kleynhans, N and Barnard, E. Efficient harvesting of Internet audio for resource-scarce ASR. 12 Annual Conference of the International Speech Communication Association (Interspeech 2011), Florence, Italy, 27-31 August 20119781618392701http://hdl.handle.net/10204/576912 Annual Conference of the International Speech Communication Association (Interspeech 2011), Florence, Italy, 27-31 August 2011Spoken recordings that have been transcribed for human reading (e.g. as captions for audiovisual material, or to provide alternative modes of access to recordings) are widely available in many languages. Such recordings and transcriptions have proven to be a valuable source of ASR data in well-resourced languages, but have not been exploited to a significant extent in under-resourced languages or dialects. Techniques used to harvest such data typically assume the availability of a fairly accurate ASR system, which is generally not available when working with resourcescarce languages. In this work, the authors define a process whereby an ASR corpus is bootstrapped using unmatched ASR models in conjunction with speech and approximate transcriptions sourced from the Internet. They introduce a new segmentation technique based on the use of a phone-internal garbage model, and demonstrate how this technique (combined with limited filtering) can be used to develop a large, high-quality corpus in an underresourced dialect with minimal effort.enSpeech recognitionUnder-resourced languagesGarbage modelingAutomatic speech recognition (ASR)Conference PresentationDavel, M., Van Heerden, C., Kleynhans, N., & Barnard, E. (2011). Efficient harvesting of Internet audio for resource-scarce ASR. The International Speech Communication Association. http://hdl.handle.net/10204/5769Davel, MH, C Van Heerden, N Kleynhans, and E Barnard. "Efficient harvesting of Internet audio for resource-scarce ASR." (2011): http://hdl.handle.net/10204/5769Davel M, Van Heerden C, Kleynhans N, Barnard E, Efficient harvesting of Internet audio for resource-scarce ASR; The International Speech Communication Association; 2011. http://hdl.handle.net/10204/5769 .TY - Conference Presentation AU - Davel, MH AU - Van Heerden, C AU - Kleynhans, N AU - Barnard, E AB - Spoken recordings that have been transcribed for human reading (e.g. as captions for audiovisual material, or to provide alternative modes of access to recordings) are widely available in many languages. Such recordings and transcriptions have proven to be a valuable source of ASR data in well-resourced languages, but have not been exploited to a significant extent in under-resourced languages or dialects. Techniques used to harvest such data typically assume the availability of a fairly accurate ASR system, which is generally not available when working with resourcescarce languages. In this work, the authors define a process whereby an ASR corpus is bootstrapped using unmatched ASR models in conjunction with speech and approximate transcriptions sourced from the Internet. They introduce a new segmentation technique based on the use of a phone-internal garbage model, and demonstrate how this technique (combined with limited filtering) can be used to develop a large, high-quality corpus in an underresourced dialect with minimal effort. DA - 2011-08 DB - ResearchSpace DP - CSIR KW - Speech recognition KW - Under-resourced languages KW - Garbage modeling KW - Automatic speech recognition (ASR) LK - https://researchspace.csir.co.za PY - 2011 SM - 9781618392701 T1 - Efficient harvesting of Internet audio for resource-scarce ASR TI - Efficient harvesting of Internet audio for resource-scarce ASR UR - http://hdl.handle.net/10204/5769 ER -