Effect of Confusion Network Combination on Speech Recognition System for Editing
This paper describes confusion network combination (CNC), which integrates multiple confusion networks, and its effectiveness on a system for editing transcription by a speech recognizer. It has been proposed that integration by CNC produces better recognition performance. We believe that this improves the working efficiency of a human editor for correcting errors. We utilized confusion networks from two recognition systems for CNC. Integration by the CNC method is performed by combining the networks based on posterior probabilities attached to each word. The experimental results showed that the improvement of recognition performance by the CNC method could reduce the working time of human editors by 5.1 seconds, on average, compared to the working time required without using this method.
Satoshi Ishimaru Hiromitsu Nishizaki Yoshihiro Sekiguchi
Department of Education Interdisciplinary Graduate School of Medicine and Engineering, University of Department of Research Interdisciplinary Graduate School of Medicine and Engineering, University of
国际会议
2011亚太信号与信息处理协会年度峰会(APSIPAASC 2011)
西安
英文
1-4
2011-10-18（万方平台首次上网日期，不代表论文的发表时间）