期刊文献+

Keyword Confidence Evaluation Algorithm Based on Word Activation Forces

Keyword Confidence Evaluation Algorithm Based on Word Activation Forces
下载PDF
导出
摘要 Many Spoken Term Detection (STD) systems use query expansion to return an increased number of keyword candidates and make posterior probability a confidence feature to reject false alarms. However, some keyword candidates hold high posterior probability although these are not recognized correctly. We investigate the Word Activation Force (WAF) model that compatibly encodes syntactical and semantic information into sparse coding directed networks. A high-level confidence feature Keyword Activation Force (KAF) based on WAF is proposed. KAF can be used for detecting false alarms by considering information about the neighbors to provide a more reliable and accurate keyword affinity. Compared with the baseline system, a relative reduction of 30.94% in average error rate could be achieved when KAF is combined with the posterior probability and the language model score. Many Spoken Term Detection (STD) systems use query expansion to return an increased number of keyword candidates and make posterior probability a confidence feature to reject false alarms. However, some keyword candidates hold high posterior probability although these are not recognized correctly. We investigate the Word Activation Force (WAF) model that compatibly encodes syntactical and semantic information into sparse coding directed networks. A high-level confidence feature Keyword Activation Force (KAF) based on WAF is proposed. KAF can be used for detecting false alarms by considering information about the neighbors to provide a more reliable and accurate keyword affinity. Compared with the baseline system, a relative reduction of 30.94% in average error rate could be achieved when KAF is combined with the posterior probability and the language model score.
出处 《China Communications》 SCIE CSCD 2012年第11期54-62,共9页 中国通信(英文版)
基金 supported by National Natural Science Foundation of China under Grants No.61005004,No.61175011,No.61171193 the Next-Generation Broadband Wireless Mobile Communications Network Technology Key Project under Grant No.2011ZX03002-005-01 the 111 Project under Grant No.B08004 Scientific Research Foundation for the Returned Overseas Chinese Scholars,State Education Ministry
关键词 speech recognition STD KAF 候选关键字 活力 Word 评估算法 信誉 后验概率 定期检测 语义信息
  • 相关文献

参考文献21

  • 1NIST. The Spoken Term Detection (STD) 2006 Evaluation Plan [EB/OL]. httpz/www, nist. gov/speech/tests/std/docs/ std06-evalplan-vlO.pdf,2006.
  • 2ALLAUZEN C, MOHR! M, SARACLAR M. General Index?ation of Weighted Autorrata-Application to Spoken Utter?ance Retrieval[C]// Proceedings of HLT-NAACL: May 2-7, 2004, Boston, Massachusetts, USA. Association for Com?putational Linguistics, 2004: 33-40.
  • 3ZHANG Pengyuan, SHAO Jian, ZHAO Qingwei, et ol. Keyword Spotting Based on Syllable Confusion NetwOlk[CY/ Proceed?ings of the 3rd International Conference on Natural Corrputa?tion: August 24-27, 2f.'JJ7, Haikou, Hainan, China. Washington, DC, USA: IEEE Corrputer Society, 2f.'JJ7: 656-659.
  • 4WESSEL F, SCHLUTER R, MACHEREYK, et al. Confidence Measures for Large Vocabulary Continuous Speech Recog?nition[J]. IEEE Transactions on Speech and Audio Process?ing, 2001, 9(3): 288-298.
  • 5CHEN Tzanhwei, CHEN Berlin, WANG Hsinmin. On Using Entropy Inforrmtion to Irrprove Posterior Probability-Based Confidence Measures [C]// Proceedings of the 9th Interna?tional Syrrposiurn on Chinese Spoken Language Process?ing: December 13-16,2006, Kent Ridge, Singapore. Springer?Verlag, 2006: 454-463.
  • 6RAZlK J, MELLA 0, FOHR D, et al. Local Word Confi?dence Measure Using Word Graph and N-Best List [C]// Proceedings of the 9th European Conference on Speech Cormrnnication and Technology: Septerroer 4-8, 2005, Lis?bon, PortugaL ISCA Archive, 2005: 3369-3372.
  • 7FABIAN T, UEB R, RUSKE G, et al. Irmact of Word Graph Density on the Quality of Posterior Probability Based Con?fidence Measures [C]// Proceedings of the 8th European Conference on Speech Commmication and Technology: September 1-4, 2003, Geneva, Switzerland. ISCA Archive, 2003: 917-920.
  • 8ROSE R, JUANG B, LEE C. A Training Procedure for Verif?ying String Hypotheses in Continuous Speech Recognition [C]// Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing: May 8-12, 1995, Detroit, Michigan, USA. IEEE Press, 1995: 281-284.
  • 9HUANG Zengyang. Hierarchical Network Concept Theory [M],Beijing, China: Tsinghua University Press, 1998.
  • 10SARlKA YA R, GAO Y, PICHENY M, et al. Semantic Confi?dence Measurerrent for Spoken Dialog System; [J]. IEEE Transactions on Speech and Audio Processing, 2005,13(4): 534-545.

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部