Highly Scalable Speech Processing on Data Stream Management System

  • Shunsuke Nishii
  • Toyotaro Suzumura
Part of the Lecture Notes in Computer Science book series (LNCS, volume 7239)


Today we require sophisticated speech processing technologies that process massive speech data simultaneously. In this paper we describe the implementation and evaluation of a Julius-backended parallel and scalable speech recognition system on the data stream management system “System S” developed by IBM Research. Our experimental result on our parallel and distributed environment with 4 nodes and 16 cores shows that the throughput can be significantly increased by a factor of 13.8 when compared with that on a single core. We also demonstrate that the beam management module in our system can keep throughput and recognition accuracy with varying input data rate.


Speech Recognition Recognition Accuracy Beam Width Speech Data Word Error Rate 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Abadi, D.J., et al.: The Design of the Borealis Stream Processing Engine. In: Proc. CIDR, pp. 277–289 (2005)Google Scholar
  2. 2.
    Wolf, J., Bansal, N., Hildrum, K., Parekh, S., Rajan, D., Wagle, R., Wu, K.-L., Fleischer, L.K.: SODA: An Optimizing Scheduler for Large-Scale Stream-Based Distributed Computer Systems. In: Issarny, V., Schantz, R. (eds.) Middleware 2008. LNCS, vol. 5346, pp. 306–325. Springer, Heidelberg (2008)CrossRefGoogle Scholar
  3. 3.
    Gedik, B., et al.: A Code Generation Approach to Optimizing High-Performance Distributed Data Stream Processing. In: Proc. USENIX, pp. 847–856 (2009)Google Scholar
  4. 4.
    Arakawa, Y., et al.: A Study for a Scalability Evaluation Model of Spoken Dialogue System. Transactions of Information Processing Society of Japan 46(9), 2269–2278 (2005) (in Japanese)MathSciNetGoogle Scholar
  5. 5.
    Tatbul, N., et al.: Load Shedding in a Data Stream Manager. In: Proc. VLDB (2003)Google Scholar
  6. 6.
    Gedik, B., et al.: SPADE: The System S Declarative Stream Processing Engine. In: Proc. SIGMOD, pp. 1123–1134 (2008)Google Scholar
  7. 7.
    Amini, L., et al.: SPC: A Distributed, Scalable Platform for Data Mining. In: DM-SSP, pp. 27–37 (2006)Google Scholar
  8. 8.
    Jain, N., et al.: Design, implementation, and evaluation of the linear road benchmark on the stream processing core. In: International Conference on Management of Data, ACM SIGMOD, Chicago, IL (2006)Google Scholar
  9. 9.
    Young, S., et al.: The HTK book (for HTK Version 3.2) (2002)Google Scholar
  10. 10.
    Lee, A., et al.: Recent Development of Open-Source Speech Recognition Engine Julius. In: Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC (2009)Google Scholar
  11. 11.
    Lee, A.: Large Vocabulary Continuous Speech Recognition Engine Julius ver. 4. IEICE technical report. Speech 107(406), pp.307-312 (2007) (in Japanese)Google Scholar
  12. 12.
    Dixon, P.R., et al.: The Titech Large Vocabulary WFST Speech Recognition System. In: IEEE ASRU, pp. 443–448 (2007)Google Scholar
  13. 13.
    Lee, A., et al.: An Efficient Two-pass Search Algorithm using Word Trellis Index. In: Proc. ICSLP, pp. 1831–1834 (1998)Google Scholar
  14. 14.
    Itahashi, S., et al.: Development of ASJ Japanese newspaper article sentences corpus. Annual Meeting of Acoustic Society of Japan 1997(2), 187–188 (1997) (in Japanese)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2012

Authors and Affiliations

  • Shunsuke Nishii
    • 1
  • Toyotaro Suzumura
    • 1
    • 2
  1. 1.Tokyo Institute of TechnologyTokyoJapan
  2. 2.IBM Research - TokyoKanagawaJapan

Personalised recommendations