Advertisement

Cross-Lingual Adaptation of Broadcast Transcription System to Polish Language Using Public Data Sources

  • Jan Nouza
  • Petr Cerva
  • Radek SafarikEmail author
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 10930)

Abstract

We present methods and procedures designed for cost-efficient adaptation of an existing speech recognition system to Polish. The system (originally built for Czech language) is adapted using common texts and speech recordings accessible from Polish web-pages. The most critical part, an acoustic model (AM) for Polish, is built in several steps, which include: (a) an initial bootstrapping phase that utilizes existing Czech AM, (b) a lightly-supervised iterative scheme for automatic collection and annotation of Polish speech data, and finally (c) acquisition of a large amount of broadcast data in an unsupervised way. The developed system has been evaluated in the task of automatic content monitoring of major Polish TV and Radio stations. Its transcription accuracy (measured on a set of 4 complete TV news shows with total duration of 105 min) is 79,2%. For clean studio speech, its accuracy gets over 92%.

Keywords

Speech recognition of polish Broadcast monitoring Acoustic model training Cross-lingual adaptation 

Notes

Acknowledgment

This work was supported by the Technology Agency of the Czech Republic (project MultiLinMedia, no. TA04010199) and by the Student Grant Scheme at the Technical University of Liberec.

References

  1. 1.
    Chaloupka, J.: Digits to words converter for slavic languages in systems of automatic speech recognition. In: Karpov, A., Potapova, R., Mporas, I. (eds.) SPECOM 2017. LNCS (LNAI), vol. 10458, pp. 312–321. Springer, Cham (2017).  https://doi.org/10.1007/978-3-319-66429-3_30CrossRefGoogle Scholar
  2. 2.
    Demenko, G., Wypych, M., Baranowska, E.: Implementation of grapheme-to-phoneme rules and extended SAMPA alphabet in Polish text-to-speech synthesis. Speech Lang. Technol. 7(17), 79–97 (2003)Google Scholar
  3. 3.
    Demenko, G., Grocholewski, S., Klessa, K., Ogorkiewicz, J., Wagner, A., Lange, M., Sledzinski, D., Cylwik, N.: JURISDIC: polish speech database for taking dictation of legal texts. In: Proceedings of LREC, pp. 1280–1287 (2008)Google Scholar
  4. 4.
    Demenko, G., et al.: Development of large vocabulary continuous speech recognition for polish. Acta Phys. Pol. A 1(121), A-86 (2012)Google Scholar
  5. 5.
    Koržinek, D., Brocki, L.: Grammar based automatic speech recognition system for the Polish language. In: Jabłoński, R., Turkowski, M., Szewczyk, R. (eds.) Recent Advances in Mechatronics, pp. 87–91. Springer, Heidelberg (2007).  https://doi.org/10.1007/978-3-540-73956-2_18
  6. 6.
    Lööf, J., Gollan, C., Ney, H.: Cross-language bootstrapping for unsupervised acoustic model training: rapid development of a Polish speech recognition system. In: Proceedings of Interspeech, pp. 88–91 (2009)Google Scholar
  7. 7.
    Marasek, K.: Large vocabulary continuous speech recognition system for Polish. Arch. Acoust. 28(4), 119–126 (2003)Google Scholar
  8. 8.
    Nouza, J., Boháč, M.: Using TTS for fast prototyping of cross-lingual ASR applications. In: Esposito, A., Vinciarelli, A., Vicsi, K., Pelachaud, C., Nijholt, A. (eds.) Analysis of Verbal and Nonverbal Communication and Enactment. The Processing Issues. LNCS, vol. 6800, pp. 154–162. Springer, Heidelberg (2011).  https://doi.org/10.1007/978-3-642-25775-9_15CrossRefGoogle Scholar
  9. 9.
    Nouza, J., Cerva, P., Kucharova, M.: Cost-efficient development of acoustic models for speech recognition of related languages. Radioengineering 22(3), 866–873 (2013)Google Scholar
  10. 10.
    Nouza, J., et al.: Speech-to-text technology to transcribe and disclose 100,000 + hours of bilingual documents from historical czech and czechoslovak radio archive. In: Proceedings of Interspeech, pp. 964–968 (2014)Google Scholar
  11. 11.
    Nouza, J., Safarik, R., Cerva, P.: ASR for south slavic languages developed in almost automated way. In: Proceedings of Interspeech, pp. 3868–3872 (2016)Google Scholar
  12. 12.
    Pawlaczyk, L., Bosky, P.: Skrybot–a System for Automatic Speech Recognition of Polish Language. Man-Machine Interactions, pp. 381–387. Springer, Heidelberg (2009)zbMATHGoogle Scholar
  13. 13.
    Schultz, T.: GlobalPhone: a multilingual speech and text database developed at karlsruhe university. In: Proceedings of ICSLP, pp. 345–348 (2002)Google Scholar
  14. 14.
    Seps, L., Malek, J., Cerva, P., Nouza, J.: Investigation of deep neural networks for robust recognition of nonlinearly distorted speech. In: Proceedings of Interspeech, pp. 363–367 (2014)Google Scholar
  15. 15.
    Vu, N.T., et al.: Rapid bootstrapping of five eastern European languages using the rapid language adaptation toolkit. In: Proceedings of Interspeech, pp. 865–868 (2010)Google Scholar
  16. 16.
    Vu, N.T., Kraus, F., Schultz, T.: Multilingual A-stabil: a new confidence score for multilingual unsupervised training. In: Proceedings of Spoken Language Technology Workshop (SLT), pp. 183–188. IEEE (2010)Google Scholar
  17. 17.
    Ziółko, M., et al.: Automatic speech recognition system dedicated for Polish. In: Proceedings of Interspeech, pp. 3315–3315 (2011)Google Scholar

Copyright information

© Springer International Publishing AG, part of Springer Nature 2018

Authors and Affiliations

  1. 1.Institute of Information Technology and ElectronicsTechnical University of LiberecLiberecCzech Republic

Personalised recommendations