Modern subword-based models for automatic speech recognition

Loading...
Thumbnail Image
Journal Title
Journal ISSN
Volume Title
School of Electrical Engineering | Doctoral thesis (article-based) | Defence date: 2019-06-17
Date
2019
Major/Subject
Mcode
Degree programme
Language
en
Pages
62 + app. 136
Series
Aalto University publication series DOCTORAL DISSERTATIONS, 97/2019
Abstract
In today's society, speech recognition systems have reached a mass audience, especially in the field of personal assistants such as Amazon Alexa or Google Home. Yet, this does not mean that speech recognition has been solved. On the contrary, for many domains, tasks, and languages such systems do not exist. Subword-based automatic speech recognition has been studied in the past for many reasons, often to overcome limitations on the size of the vocabulary. Specifically for agglutinative languages, where new words can be created on the fly, handling these limitations is possible using a subword-based automatic speech recognition (ASR) system. Though, over time subword-based systems lost a bit of popularity as system resources increased and word-based models with large vocabularies became possible. Still, subword-based models in modern ASR systems can predict words that have never been seen before and better use the available language modeling resources. Furthermore, subword models have smaller vocabularies, which makes neural network language models (NNLMs) easier to train and use.  Hence, in this thesis, we study subword models for ASR and make two major contributions. First, this thesis reintroduces subword-based modeling in a modern framework based on weighted finite-state transducers and describe the necessary tools for making a sound and effective system. It does this through careful modification of the lexicon FST part of a WFST-based recognizer. Secondly, extensive experiments using are done using subwords, with different types of language models including n-gram models and NNLMs. These experiments are performed on six different languages setting the new best-published result for any of these datasets. Overall, we show that subword-based models can outperform word-based models in terms of ASR performance for many different types of languages. This thesis also details design choices needed when building modern subword ASR systems, including the choices of the segmentation algorithm, vocabulary size and subword marking style. In addition, it includes techniques to combine speech recognition models trained on different units through system combination. Lastly, it evaluates the use of the smallest possible subword unit; characters and shows that these models can be smaller and yet be competitive to word-based models.
Description
Supervising professor
Kurimo, Mikko, Assoc. Prof., Aalto University, Department of Signal Processing and Acoustics, Finland
Thesis advisor
Virpioja, Sami, Dr., Aalto University, Department of Signal Processing and Acoustics, Finland
Keywords
automatic speech recognition, language modeling, subword models
Other note
Parts
  • [Publication 1]: Sami Virpioja, Peter Smit, Stig-Arne Grönroos, Mikko Kurimo. Morfessor 2.0: Python Implementation and Extensions for Morfessor Baseline.
    Full text in Acris/Aaltodoc: http://urn.fi/URN:ISBN:978-952-60-5501-5
  • [Publication 2]: Peter Smit, Sami Virpioja, Mikko Kurimo. Improved Subword Modeling for WFST-Based Speech Recognition. In Annual Conference of the International Speech Communication Association (INTERSPEECH), Stockholm, pages 2551–2555 , August 2017.
    Full text in Acris/Aaltodoc: http://urn.fi/URN:NBN:fi:aalto-201710157202
    DOI: 10.21437/Interspeech.2017-103 View at publisher
  • [Publication 3]: Peter Smit, Juho Leinonen, Kristiina Jokinen, Mikko Kurimo. Automatic Speech Recognition for Northern Sámi with comparison to other Uralic Languages. In Proceedings of the Second International Workshop on Computational Linguistics for Uralic Languages, Szeged, pages 80–91, January 2016.
    Full text in Acris/Aaltodoc: http://urn.fi/URN:NBN:fi:aalto-201701191109
  • [Publication 4]: Juho Leinonen, Peter Smit, Sami Virpioja, Mikko Kurimo. New Baseline in Automatic Speech Recognition for Northern Sámi. In Fourth International Workshop on Computational Linguistics for Uralic Languages, Helsinki, pages 89–99, January 2018.
    Full text in Acris/Aaltodoc: http://urn.fi/URN:NBN:fi:aalto-201802091229
    DOI: 10.18653/v1/W18-0208 View at publisher
  • [Publication 5]: Peter Smit, Siva Reddy Gangireddy, Seppo Enarvi, Sami Virpioja, Mikko Kurimo. Character-based units for Unlimited Vocabulary Continuous Speech Recognition. In IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), Okinawa, pages 149–156, December 2017.
    Full text in Acris/Aaltodoc: http://urn.fi/URN:NBN:fi:aalto-201802091465
    DOI: 10.1109/ASRU.2017.8268929 View at publisher
  • [Publication 6]: Peter Smit, Siva Reddy Gangireddy, Seppo Enarvi, Sami Virpioja, Mikko Kurimo. Aalto system for the 2017 Arabic multi-genre broadcast challenge. In IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), Okinawa, pages 338–345, December 2017.
    Full text in Acris/Aaltodoc: http://urn.fi/URN:NBN:fi:aalto-201802091512
    DOI: 10.1109/ASRU.2017.8268955 View at publisher
  • [Publication 7]: Seppo Enarvi, Peter Smit, Sami Virpioja, Mikko Kurimo. Automatic Speech Recognition with Very Large Conversational Finnish and Estonian Vocabularies. IEEE/ACM Transactions on Audio, Speech, and Language Processing, volume 25, issue 11, pages 2085–2097, November 2017.
    Full text in Acris/Aaltodoc: http://urn.fi/URN:NBN:fi:aalto-201710157079
    DOI: 10.1109/TASLP.2017.2743344 View at publisher
  • [Publication 8]: Peter Smit, Sami Virpioja, Mikko Kurimo. Advances in Subword-based HMM-DNN Speech Recognition Across Languages. Submitted to Language Resources and Evaluation, 29 November 2018.
Citation