Out-of-vocabulary word recognition using a hierarchical language model based on multiple Markov models

Hirofumi Yamamoto, Hiroaki Kokubo, Genichiro Kikui, Yoshihiko Ogawa, Yoshinori Sagisaka

    Research output: Contribution to journalArticle

    1 Citation (Scopus)

    Abstract

    In this paper we propose a language model to solve the issue of task-dependent out-of-vocabulary words in speech recognition. Language model adaptation is a standard method to enable the application of a language model to a new task; however, this approach is not able to deal with the issue of out-of-vocabulary proper names that appear in a task-dependent fashion. In this paper we attempt to solve this issue using a hierarchical language model. In the hierarchical model we use two independent Markov models to constrain the transition probabilities and phonetic sequence emission probabilities of out-of-vocabulary words. In this way we express the emission probabilities of out-of-vocabulary words in the form of a double Markov model that combines both sets of probabilities. We have conducted speech recognition experiments using Japanese dialogue data in the appointments domain. The results show that for sentences containing one or more out-of-vocabulary words, this approach gives a word accuracy rate of 86.7% compared to word accuracy rate of 78.2% when no strategy for out-of-vocabulary words is employed. This corresponds to an elimination of 34.4% of the baseline errors and confirms the effectiveness of the approach.

    Original languageEnglish
    Pages (from-to)55-64
    Number of pages10
    JournalElectronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi)
    Volume88
    Issue number12
    DOIs
    Publication statusPublished - 2005 Dec

    Fingerprint

    speech recognition
    Speech recognition
    phonetics
    sentences
    Speech analysis
    transition probabilities
    elimination
    Experiments

    Keywords

    • Hierarchical language model
    • Markov model
    • N-gram
    • Out-of-vocabulary words

    ASJC Scopus subject areas

    • Electrical and Electronic Engineering

    Cite this

    Out-of-vocabulary word recognition using a hierarchical language model based on multiple Markov models. / Yamamoto, Hirofumi; Kokubo, Hiroaki; Kikui, Genichiro; Ogawa, Yoshihiko; Sagisaka, Yoshinori.

    In: Electronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi), Vol. 88, No. 12, 12.2005, p. 55-64.

    Research output: Contribution to journalArticle

    @article{7b2dd0b931194d8da429a9bad1272069,
    title = "Out-of-vocabulary word recognition using a hierarchical language model based on multiple Markov models",
    abstract = "In this paper we propose a language model to solve the issue of task-dependent out-of-vocabulary words in speech recognition. Language model adaptation is a standard method to enable the application of a language model to a new task; however, this approach is not able to deal with the issue of out-of-vocabulary proper names that appear in a task-dependent fashion. In this paper we attempt to solve this issue using a hierarchical language model. In the hierarchical model we use two independent Markov models to constrain the transition probabilities and phonetic sequence emission probabilities of out-of-vocabulary words. In this way we express the emission probabilities of out-of-vocabulary words in the form of a double Markov model that combines both sets of probabilities. We have conducted speech recognition experiments using Japanese dialogue data in the appointments domain. The results show that for sentences containing one or more out-of-vocabulary words, this approach gives a word accuracy rate of 86.7{\%} compared to word accuracy rate of 78.2{\%} when no strategy for out-of-vocabulary words is employed. This corresponds to an elimination of 34.4{\%} of the baseline errors and confirms the effectiveness of the approach.",
    keywords = "Hierarchical language model, Markov model, N-gram, Out-of-vocabulary words",
    author = "Hirofumi Yamamoto and Hiroaki Kokubo and Genichiro Kikui and Yoshihiko Ogawa and Yoshinori Sagisaka",
    year = "2005",
    month = "12",
    doi = "10.1002/ecjb.20238",
    language = "English",
    volume = "88",
    pages = "55--64",
    journal = "Electronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi)",
    issn = "8756-663X",
    publisher = "Scripta Technica",
    number = "12",

    }

    TY - JOUR

    T1 - Out-of-vocabulary word recognition using a hierarchical language model based on multiple Markov models

    AU - Yamamoto, Hirofumi

    AU - Kokubo, Hiroaki

    AU - Kikui, Genichiro

    AU - Ogawa, Yoshihiko

    AU - Sagisaka, Yoshinori

    PY - 2005/12

    Y1 - 2005/12

    N2 - In this paper we propose a language model to solve the issue of task-dependent out-of-vocabulary words in speech recognition. Language model adaptation is a standard method to enable the application of a language model to a new task; however, this approach is not able to deal with the issue of out-of-vocabulary proper names that appear in a task-dependent fashion. In this paper we attempt to solve this issue using a hierarchical language model. In the hierarchical model we use two independent Markov models to constrain the transition probabilities and phonetic sequence emission probabilities of out-of-vocabulary words. In this way we express the emission probabilities of out-of-vocabulary words in the form of a double Markov model that combines both sets of probabilities. We have conducted speech recognition experiments using Japanese dialogue data in the appointments domain. The results show that for sentences containing one or more out-of-vocabulary words, this approach gives a word accuracy rate of 86.7% compared to word accuracy rate of 78.2% when no strategy for out-of-vocabulary words is employed. This corresponds to an elimination of 34.4% of the baseline errors and confirms the effectiveness of the approach.

    AB - In this paper we propose a language model to solve the issue of task-dependent out-of-vocabulary words in speech recognition. Language model adaptation is a standard method to enable the application of a language model to a new task; however, this approach is not able to deal with the issue of out-of-vocabulary proper names that appear in a task-dependent fashion. In this paper we attempt to solve this issue using a hierarchical language model. In the hierarchical model we use two independent Markov models to constrain the transition probabilities and phonetic sequence emission probabilities of out-of-vocabulary words. In this way we express the emission probabilities of out-of-vocabulary words in the form of a double Markov model that combines both sets of probabilities. We have conducted speech recognition experiments using Japanese dialogue data in the appointments domain. The results show that for sentences containing one or more out-of-vocabulary words, this approach gives a word accuracy rate of 86.7% compared to word accuracy rate of 78.2% when no strategy for out-of-vocabulary words is employed. This corresponds to an elimination of 34.4% of the baseline errors and confirms the effectiveness of the approach.

    KW - Hierarchical language model

    KW - Markov model

    KW - N-gram

    KW - Out-of-vocabulary words

    UR - http://www.scopus.com/inward/record.url?scp=29744454150&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=29744454150&partnerID=8YFLogxK

    U2 - 10.1002/ecjb.20238

    DO - 10.1002/ecjb.20238

    M3 - Article

    AN - SCOPUS:29744454150

    VL - 88

    SP - 55

    EP - 64

    JO - Electronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi)

    JF - Electronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi)

    SN - 8756-663X

    IS - 12

    ER -