A purely monotonic approach to machine translation for similar languages

Ye Kyaw Thu, Andrew Finch, Eiichiro Sumita, Yoshinori Sagisaka

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    This paper investigates the effect of taking a strictly monotonic approach to machine translation for a restricted set of suitable language pairs. We studied the effect of decoding monotonically for a set of language pairs which has similar word order characteristics and found that for some language pairs - namely language pairs where both languages are in SOV order - there was almost no difference in machine translation quality. The results of this experiment motivated the extension of the monotonic approach into the alignment stage of the training. We used a Bayesian non-parametric aligner that has been shown to out-perform GIZA++ in combination with the grow-diag-final- and heuristic on transliteration data. Our results show that the monotonic aligner was able to match the performance of the GIZA++ baseline, and gains in translation performance were obtained by integrating both aligners into the systems.

    Original languageEnglish
    Title of host publicationProceedings - 2013 International Conference on Asian Language Processing, IALP 2013
    Pages107-110
    Number of pages4
    DOIs
    Publication statusPublished - 2013
    Event2013 International Conference on Asian Language Processing, IALP 2013 - Urumqi, Xinjiang
    Duration: 2013 Aug 172013 Aug 19

    Other

    Other2013 International Conference on Asian Language Processing, IALP 2013
    CityUrumqi, Xinjiang
    Period13/8/1713/8/19

    Fingerprint

    Decoding
    Experiments

    Keywords

    • bilingual alignment
    • machine translation
    • monotonic decoding

    ASJC Scopus subject areas

    • Software

    Cite this

    Thu, Y. K., Finch, A., Sumita, E., & Sagisaka, Y. (2013). A purely monotonic approach to machine translation for similar languages. In Proceedings - 2013 International Conference on Asian Language Processing, IALP 2013 (pp. 107-110). [6646015] https://doi.org/10.1109/IALP.2013.31

    A purely monotonic approach to machine translation for similar languages. / Thu, Ye Kyaw; Finch, Andrew; Sumita, Eiichiro; Sagisaka, Yoshinori.

    Proceedings - 2013 International Conference on Asian Language Processing, IALP 2013. 2013. p. 107-110 6646015.

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Thu, YK, Finch, A, Sumita, E & Sagisaka, Y 2013, A purely monotonic approach to machine translation for similar languages. in Proceedings - 2013 International Conference on Asian Language Processing, IALP 2013., 6646015, pp. 107-110, 2013 International Conference on Asian Language Processing, IALP 2013, Urumqi, Xinjiang, 13/8/17. https://doi.org/10.1109/IALP.2013.31
    Thu YK, Finch A, Sumita E, Sagisaka Y. A purely monotonic approach to machine translation for similar languages. In Proceedings - 2013 International Conference on Asian Language Processing, IALP 2013. 2013. p. 107-110. 6646015 https://doi.org/10.1109/IALP.2013.31
    Thu, Ye Kyaw ; Finch, Andrew ; Sumita, Eiichiro ; Sagisaka, Yoshinori. / A purely monotonic approach to machine translation for similar languages. Proceedings - 2013 International Conference on Asian Language Processing, IALP 2013. 2013. pp. 107-110
    @inproceedings{2eb297c1bb9b4f3a8ddad46a3cf88b20,
    title = "A purely monotonic approach to machine translation for similar languages",
    abstract = "This paper investigates the effect of taking a strictly monotonic approach to machine translation for a restricted set of suitable language pairs. We studied the effect of decoding monotonically for a set of language pairs which has similar word order characteristics and found that for some language pairs - namely language pairs where both languages are in SOV order - there was almost no difference in machine translation quality. The results of this experiment motivated the extension of the monotonic approach into the alignment stage of the training. We used a Bayesian non-parametric aligner that has been shown to out-perform GIZA++ in combination with the grow-diag-final- and heuristic on transliteration data. Our results show that the monotonic aligner was able to match the performance of the GIZA++ baseline, and gains in translation performance were obtained by integrating both aligners into the systems.",
    keywords = "bilingual alignment, machine translation, monotonic decoding",
    author = "Thu, {Ye Kyaw} and Andrew Finch and Eiichiro Sumita and Yoshinori Sagisaka",
    year = "2013",
    doi = "10.1109/IALP.2013.31",
    language = "English",
    isbn = "9780769550633",
    pages = "107--110",
    booktitle = "Proceedings - 2013 International Conference on Asian Language Processing, IALP 2013",

    }

    TY - GEN

    T1 - A purely monotonic approach to machine translation for similar languages

    AU - Thu, Ye Kyaw

    AU - Finch, Andrew

    AU - Sumita, Eiichiro

    AU - Sagisaka, Yoshinori

    PY - 2013

    Y1 - 2013

    N2 - This paper investigates the effect of taking a strictly monotonic approach to machine translation for a restricted set of suitable language pairs. We studied the effect of decoding monotonically for a set of language pairs which has similar word order characteristics and found that for some language pairs - namely language pairs where both languages are in SOV order - there was almost no difference in machine translation quality. The results of this experiment motivated the extension of the monotonic approach into the alignment stage of the training. We used a Bayesian non-parametric aligner that has been shown to out-perform GIZA++ in combination with the grow-diag-final- and heuristic on transliteration data. Our results show that the monotonic aligner was able to match the performance of the GIZA++ baseline, and gains in translation performance were obtained by integrating both aligners into the systems.

    AB - This paper investigates the effect of taking a strictly monotonic approach to machine translation for a restricted set of suitable language pairs. We studied the effect of decoding monotonically for a set of language pairs which has similar word order characteristics and found that for some language pairs - namely language pairs where both languages are in SOV order - there was almost no difference in machine translation quality. The results of this experiment motivated the extension of the monotonic approach into the alignment stage of the training. We used a Bayesian non-parametric aligner that has been shown to out-perform GIZA++ in combination with the grow-diag-final- and heuristic on transliteration data. Our results show that the monotonic aligner was able to match the performance of the GIZA++ baseline, and gains in translation performance were obtained by integrating both aligners into the systems.

    KW - bilingual alignment

    KW - machine translation

    KW - monotonic decoding

    UR - http://www.scopus.com/inward/record.url?scp=84891284230&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84891284230&partnerID=8YFLogxK

    U2 - 10.1109/IALP.2013.31

    DO - 10.1109/IALP.2013.31

    M3 - Conference contribution

    SN - 9780769550633

    SP - 107

    EP - 110

    BT - Proceedings - 2013 International Conference on Asian Language Processing, IALP 2013

    ER -