Noise in Speech-to-Text Voice: Analysis of Errors and Feasibility of Phonetic Similarity for Their Correction

Hanna SUOMINEN, Gabriela Ferraro

    Research output: A Conference proceeding or a Chapter in BookConference contribution

    Abstract

    In Australian healthcare, failures in information flow cause over one-tenth of preventable adverse events and are tangible in clinical handover. Regardless of a good verbal handover, anything from two-thirds to all of this information is lost after 3– 5 shifts if notes are taken by hand or not taken. Speech to text (SST) and information extraction (IE) have been proposed for taking the notes and filling in a handover form with extrapolated evaluations from related studies promising over 90 per cent correctness for both STT and IE. However, this cascading evokes a fruitful methodological challenge: the severe implications that errors may have in clinical decision-making call for superiority in STT; the correctness percentage measured in a peaceful laboratory is decreased to 77 by noise in clinical practise; and the STT errors multiply when cascaded with IE. We provide an analysis of STT errors and dis- cuss the feasibility of phonetic similarity for their correction in this paper. Our data consists of one hundred simulated handover records in Australian English with STT recognising 73 per cent of the 7 ; 277 words (1 h 8 min 5 s) correctly. In text relevant to the form, 836 unique error types are present. The most common errors include inserting and , in , are , arm , is , a , the , or am ( 5 n 94 ), deleting is ( n = 17 ), and substituting and , obs are , 2 , he with in, also, to, or and she (7≤n≤11), respectively. Eighteen per cent of word substitutions sound exactly the same as the correct word and 26 per cent have a similarity percentage above 75. This encourages using phonetic similarity to improve STT.
    Original languageEnglish
    Title of host publicationProceedings of Australasian Language Technology Association Workshop
    PublisherAssociation for Computational Linguistics
    Pages34-42
    Number of pages9
    Publication statusPublished - 2013

    Fingerprint

    Phonetics
    Information Storage and Retrieval
    Noise
    Patient Handoff
    Arm
    Hand
    Delivery of Health Care

    Cite this

    SUOMINEN, H., & Ferraro, G. (2013). Noise in Speech-to-Text Voice: Analysis of Errors and Feasibility of Phonetic Similarity for Their Correction. In Proceedings of Australasian Language Technology Association Workshop (pp. 34-42). [UL13-1006] Association for Computational Linguistics.
    SUOMINEN, Hanna ; Ferraro, Gabriela. / Noise in Speech-to-Text Voice: Analysis of Errors and Feasibility of Phonetic Similarity for Their Correction. Proceedings of Australasian Language Technology Association Workshop. Association for Computational Linguistics, 2013. pp. 34-42
    @inproceedings{da0452d7dbe746cc88181a0e7ae14e14,
    title = "Noise in Speech-to-Text Voice: Analysis of Errors and Feasibility of Phonetic Similarity for Their Correction",
    abstract = "In Australian healthcare, failures in information flow cause over one-tenth of preventable adverse events and are tangible in clinical handover. Regardless of a good verbal handover, anything from two-thirds to all of this information is lost after 3– 5 shifts if notes are taken by hand or not taken. Speech to text (SST) and information extraction (IE) have been proposed for taking the notes and filling in a handover form with extrapolated evaluations from related studies promising over 90 per cent correctness for both STT and IE. However, this cascading evokes a fruitful methodological challenge: the severe implications that errors may have in clinical decision-making call for superiority in STT; the correctness percentage measured in a peaceful laboratory is decreased to 77 by noise in clinical practise; and the STT errors multiply when cascaded with IE. We provide an analysis of STT errors and dis- cuss the feasibility of phonetic similarity for their correction in this paper. Our data consists of one hundred simulated handover records in Australian English with STT recognising 73 per cent of the 7 ; 277 words (1 h 8 min 5 s) correctly. In text relevant to the form, 836 unique error types are present. The most common errors include inserting and , in , are , arm , is , a , the , or am ( 5 n 94 ), deleting is ( n = 17 ), and substituting and , obs are , 2 , he with in, also, to, or and she (7≤n≤11), respectively. Eighteen per cent of word substitutions sound exactly the same as the correct word and 26 per cent have a similarity percentage above 75. This encourages using phonetic similarity to improve STT.",
    keywords = "Speech-to-text",
    author = "Hanna SUOMINEN and Gabriela Ferraro",
    year = "2013",
    language = "English",
    pages = "34--42",
    booktitle = "Proceedings of Australasian Language Technology Association Workshop",
    publisher = "Association for Computational Linguistics",

    }

    SUOMINEN, H & Ferraro, G 2013, Noise in Speech-to-Text Voice: Analysis of Errors and Feasibility of Phonetic Similarity for Their Correction. in Proceedings of Australasian Language Technology Association Workshop., UL13-1006, Association for Computational Linguistics, pp. 34-42.

    Noise in Speech-to-Text Voice: Analysis of Errors and Feasibility of Phonetic Similarity for Their Correction. / SUOMINEN, Hanna; Ferraro, Gabriela.

    Proceedings of Australasian Language Technology Association Workshop. Association for Computational Linguistics, 2013. p. 34-42 UL13-1006.

    Research output: A Conference proceeding or a Chapter in BookConference contribution

    TY - GEN

    T1 - Noise in Speech-to-Text Voice: Analysis of Errors and Feasibility of Phonetic Similarity for Their Correction

    AU - SUOMINEN, Hanna

    AU - Ferraro, Gabriela

    PY - 2013

    Y1 - 2013

    N2 - In Australian healthcare, failures in information flow cause over one-tenth of preventable adverse events and are tangible in clinical handover. Regardless of a good verbal handover, anything from two-thirds to all of this information is lost after 3– 5 shifts if notes are taken by hand or not taken. Speech to text (SST) and information extraction (IE) have been proposed for taking the notes and filling in a handover form with extrapolated evaluations from related studies promising over 90 per cent correctness for both STT and IE. However, this cascading evokes a fruitful methodological challenge: the severe implications that errors may have in clinical decision-making call for superiority in STT; the correctness percentage measured in a peaceful laboratory is decreased to 77 by noise in clinical practise; and the STT errors multiply when cascaded with IE. We provide an analysis of STT errors and dis- cuss the feasibility of phonetic similarity for their correction in this paper. Our data consists of one hundred simulated handover records in Australian English with STT recognising 73 per cent of the 7 ; 277 words (1 h 8 min 5 s) correctly. In text relevant to the form, 836 unique error types are present. The most common errors include inserting and , in , are , arm , is , a , the , or am ( 5 n 94 ), deleting is ( n = 17 ), and substituting and , obs are , 2 , he with in, also, to, or and she (7≤n≤11), respectively. Eighteen per cent of word substitutions sound exactly the same as the correct word and 26 per cent have a similarity percentage above 75. This encourages using phonetic similarity to improve STT.

    AB - In Australian healthcare, failures in information flow cause over one-tenth of preventable adverse events and are tangible in clinical handover. Regardless of a good verbal handover, anything from two-thirds to all of this information is lost after 3– 5 shifts if notes are taken by hand or not taken. Speech to text (SST) and information extraction (IE) have been proposed for taking the notes and filling in a handover form with extrapolated evaluations from related studies promising over 90 per cent correctness for both STT and IE. However, this cascading evokes a fruitful methodological challenge: the severe implications that errors may have in clinical decision-making call for superiority in STT; the correctness percentage measured in a peaceful laboratory is decreased to 77 by noise in clinical practise; and the STT errors multiply when cascaded with IE. We provide an analysis of STT errors and dis- cuss the feasibility of phonetic similarity for their correction in this paper. Our data consists of one hundred simulated handover records in Australian English with STT recognising 73 per cent of the 7 ; 277 words (1 h 8 min 5 s) correctly. In text relevant to the form, 836 unique error types are present. The most common errors include inserting and , in , are , arm , is , a , the , or am ( 5 n 94 ), deleting is ( n = 17 ), and substituting and , obs are , 2 , he with in, also, to, or and she (7≤n≤11), respectively. Eighteen per cent of word substitutions sound exactly the same as the correct word and 26 per cent have a similarity percentage above 75. This encourages using phonetic similarity to improve STT.

    KW - Speech-to-text

    M3 - Conference contribution

    SP - 34

    EP - 42

    BT - Proceedings of Australasian Language Technology Association Workshop

    PB - Association for Computational Linguistics

    ER -

    SUOMINEN H, Ferraro G. Noise in Speech-to-Text Voice: Analysis of Errors and Feasibility of Phonetic Similarity for Their Correction. In Proceedings of Australasian Language Technology Association Workshop. Association for Computational Linguistics. 2013. p. 34-42. UL13-1006