Disfluencies in alcoholized speech

Disfluencies in alcoholized speech
复制标题

酗酒后言语不流畅

DOI：
--
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
F. Schiel
F. Schiel
中科院分区：
--
文献类型：
--
作者：
S. Barfüßer;F. Schiel

文献摘要

被引文献

相似文献

s，IAFPA 2010，Trier大学语音系，德国，酒精中毒语音的不流利性S。Barfüßer，F. Schiel Institut für Phonetik und Sprachverarbeitung Ludwig Maximilians Universität，München，德国{bine|本研究的目的是澄清在酒精与清醒言语的影响下，言语中产生的不流利的数量是否存在显着差异，从而可以用作醉酒的指标。更具体地说，我们调查了9个语言和准语言的事件，可以概括为更广泛的术语不流利，可以被视为从正常的，流畅的讲话可察觉的偏差。除事件计数外，还检查了填充停顿（犹豫）的持续时间。方法尽管许多研究都是关于酒精中毒的语言（参见（Chin & Pisoni，1997）和（Künzel et al.，1992年）的概述），存在的研究与大量的对象，两种性别或与血液酒精浓度的精确测量。在这项研究中，我们使用了酒精语言语料库（ALC）1中的128名酒精中毒受试者（66名男性和62名女性）的语言，血液酒精浓度范围为0.5‰至1.68‰。以下事件作为酒精中毒（二进制），性别和言语类型（阅读，自发和引发命令和控制）的函数进行计数：犹豫的持续时间，犹豫的次数，长停顿，短停顿（<1秒），单词的错误发音，重复，修复，声音的异常延长和单词中断。事件比例采用以说话人为随机因素的混合模型进行统计分析，而时长测量采用经典的RM-ANOVA进行检验。结果酒精中毒对8个事件有显著影响，如下表所示（“total”表示所有不流利计数器的总和;“↑”数字因酒精中毒而显著增加;“-”无显著变化）：total hes.dur hes. num.长。p。短。p。mispr。重复修理很长时间。内部↑ ↑此外，事实证明，演讲的类型也经常影响结果。在某些语音类型中，某些事件的发生率很小，这是一个问题（例如，命令和控制语音中的重复）。在演讲中，我们将介绍计数事件的例子，更详细的统计结果，并讨论自动酒精检测系统的影响。参考文献Barfüßer，S.，Gilg，Th.，Heinrich，Chr.，Schiel，F.（2008年）。酒精语言语料库。2008年LREC会议论文集Chin，S.B.和Pisoni，D. B.（1997年）。酒精和演讲中国科学出版社.昆策尔J.，布劳恩，A.，埃斯霍尔特大学（1992年）。酒精对Sprache和Stimme的影响。Heidelberg：Kriminalistik Verlag，见http://www.bas.uni-muenchen.de/Bas
s, IAFPA 2010, Department of Phonetics, Trier University, Germany Disfluencies in alcoholized speech S. Barfüßer, F. Schiel Institut für Phonetik und Sprachverarbeitung Ludwig­Maximilians­Universität, München, Germany {bine|schiel}@phonetik.uni-muenchen.de Introduction and goals The aim of this study is to clarify whether there is a significant difference in the number of disfluencies produced in speech under the influence of alcohol vs. sober speech, and thus can be used as an indicator for intoxication. More specifically we investigated nine linguistic and para-linguistic events which can be summarized under the broader term disfluencies and can be regarded as perceptible deviations from normal, fluent speech. Aside from the event counts the duration of filled pauses (hesitations) was examined as well. Method Although a number of studies have been dealing with alcoholized speech (see (Chin & Pisoni, 1997) and (Künzel et al., 1992) for an overview) there exist few studies with a large number of subjects of both genders or with a precise measurement of the blood alcohol concentration. For this study we used the speech of 128 alcoholized subjects (66 male and 62 female) from the Alcohol Language Corpus (ALC)1 with a blood alcohol concentration ranging from 0.5‰ to 1.68‰. The following events were counted as a function of alcoholisation (binary), gender and speech type (read, spontaneous and elicited command&control): Durations of hesitations, number of hesitations, long pauses, short pauses (<1sec), mispronunciation of words, repetitions, repairs, unusual elongations of sounds and word interruptions. Event proportion were statistically analysed using mixed models with the speaker as random factor, while the duration measurement was tested using classical RM-ANOVA. Results Alcoholisation had a significant effect on eight of the events as shown in the following table (‘total’ denotes the sum of all disfluency counters; ‘↑’ numbers increase significantly for alcoholisation; ‘-’ no significant change): total hes.dur hes.num. long.p. short.p. mispr. repeat repair elong. interr. ↑ ↑ ↑ ↑ ­ ↑ ↓ ­ ↑ ↑ The gender of the subject only had an influence on the number of hesitations. Further it turned out that the type of speech also affected the results frequently. The small occurrence rate of some events in certain speech types turned out to be a problem (e.g. repetitions in command&control speech). In the talk we will present examples of the counted events, more detailed statistical results and discuss the implications for automatic alcohol detection systems. References Barfüßer, S., Gilg, Th., Heinrich, Chr., Schiel, F. (2008). ALC Alcohol Language Corpus. Proc. of LREC 2008 Chin, S.B. and Pisoni, D.B. (1997). Alcohol and Speech. Academic Press. Künzel, H. J., Braun, A., Eysholdt, U. (1992). Einfluss von Alkohol auf Sprache und Stimme. Heidelberg: Kriminalistik Verlag see http://www.bas.uni-muenchen.de/Bas