NineOneOne: Exploratory Research on Recognizing Non-English Speech for Emergency Triage in Disaster Response

这项探索性研究试图证明，自动语音识别(ASR)和机器翻译(MT)的新方法能够很好地自动、实时地进行紧急分流。在任何重大灾难中，如飓风，在短时间内都会有大量的9-1-1紧急呼叫；其中许多是英语以外的语言。即使在最好的情况下，在美国的许多9-1-1中心，对于非英语呼叫的分流，也存在着时间上的短缺。对于紧急分流，不需要完全识别和翻译；只需要识别紧急情况的类型和相关的细节，如位置。然而，需要新的先进的ASR方法来处理非英语9-1-1呼叫的困难特征，这些呼叫在噪声环境中包含情绪痛苦的语音，并且可能包括混合了其他语言的英语短语。对于ASR，探索性系统使用由发音特征增强的多语言声学模型，以及结合n元语法语言模型的多语言语法来识别语音。至于机器翻译，探索性系统只将ASR输出归类为“DomainActions”，即与分诊相关的类别，如“Request-Ambuance”或“Report-Flowing”。该探索性系统适用于西班牙语，但所使用的方法适用于任何语言，并可移植到其他任务域。演示用于9-1-1呼叫中心的语音翻译技术的可行性，应使后续项目能够在重大灾难期间更好地服务于所有非英语国家的人。我们希望将得到的转录和注释语音语料库供其他研究人员使用。

This exploratory research seeks to demonstrate that new approaches toAutomatic Speech Recognition (ASR) and Machine Translation (MT) permitautomatic, real-time translation well enough for emergency triage. Inany major disaster, such as a hurricane, there will be a huge numberof 9-1-1 emergency calls in a short period of time; many will be inlanguages other than English. Even in the best of times, there is achronic shortage of translation for triage of non-English calls atmany U.S. 9-1-1 centers.For emergency triage, full recognition and translation areunnecessary; only the type of emergency and relevant details, such aslocation, need be recognized. Nevertheless, new advanced ASR methodsare needed to handle the difficult characteristics of non-English9-1-1 calls, which contain emotionally-distressed speech in noisyenvironments and may include English phrases mixed in with the otherlanguage. For ASR, the exploratory system uses multilingual acousticmodels enhanced by articulatory features, and multilingual grammarscombined with n-gram language models, to recognize speech. As for MT,the exploratory system only classifies the ASR outputs into ``DomainActions'', categories relevant to triage such as ``Request-Ambulance''or ``Report-Flooding''. The exploratory system works with Spanish,but the approaches used are applicable to any language, andtransferable to other task domains.Demonstrating the feasibility of speech translation technology forserving 9-1-1 call centers should enable follow-on projects to betterserve all non-English speakers during major disasters. We expect tomake the resulting transcribed and annotated speech corpus availablefor other researchers.