Doctoral Dissertation Research: Interviewer Voice Characteristics and Data Quality

博士论文研究：采访者声音特征与数据质量

• 批准号: 1356985
• 负责人: Kristen Olson
• 金额: $ 1.6万
• 依托单位: University of Nebraska-Lincoln
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-06-15 至 2016-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1356985&HistoricalAwards=false
• 关键词: Doctoral; Dissertation; Research; Interviewer; Voice

Telephone interviews frequently contain socially desirable, socially undesirable, and complex questions that tend to produce problems for respondents. A speaker may be especially likely to change vocal patterns for these types of questions. These changes may either directly affect data quality or indirectly affect data quality through the listener's perception of the voice. As telephone surveys continue to be the primary mode of data collection for many large national studies, it is important to understand how interviewer voices affect data quality. This study will examine whether interviewer voice characteristics affect data quality in socially desirable, undesirable, and complex survey questions. Interviewer voices are the primary means of communication to respondents in telephone interviews. If voice characteristics of interviewers affect data quality, those overseeing interviews may be able to select or train interviewers to modify some of their vocal characteristics with the goal of maximizing data quality. Moreover, results from this research will be useful for selecting interviewers based on voice characteristics for audio computer-assisted self-interviewing (ACASI), telephone audio-CASI (T-ACASI), and interactive voice response (IVR) systems with the goal of minimizing measurement error. As a Doctoral Dissertation Research Improvement award, support is provided to enable a promising student to establish a strong, independent research career.Specifically, this study has three objectives. The project will evaluate whether a telephone survey interviewer's objective voice characteristics including speech rate, pitch, intonation, and disfluency are associated with a listener's subjective perception of these voice characteristics and their assessment of five subjective interviewer traits (credibility, confidence, reliability, trustworthiness, and easiness to understand). The project also will examine whether objective voice characteristics of telephone survey interviewers affect data quality in socially desirable, socially undesirable, and complex questions. Finally, the project will investigate whether subjective perceptions of an interviewer's voice mediate the relationship between objective voice characteristics and data quality. The study will objectively measure interviewer's voice characteristics by using the Praat computer software program and will use raters to subjectively evaluate voice characteristics (e.g. pitch and speaking rate) and interviewer traits (e.g. credibility, confidence) on seven-point scales. Hierarchical logistic regression models will be used to examine the association between the objective and subjective voice characteristics and data quality. Measures of data quality include item nonresponse for all questions, rounding answers (e.g. 5, 10) for complex and neutral questions, and the directional hypothesis of more/less is better for socially undesirable/desirable questions.

电话采访经常包含社会上可取的，社会上不可取的，复杂的问题，往往会产生问题的受访者。 说话者可能特别有可能为这些类型的问题改变声音模式。 这些变化可能直接影响数据质量，也可能通过收听者对语音的感知间接影响数据质量。 由于电话调查仍然是许多大型国家研究的主要数据收集方式，因此了解采访者的声音如何影响数据质量非常重要。 本研究将探讨是否采访者的声音特征影响数据质量在社会上可取的，不可取的，和复杂的调查问题。 在电话采访中，采访者的声音是与受访者沟通的主要方式。 如果访谈者的声音特征影响数据质量，那么那些监督访谈的人可以选择或训练访谈者修改他们的一些声音特征，以最大限度地提高数据质量。 此外，本研究的结果将是有用的，以最大限度地减少测量误差为目标的音频计算机辅助自我采访（ACASI），电话音频CASI（T-ACASI）和交互式语音应答（IVR）系统的语音特征的基础上选择采访者。 作为博士论文研究改进奖，提供支持，使有前途的学生建立一个强大的，独立的研究生涯。具体来说，这项研究有三个目标。 该项目将评估电话调查采访者的客观语音特征，包括语速，音高，语调和不流利是否与听众对这些语音特征的主观感知以及他们对五个主观采访者特征（可信度，信心，可靠性，可信度和易于理解）的评估相关。 该项目还将研究电话调查采访者的客观声音特征是否会影响社会期望、社会不期望和复杂问题的数据质量。 最后，该项目将调查是否采访者的声音的主观看法调解客观的声音特征和数据质量之间的关系。 这项研究将使用Praat计算机软件程序客观地测量面试官的声音特征，并使用评分员主观地评估声音特征（例如音高和语速）和面试官特质（例如可信度，信心）。 分层逻辑回归模型将用于检查客观和主观语音特征与数据质量之间的关联。 数据质量的测量包括所有问题的项目无应答，复杂和中性问题的四舍五入答案（例如5，10），以及多/少的方向性假设更适合社会不良/可取的问题。