Evaluating the feasibility of an automated classifier for target-child-directed speech from LENA recordings

Author:

Bang Janet YORCID,Kachergis GeorgeORCID,Weisleder Adriana,Marchman Virginia A.ORCID

Abstract

Recent advances in recording technology provide unique opportunities to observe children’s everyday speech environments using daylong audio recordings. A number of studies have proposed that speech may support learning differently when speech is directed to a child (target child directed speech, tCDS) than when that speech is directed to others (ODS) (Shneidman & Goldin-Meadow, 2012; Weisleder & Fernald, 2013). To identify periods of tCDS and ODS, researchers typically rely on the time-consuming and laborious work of human listeners who consider numerous features when making judgments. Human listeners are also used to identify periods when children are sleeping or awake. In this paper, we detail our efforts to automate these processes. We analyzed over 1,000 hours of audio from daylong recordings of 153 English- and Spanish-speaking families in the U.S. with 17- to 28-month-old children that had been previously coded for periods of sleep, tCDS, and ODS. We first explored patterns of features that characterized periods of tCDS and ODS. Then, we evaluated two classifiers that were trained using automated measures generated from LENATM, including frequency (AWC, CTC, CVC) and duration (meaningful speech, distant speech, TV, noise, silence) measures. Results revealed high sensitivity and specificity in classifying periods of sleep, and moderate sensitivity and specificity in classifying periods of tCDS and ODS. Model-derived predictions from our tCDS/ODS classifier yielded similar patterns of correlations as previously-published findings, with variation in tCDS, but not ODS, positively linked to children’s later vocabularies (Weisleder & Fernald, 2013). This work offers promising tools for streamlining work with daylong recordings, thereby facilitating research that aims to better understand how children learn from their everyday speech environments.

Publisher

Center for Open Science

Cited by 2 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3