Quantifying talker variability in North-American infant's daily input

Words sound slightly different each time they are said, both by the same talker and across talkers. Rather than hurting learning, lab studies suggest that talker variability helps infants learn similar sounding words. However, very little is known about how much variability infants hear within a single talker or across talkers in naturalistic input. Here, we quantified these types of talker variability for highly frequent words spoken to 44 infants, from naturalistic recordings sampled longitudinally over a year of life (from 6-17 months). We used non-contrastive acoustic measurements (e.g. mean pitch, duration, harmonics-to-noise ratio) and holistic measures of sound similarity (normalized acoustic distance) to quantify acoustic variability. We find three key results. First, pitch-based variability was generally lower for infants' top talker than across their other talkers, but overall acoustic distance is higher for tokens from the top talker vesus the others. Second, the amount of acoustic variability infants heard was not simply redundant with other properties of the input such as number of talkers or tokens, or proportion of speech from particular sources (e.g. women, children, electronics). Finally, we find that patterns of acoustic variability heard in naturalistic input were similar to those found with in-lab stimuli that facilitated word learning. This large-scale quantification of talker variability in infants' everyday input sets the stage for linking naturally-occurring variability ‘in the wild’ to early word learning.

Albayzin 2018 Evaluation: The IberSpeech-RTVE Challenge on Speech Technologies for Spanish Broadcast Media

The IberSpeech-RTVE Challenge presented at IberSpeech 2018 is a new Albayzin evaluation series supported by the Spanish Thematic Network on Speech Technologies (Red Temática en Tecnologías del Habla (RTTH)). That series was focused on speech-to-text transcription, speaker diarization, and multimodal diarization of television programs. For this purpose, the Corporacion Radio Television Española (RTVE), the main public service broadcaster in Spain, and the RTVE Chair at the University of Zaragoza made more than 500 h of broadcast content and subtitles available for scientists. The dataset included about 20 programs of different kinds and topics produced and broadcast by RTVE between 2015 and 2018. The programs presented different challenges from the point of view of speech technologies such as: the diversity of Spanish accents, overlapping speech, spontaneous speech, acoustic variability, background noise, or specific vocabulary. This paper describes the database and the evaluation process and summarizes the results obtained.

Social context influences the acoustic properties of laughter

Laughter serves different functions depending on the social context. Sometimes laughter feels good for producers and listeners, serving as a social reward. Sometimes laughter smooths over social tension, promoting affiliation. And sometimes laughter reprimands or teases the target, establishing dominance. The present study examines whether the acoustic properties of spontaneous laughter reflect the context in which it occurs. Pairs of participants (complete recordings N = 141) watched and discussed humorous videos associated with the social tasks of reward, affiliation, and dominance. The acoustic profiles of the 3,370 laughs extracted from their conversations depended on the social context. Affiliation laughter was shorter, quieter, and muffled, while dominance laughter was more unpleasant (e.g., brighter, less voiced, noisier). The combined acoustic variables discriminated at greater than chance accuracy between the social contexts. This research accounts for some of the acoustic variability of laughter and highlights the functional flexibility of nonverbal signals.