Publications

Abstract

- * indicates joint first authorship - Abstract: A range of demographic variables influence how much speech young children hear. However, because studies have used vastly different sampling methods, quantitative comparison of interlocking demographic effects has been nearly impossible, across or within studies. We harnessed a unique collection of existing naturalistic, day-long recordings from 61 homes across four North American cities to examine language input as a function of age, gender, and maternal education. We analyzed adult speech heard by 3- to 20-month-olds who wore audio recorders for an entire day. We annotated speaker gender and speech register (child-directed or adult-directed) for 10,861 utterances from female and male adults in these recordings. Examining age, gender, and maternal education collectively in this ecologically-valid dataset, we find several key results. First, the speaker gender imbalance in the input is striking: children heard 2--3x more speech from females than males. Second, children in higher-maternal-education homes heard more child-directed speech than those in lower-maternal education homes. Finally, our analyses revealed a previously unreported effect: the proportion of child-directed speech in the input increases with age, due to a decrease in adult-directed speech with age. This large-scale analysis is an important step forward in collectively examining demographic variables that influence early development, made possible by pooled, comparable, day-long recordings of children's language environments. The audio recordings, annotations, and annotation software are readily available for re-use and re-analysis by other researchers.

Abstract

Recent years have seen rapid technological development of devices that can record communicative behavior as participants go about daily life. This paper is intended as an end-to-end methodological guidebook for potential users of these technologies, including researchers who want to study children’s or adults’ communicative behavior in everyday contexts. We explain how long-format speech environment (LFSE) recordings provide a unique view on language use and how they can be used to complement other measures at the individual and group level. We aim to help potential users of these technologies make informed decisions regarding research design, hardware, software, and archiving. We also provide information regarding ethics and implementation, issues that are difficult to navigate for those new to this technology, and on which little or no resources are available. This guidebook offers a concise summary of information for new users and points to sources of more detailed information for more advanced users. Links to discussion groups and community-augmented databases are also provided to help readers stay up-to-date on the latest developments.

Abstract

Odor naming is enhanced in communities where communication about odors is a central part of daily life (e.g., wine experts, flavorists, and some hunter‐gatherer groups). In this study, we investigated how expert knowledge and daily experience affect the ability to name odors in a group of experts that has not previously been investigated in this context—Iranian herbalists; also called attars—as well as cooks and laypeople. We assessed naming accuracy and consistency for 16 herb and spice odors, collected judgments of odor perception, and evaluated participants' odor meta‐awareness. Participants' responses were overall more consistent and accurate for more frequent and familiar odors. Moreover, attars were more accurate than both cooks and laypeople at naming odors, although cooks did not perform significantly better than laypeople. Attars' perceptual ratings of odors and their overall odor meta‐awareness suggest they are also more attuned to odors than the other two groups. To conclude, Iranian attars—but not cooks—are better odor namers than laypeople. They also have greater meta‐awareness and differential perceptual responses to odors. These findings further highlight the critical role that expertise and type of experience have on olfactory functions.

Abstract

Automatic word count estimation (WCE) from audio recordings can be used to quantify the amount of verbal communication in a recording environment. One key application of WCE is to measure language input heard by infants and toddlers in their natural environments, as captured by daylong recordings from microphones worn by the infants. Although WCE is nearly trivial for high-quality signals in high-resource languages, daylong recordings are substantially more challenging due to the unconstrained acoustic environments and the presence of near- and far-field speech. Moreover, many use cases of interest involve languages for which reliable ASR systems or even well-defined lexicons are not available. A good WCE system should also perform similarly for low- and high-resource languages in order to enable unbiased comparisons across different cultures and environments. Unfortunately, the current state-of-the-art solution, the LENA system, is based on proprietary software and has only been optimized for American English, limiting its applicability. In this paper, we build on existing work on WCE and present the steps we have taken towards a freely available system for WCE that can be adapted to different languages or dialects with a limited amount of orthographically transcribed speech data. Our system is based on language-independent syllabification of speech, followed by a language-dependent mapping from syllable counts (and a number of other acoustic features) to the corresponding word count estimates. We evaluate our system on samples from daylong infant recordings from six different corpora consisting of several languages and socioeconomic environments, all manually annotated with the same protocol to allow direct comparison. We compare a number of alternative techniques for the two key components in our system: speech activity detection and automatic syllabification of speech. As a result, we show that our system can reach relatively consistent WCE accuracy across multiple corpora and languages (with some limitations). In addition, the system outperforms LENA on three of the four corpora consisting of different varieties of English. We also demonstrate how an automatic neural network-based syllabifier, when trained on multiple languages, generalizes well to novel languages beyond the training data, outperforming two previously proposed unsupervised syllabifiers as a feature extractor for WCE.

Abstract

Interoperable annotation formats are fundamental to the utility, expansion, and sustainability of collective data repositories.In language development research, shared annotation schemes have been critical to facilitating the transition from raw acoustic data to searchable, structured corpora. Current schemes typically require comprehensive and manual annotation of utterance boundaries and orthographic speech content, with an additional, optional range of tags of interest. These schemes have been enormously successful for datasets on the scale of dozens of recording hours but are untenable for long-format recording corpora, which routinely contain hundreds to thousands of audio hours. Long-format corpora would benefit greatly from (semi-)automated analyses, both on the earliest steps of annotation—voice activity detection, utterance segmentation, and speaker diarization—as well as later steps—e.g., classification-based codes such as child-vs-adult-directed speech, and speech recognition to produce phonetic/orthographic representations. We present an annotation workflow specifically designed for long-format corpora which can be tailored by individual researchers and which interfaces with the current dominant scheme for short-format recordings. The workflow allows semi-automated annotation and analyses at higher linguistic levels. We give one example of how the workflow has been successfully implemented in a large cross-database project.

Abstract

Children begin developing turn-taking skills in infancy but take several years to fluidly integrate their growing knowledge of language into their turn-taking behavior. In two eye-tracking experiments, we measured children’s anticipatory gaze to upcoming responders while controlling linguistic cues to turn structure. In Experiment 1, we showed English and non-English conversations to English-speaking adults and children. In Experiment 2, we phonetically controlled lexicosyntactic and prosodic cues in English-only speech. Children spontaneously made anticipatory gaze switches by age two and continued improving through age six. In both experiments, children and adults made more anticipatory switches after hearing questions. Consistent with prior findings on adult turn prediction, prosodic information alone did not increase children’s anticipatory gaze shifts. But, unlike prior work with adults, lexical information alone was not sucient either—children’s performance was best overall with lexicosyntax and prosody together. Our findings support an account in which turn tracking and turn prediction emerge in infancy and then gradually become integrated with children’s online linguistic processing.

Abstract

Child-directed speech is argued to facilitate language development, and is found cross-linguistically and cross-culturally to varying degrees. However, previous research has generally focused on short samples of child-caregiver interaction, often in the lab or with experimenters present. We test the generalizability of this phenomenon with an initial descriptive analysis of the speech heard by young children in a large, unique collection of naturalistic, daylong home recordings. Trained annotators coded automatically-detected adult speech 'utterances' from 61 homes across 4 North American cities, gathered from children (age 2-24 months) wearing audio recorders during a typical day. Coders marked the speaker gender (male/female) and intended addressee (child/adult), yielding 10,886 addressee and gender tags from 2,523 minutes of audio (cf. HB-CHAAC Interspeech ComParE challenge; Schuller et al., in press). Automated speaker-diarization (LENA) incorrectly gender-tagged 30% of male adult utterances, compared to manually-coded consensus. Furthermore, we find effects of SES and gender on child-directed and overall speech, increasing child-directed speech with child age, and interactions of speaker gender, child gender, and child age: female caretakers increased their child-directed speech more with age than male caretakers did, but only for male infants. Implications for language acquisition and existing classification algorithms are discussed.

Abstract

The INTERSPEECH 2017 Computational Paralinguistics Challenge addresses three different problems for the first time in research competition under well-defined conditions: In the Addressee sub-challenge, it has to be determined whether speech produced by an adult is directed towards another adult or towards a child; in the Cold sub-challenge, speech under cold has to be told apart from ‘healthy’ speech; and in the Snoring subchallenge, four different types of snoring have to be classified. In this paper, we describe these sub-challenges, their conditions, and the baseline feature extraction and classifiers, which include data-learnt feature representations by end-to-end learning with convolutional and recurrent neural networks, and bag-of-audiowords for the first time in the challenge series

Abstract

Young children answer questions with longer delays than adults do, and they don't reach typical adult response times until several years later. We hypothesized that this prolonged pattern of delay in children's timing results from competing demands: to give an answer, children must understand a question while simultaneously planning and initiating their response. Even as children get older and more efficient in this process, the demands on them increase because their verbal responses become more complex. We analyzed conversational question-answer sequences between caregivers and their children from ages 1;8 to 3;5, finding that children (1) initiate simple answers more quickly than complex ones, (2) initiate simple answers quickly from an early age, and (3) initiate complex answers more quickly as they grow older. Our results suggest that children aim to respond quickly from the start, improving on earlier-acquired answer types while they begin to practice later-acquired, slower ones.

Abstract

The core use of language is in face-to-face conversation. This is characterized by rapid turn-taking. This turn-taking poses a number central puzzles for the psychology of language.

Consider, for example, that in large corpora the gap between turns is on the order of 100 to 300 ms, but the latencies involved in language production require minimally between 600ms (for a single word) or 1500 ms (for as simple sentence). This implies that participants in conversation are predicting the ends of the incoming turn and preparing in advance. But how is this done? What aspects of this prediction are done when? What happens when the prediction is wrong? What stops participants coming in too early? If the system is running on prediction, why is there consistently a mode of 100 to 300 ms in response time?

The timing puzzle raises further puzzles: it seems that comprehension must run parallel with the preparation for production, but it has been presumed that there are strict cognitive limitations on more than one central process running at a time. How is this bottleneck overcome? Far from being 'easy' as some psychologists have suggested, conversation may be one of the most demanding cognitive tasks in our everyday lives. Further questions naturally arise: how do children learn to master this demanding task, and what is the developmental trajectory in this domain?

Research shows that aspects of turn-taking such as its timing are remarkably stable across languages and cultures, but the word order of languages varies enormously. How then does prediction of the incoming turn work when the verb (often the informational nugget in a clause) is at the end? Conversely, how can production work fast enough in languages that have the verb at the beginning, thereby requiring early planning of the whole clause? What happens when one changes modality, as in sign languages -- with the loss of channel constraints is turn-taking much freer? And what about face-to-face communication amongst hearing individuals -- do gestures, gaze, and other body behaviors facilitate turn-taking? One can also ask the phylogenetic question: how did such a system evolve? There seem to be parallels (analogies) in duetting bird species, and in a variety of monkey species, but there is little evidence of anything like this among the great apes.

All this constitutes a neglected set of problems at the heart of the psychology of language and of the language sciences. This research topic welcomes contributions from right across the board, for example from psycholinguists, developmental psychologists, students of dialogue and conversation analysis, linguists interested in the use of language, phoneticians, corpus analysts and comparative ethologists or psychologists. We welcome contributions of all sorts, for example original research papers, opinion pieces, and reviews of work in subfields that may not be fully understood in other subfields.