Describing the Spoken Language Skills of Typically Developing Afrikaans-Speaking Children Using Language Sample Analysis: A Pilot Study.

PURPOSE: Language sample analysis is widely regarded as the gold standard of language assessment. However, the uncertainty regarding the optimal length of sample and the limited availability of developmental language data for nonmainstream languages such as Afrikaans complicate reliable use of the method. The study aimed to provide guidelines on representative length of sample and concurrently provide a preliminary description of the spoken language skills of Afrikaans-speaking children. METHOD: The study involved 30 typically developing Afrikaans-speaking children aged between 3;6 and 9;6 (years;months). A descriptive research design was used to transcribe and analyze 1-hr interactions collected in natural environments of participants who were recruited using referral sampling. Video and audio recordings of the samples were transcribed using adapted Sampling Utterances and Grammatical Analysis Revised analysis procedures. RESULTS: Results indicated that mean length of utterance in words per minute, number of different words per minute, and total number of words per minute stabilized at 30 min and no significant differences were found between 30 min and longer time segments. Morphology results concur with existing developmental findings in Afrikaans. Lexical diversity results correlated with the findings of the lexical specificity and accuracy in the Prutting and Kirchner Pragmatic Protocol (Prutting & Kirchner, 1987). The developmental trajectories for pragmatic and phonological development were consistent with existing guidelines. CONCLUSIONS: The study concluded that a 30-min interaction segment provides a representative language sample for Afrikaans-speaking children who are between 3;6 and 9;6. It provides promising preliminary developmental data and clinical guidelines, confirming the potential of language sample analysis (LSA) as a reliable component of language assessment in Afrikaans.

Multidisciplinary Perspectives on Automatic Analysis of Children's Language Samples: Where Do We Go from Here?

BACKGROUND: Language sample analysis (LSA) is invaluable to describe and understand child language use and development for clinical purposes and research. Digital tools supporting LSA are available, but many of the LSA steps have not been automated. Nevertheless, programs that include automatic speech recognition (ASR), the first step of LSA, have already reached mainstream applicability. SUMMARY: To better understand the complexity, challenges, and future needs of automatic LSA from a technological perspective, including the tasks of transcribing, annotating, and analysing natural child language samples, this article takes on a multidisciplinary view. Requirements of a fully automated LSA process are characterized, features of existing LSA software tools compared, and prior work from the disciplines of information science and computational linguistics reviewed. KEY MESSAGES: Existing tools vary in their extent of automation provided across the process of LSA. Advances in machine learning for speech recognition and processing have potential to facilitate LSA, but the specifics of child speech and language as well as the lack of child data complicate software design. A transdisciplinary approach is recommended as feasible to support future software development for LSA.

NCHLT Auxiliary speech data for ASR technology development in South Africa.

The aim of the National Centre for Human Language Technology (NCHLT) project was to create speech and text resources that would enable Human Language Technology (HLT) development for the 11 official languages of South Africa. The speech data described in this paper was collected during the NCHLT project using a smartphone application. The official NCHLT Speech Corpus was released in 2014, but it did not include all recordings that were made during the data collection campaign. This paper describes the additional data that was recently released as auxiliary corpora [2]. The auxiliary data sets contain between 20 and 170 hours of speech data per language as well as the transcriptions associated with each utterance. In terms of the resources required for HLT development South Africa's official languages are all under-resourced. The data described in this paper contributes toward alleviating this situation, specifically for the development of speech technology.

Learning new words from an interactive electronic storybook intervention.

BACKGROUND:  Children who enter school with limited vocabulary knowledge are at risk for reading failure. This study investigated the efficacy of an interactive e-book, implemented as a mobile application, to facilitate vocabulary learning in Grade 1 isiXhosa-speaking children (n = 65). OBJECTIVE:  The purpose was to measure if an e-book intervention, specifically developed for use in the South African context, could facilitate the acquisition and retention of new words at different levels of lexical representation. METHOD:  A randomised pre-test and/or post-test between-subject design was used where an experimental group that received the e-book intervention was compared to a control group before the control group received a delayed intervention. Follow-up testing was performed to measure retention of the new vocabulary after eight weeks. Mixed-model repeated-measure Analysis of Variance (ANOVAs) were used to determine differences between the participants in the experimental and control groups. RESULTS:  The short-term e-book intervention not only facilitated fast-mapping of new words but enabled participants to develop more robust lexical representations of the newly acquired words. Follow-up assessment showed that they retained their newly acquired word knowledge. CONCLUSION:  Multimedia technology can be used to provide explicit and embedded vocabulary training to young children at risk for academic failure. These findings are particularly relevant for South African environments where there is limited parental support and lack of educational resources to promote vocabulary learning in young children.

Evaluation of formant-like features on an automatic vowel classification task.

Numerous attempts have been made to find low-dimensional, formant-related representations of speech signals that are suitable for automatic speech recognition. However, it is often not known how these features behave in comparison with true formants. The purpose of this study was to compare two sets of automatically extracted formant-like features, i.e., robust formants and HMM2 features, to hand-labeled formants. The robust formant features were derived by means of the split Levinson algorithm while the HMM2 features correspond to the frequency segmentation of speech signals obtained by two-dimensional hidden Markov models. Mel-frequency cepstral coefficients (MFCCs) were also included in the investigation as an example of state-of-the-art automatic speech recognition features. The feature sets were compared in terms of their performance on a vowel classification task. The speech data and hand-labeled formants that were used in this study are a subset of the American English vowels database presented in Hillenbrand et al. [J. Acoust. Soc. Am. 97, 3099-3111 (1995)]. Classification performance was measured on the original, clean data and in noisy acoustic conditions. When using clean data, the classification performance of the formant-like features compared very well to the performance of the hand-labeled formants in a gender-dependent experiment, but was inferior to the hand-labeled formants in a gender-independent experiment. The results that were obtained in noisy acoustic conditions indicated that the formant-like features used in this study are not inherently noise robust. For clean and noisy data as well as for the gender-dependent and gender-independent experiments the MFCCs achieved the same or superior results as the formant features, but at the price of a much higher feature dimensionality.