A Longitudinal Evaluation of Tablet-Based Child Speech Therapy with Apraxia World

A word practice game that presents words in preset order without recommendations or predictions based on real-time learner performance likely represents a non-AI clinical technology, whereas one that automatically uses speech analysis to monitor performance and/or adapt recommendations likely represents clinical AI.

“A word practice game that presents words in this preset order, without recommendations or predictions based on learner performance in real time, likely represents a non-AI clinical technology, whereas a word practice game that automatically uses speech analysis to monitor learner performance and/or adapt word practice recommendations based on learner performance would likely represent a clinical”

Say Bananas! can be considered an AI clinician due to its ability to recommend speech practice trials and use automatic speech analysis to provide feedback.

“Similarly, Say Bananas! can be considered an AI clinician due to its ability to recommend speech practice trials during use of the program and use automatic speech analysis to provide the learner with feedback.”

Two systematic reviews examining automatic speech analysis for computerized speech sound learning systems available before December 2016 and February 2022 highlighted an overall lack of high-quality randomized controlled trials examining effectiveness/efficacy of the systems.

“For example, two systematic reviews that specifically examine the use of automatic speech analysis for computerized speech sound learning systems available before December 2016 (McKechnie et al., 2018) and February 2022 (Deka et al., 2024) have highlighted an overall lack of high-quality randomized controlled trials examining effectiveness/efficacy of the systems (cf. the high-quality clinical tri”

Speaker-specific methods (e.g., template matching as in Hair et al., 2021) could be used to provide evidence that test set performance does not differ based on demographic factors for the individual speakers recruited for the study.

“Alternately, speaker-specific methods (e.g., template matching as in Hair et al., 2021) could be used to provide evidence that test set performance does not differ based on demographic factors for the individual speakers recruited for the study.”

SayBananas! app involves tasks for the child to listen and produce selected target words, and provides feedback on target word accuracy as a correct/incorrect decision.

“This app involves tasks for the child user to both listen and produce selected target words, and is provided with feedback on target word accuracy, as a correct/incorrect decision (Hair et al. 2021; McLeod et al. 2023).”

Hair et al. (2021) studied 10 children and observed improved target sound accuracy and high engagement over a 4 plus 4 week intervention.

“The researchers observed improved target sound accuracy in the children and high levels of engagement throughout the extended intervention.”

The observed dose frequency ranging between four and seven sessions per week in the present study is comparable to the frequency of four sessions per week in Hair et al. (2021) and McLeod et al. (2023), and higher than the frequency of 2-3 sessions per week in Peterson et al. (2022).

“Regarding dose frequency, on the other hand, the observed frequency ranging between four and seven sessions per week in the present study, is comparable to the frequency of four sessions per week in Hair et al. (2021) and McLeod et al. (2023), and higher than the frequency of 2–3 sessions per week in Peterson et al. (2022).”

The results that intervention with Pop2TalkNordic was not successful for the participating children runs counter to the positive results reported in previous studies of app-based speech intervention for children with SSD.

“The results that intervention with Pop2TalkNordic was not successful for the participating children runs counter to the positive results reported in previous studies of app‐based speech intervention for children with SSD (Hair et al. 2021; McLeod et al. 2023; Peterson et al. 2022).”

Compared to the observed dose of 76 production efforts in Hair et al. (2021) and 200 in Peterson et al. (2022), the present study's dose is markedly lower, whereas it is more similar to the 45 observed in McLeod et al. (2023).

“Compared to the observed dose of 76 production efforts in Hair et al. (2021), and 200 in Peterson et al. (2022), it is markedly lower, whereas it is more similar to the 45 observed in McLeod et al. (2023).”

Pop2TalkNordic offers more auditory exposure to target words than opportunities to production compared to both staRt and SayBananas!.

“It should be noted, however, that there is a critical difference between these speech‐training tools, where Pop2TalkNordic offers more auditory exposure to target words than opportunities to production compared to both staRt and SayBananas!.”

The current study employed a shorter and less supervised form of intervention as compared to previous investigations.

“In terms of intervention delivery, the current study employed a shorter and less supervised form of intervention as compared to previous investigations.”