Speech is a fundamental means of human communication that can convey numerous meanings and intentions through even a single syllable. Speech sound disorders refer to conditions in which difficulties in speech sound production reduce the efficiency of ...
Speech is a fundamental means of human communication that can convey numerous meanings and intentions through even a single syllable. Speech sound disorders refer to conditions in which difficulties in speech sound production reduce the efficiency of verbal communication. Speech sound disorders, which show high prevalence among preschool children, require careful and meticulous approaches as they coincide with a period of rapid cognitive, linguistic, and emotional development. Children at this stage show considerable individual differences in language development rate and phonological awareness levels, necessitating a whole language approach tailored to each child's developmental stage and linguistic-cognitive level. Developmental children may experience unstable articulation production when they allocate considerable cognitive resources to higher-level language processing such as meaning construction and sentence structuring, requiring an environment that reduces cognitive load through repetitive story structures and allows focus on articulation.
This study developed a customized storybook-based articulation intervention program using generative AI that considers children's target phonemes, stimulability contexts, and preferences, and examined the effectiveness of an integrated intervention combining clinic-based treatment and home-based practice.
Four children aged 3-5 years with speech sound disorders who showed reduced intelligibility in connected speech and spontaneous speech compared to word-level accuracy were selected, and a multiple baseline design across subjects was applied.
The storybook-based articulation intervention program selected target phonemes through comprehensive analysis of each child's spontaneous speech, contextual tests, and stimulability assessments, then constructed stories appropriate to the child's reading developmental level and created visual materials suitable for the child's age and preferences using generative AI (ChatGPT, Gemini) and Canva. Target phonemes were combined with seven vowels and placed within story sentences to enable practice in various phonological environments.
The intervention consisted of 1-2 clinic sessions per week and home-based repeated reading. After each session, the researcher reviewed intervention videos with parents to explain error patterns and intervention principles and guided home assignments. Parents transmitted task videos after completing assignments at home following the researcher's feedback, and the researcher provided feedback within 24 hours.
Results showed that all four children improved in articulation accuracy and speech intelligibility in both trained words and generalization sentences, with effects maintained stably during the maintenance phase.
The four children exhibited different characteristics including fast speech rate, bilingual environment, lack of self-monitoring, and insufficient articulatory motor precision. The customized storybook intervention demonstrated significant improvement in articulation across all children despite their diverse characteristics, confirming the effectiveness of individualized intervention.
Repeated book reading showed effectiveness in multiple aspects. The researcher could detect subtle changes in children and respond immediately when new interventions were needed, while parents experienced increased participation motivation as the study progressed by directly observing children's changes through video recordings. Additionally, the researcher could directly monitor parents' home instruction methods through task videos and provide feedback, thereby enhancing the quality of home connection. These results can be interpreted as the effects of an integrated intervention approach that organically combined sensitive monitoring of children's speech changes, immediate customized material production using generative AI, systematic home connection, and continuous monitoring.
Particularly, this study identified differences between phoneme-level accuracy and utterance-level intelligibility. One child achieved the highest scores in consonant accuracy and word-level accuracy, yet showed relatively low sentence-level accuracy (the proportion of sentences produced entirely without errors). This demonstrates that phoneme-level evaluation alone is insufficient for fully assessing intervention effectiveness, and integrated evaluation at word and sentence levels is necessary.
This study holds significance in presenting concrete data on differences between phoneme accuracy and speech intelligibility, proposing sentence-level accuracy criteria, and empirically demonstrating the necessity of individualized intervention and the importance of home-clinic collaborative structures.