This article is based on a part of the first author's Master thesis. Objectives:The purpose of this study was to investigate age-related differences according to frequency of words and use of phonological rules in the Korean word recognition process between younger and elderly groups through Event-Related Potentials (ERP) analysis. Methods: A total of 35 participants participated in this experiment. They were asked to judge whether the pronunciation of visual words was identical to the corresponding auditory words. Behavioral data and electrophysiological data were collected. Results: Behavioral results revealed that the elderly group showed significantly lower accuracy and longer response time than the younger group. ERP analysis showed that there was no significant difference between the two groups in the 150-300 ms range. However, for the younger group, the N400 component was observed in the 300-500 ms range regardless of word frequency when the phonological rule was not applied and was more apparent under the low word frequency condition. The topographic patterns of the grand average ERP waveforms for the elderly group showed that the N400 component appeared only under high word frequency when the phonological rule was not applied. However, when the phonological rule was applied, the N400 component was observed only under the condition of low word frequency. Conclusion: Differences in the pattern of applying phonological rules associated with word frequency were found between groups. The current study indicates that word frequency and aging may affect the ability to apply phonological rules. The result of age-related differences in ERP analysis reflects a decline in the simultaneous neural processing of phonological and semantic information.

Objectives: A systematic review of the literature was undertaken (1) to investigate research trends on how artificial intelligence is being used for assessment and diagnosis in the field of communication disorders and (2) to suggest consideration and a directions for the effective use of artificial intelligence in clinical settings. Methods: A total of 328 articles published in foreign journals between January 2016 and August 2021 were searched using 6 databases and a manual search, and 18 articles were finally selected according to PICO strategy (Population, Intervention, Comparison, Outcome) inclusion and exclusion criteria. Four authors determined the report selection and data extraction. They also independently analyzed the quality of the selected papers using QUADAS-II (Quality Assessment of Diagnostic Accuracy Studies-II). Results: Firstly, the selected studies had a generally low risk of bias. Secondly, the major subjects of studies were children with communication disorders. Thirdly, most of the studies included in the analysis were experimental studies to verify the effectiveness of using artificial intelligence. Lastly, the extracted features for assessment and diagnosis were biased against acoustic features at the levels of phoneme and word in speaking tasks. The performance of artificial intelligence in the selected studies differed according to the research purpose and evaluation metrics. Conclusion: This study suggests that in order for artificial intelligence to be used in the assessment and diagnosis system, it is essential to acquire clinically reliable and high-quality big data on the characteristics of speech and language of people with communication disorders.