DOCOMO adapts AI dialogue to a preschool child’s vocabulary
NTT DOCOMO has developed an approach that limits an AI character’s vocabulary according to a child’s age in months. An Android app uses that dialogue to encourage routines such as toothbrushing. Earlier observations used manually selected words; household trials are planned to assess the automated system’s effectiveness and ease of use.
Artificial Intelligence··Morning
Age in months guides the words
NTT DOCOMO has developed dialogue technology that adjusts the words an AI character uses to a preschool child’s age. The approach announced on 27 October combines a language model with a vocabulary-development database. It is intended to support everyday routines, including toothbrushing, through conversation in an Android application under development.[1], [2]
The database, built by NTT’s Communication Science Laboratories, contains information on when children acquire particular words. Parents enter the child’s age in months. The system uses that age to avoid vocabulary considered difficult for the child’s developmental stage.[1]
The character responds to routines and past behavior
Prompts are personalized using the child’s characteristics, ordinary conversations and earlier behavioral outcomes. The character can speak at a configured time or for a selected routine, such as changing clothes or preparing for bed. Simple dialogue sits alongside an activity log and stickers collected when a routine is completed. DOCOMO’s stated aim is to encourage children to act voluntarily while easing caregivers’ repeated work.[1]
Household trials follow a limited manual test
In an earlier small test, children responded to synthesized speech and were observed beginning activities such as toothbrushing. The prompt words were chosen manually under restricted conditions. DOCOMO explicitly says those observations do not establish the effectiveness of the automated technology or app. The new process automates wording, speech synthesis, prompting and adjustment after an outcome. The company plans household trials to examine effectiveness and usability, and identifies better recognition of younger children’s speech as further work.[1]
