Dialogue Act Analysis of Facilitator-Children Multilingual Communication

Multilingual communication among children is beneficial for cultivating an attitude that respects cultural diversity at an early age. Improvements in translation accuracy by neural machine translation can enhance children’s mutual understanding. However, a significant challenge remains in involving children who speak low-resource languages (with limited language resources) in multilingual communication due to poor translation quality in those languages, resulting in a low number of utterances from the children. Meanwhile, facilitators play an important role in encouraging children’s dialogue. Therefore, we have analyzed the facilitator’s dialogue acts in a children’s intercultural workshop to clarify which acts can promote their utterances. Specifically, we annotated all facilitator utterances with a dialogue act corresponding to each instance of children’s language use in the actual multilingual dialogue log. Subsequently, we tallied the number of responses to each dialogue act. With this data, we compared the number of children’s responses for each dialogue act across languages, analyzed the number of children’s responses in each language for each dialogue act, and examined the patterns of change in dialogue acts for language pairs where translation was involved. The analysis finds that although no act has significant differences in the number of responses across different languages, the facilitator’s utterance act changes caused by machine translation can result in a difference in the number of children’s responses.