Year: 2026
Type: conference-paper
Abstract: Child-directed Speech (CDS) has been shown to better support language learning when used as training data for computational models.Synthetically generated input attempts to replicate this advantage by recreating targeted linguistic properties of CDS.Recently, the use of questions in CDS has been suggested as a linguistic property that may entail an effective discourse structure for model training.However, previous work has shown inconsistent improvement over baseline using questions in the training data.In this study, we extend Poh et al. ( 2025) by revising the prompts, and introducing sampling controls that align both generation and sampling with properties of CDS.We show that prompt language substantially changes whether synthetic questions match CDS on surface properties such as MLU and question type.Despite marked improvements over baseline, enhanced CDS-likeness does not translate into consistent downstream gains.Overall, our results show that the role of questions in training data is a topic worth examining further.
Authors: Whitney Poh, Michael Tombolini, Libby Barak
Institution: Montclair State University
Language: English
Subfield: Developmental and Educational Psychology
Field: Psychology
Domain: Social Sciences
SDG: Quality education
Open Access status: gold
Funder: Montclair State University
OpenAlex ID: W7166834311