CHILDES corpora are comparable corpora made up from transcripts of child language. Most of these transcripts record spontaneous conversational interactions. Often the speakers involved are young monolingual children conversing with their parents or siblings. Corpora also comprise transcripts of bilingual children, older school-aged children, adult second-language learners, children with various types of language disabilities and aphasics who are trying to recover from language loss.
These corpora belong to the child language project within the TalkBank system that was created for sharing and studying conversational interactions. Current CHILDES corpora in Sketch Engine include 24 languages.
CHILDES TalkBank web page is available at http://childes.talkbank.org/
Detailed information about each corpus within CHILDES collection can be found at http://childes.talkbank.org/access/