A Phonemic Corpus of Polish Child-Directed Speech

摘要

Recent advances in modeling early language acquisition are due not only to the development of machine-learning techniques, but also to the increasing availability of data on child language and child-adult interaction. In the absence of recordings of child-directed speech, or when models explicitly require such a representation for training data, phonemic transcriptions are commonly used as input data. We present a novel (and to our knowledge, the first) phonemic corpus of Polish child-directed speech. It is derived from the Weist corpus of Polish, freely available from the seminal CHILDES database. For the sake of reproducibility, and to exemplify the typical trade-off between ecological validity and sample size, we report all preprocessing operations and transcription guidelines. Contributed linguistic resources include updated CHAT-formatted transcripts with phonemic transcriptions in a novel phonology tier, as well as by-product data, such as a phonemic lexicon of Polish. All resources are distributed under the LGPL-LR license.