IPA-CHILDES & G2P+: Feature-Rich Resources for Cross-Lingual Phonology and Phonemic Language Modeling
This provides resources for cross-lingual phonological research and phonemic language modeling, addressing gaps in multilingual coverage and spontaneous speech datasets, but it is incremental as it builds on existing tools and databases.
The paper tackled the problem of inconsistent phonemic representations in grapheme-to-phoneme conversion by introducing G2P+, a tool that uses established phonemic inventories, and IPA CHILDES, a phonemic dataset of child-directed speech across 31 languages. The result showed that training phoneme language models on this dataset allowed learning of major class and place features cross-lingually from distributional properties.
In this paper, we introduce two resources: (i) G2P+, a tool for converting orthographic datasets to a consistent phonemic representation; and (ii) IPA CHILDES, a phonemic dataset of child-centered speech across 31 languages. Prior tools for grapheme-to-phoneme conversion result in phonemic vocabularies that are inconsistent with established phonemic inventories, an issue which G2P+ addresses by leveraging the inventories in the Phoible database. Using this tool, we augment CHILDES with phonemic transcriptions to produce IPA CHILDES. This new resource fills several gaps in existing phonemic datasets, which often lack multilingual coverage, spontaneous speech, and a focus on child-directed language. We demonstrate the utility of this dataset for phonological research by training phoneme language models on 11 languages and probing them for distinctive features, finding that the distributional properties of phonemes are sufficient to learn major class and place features cross-lingually.