Technology Category
Natural Language Processing: SpeechMachine Learning: Large Multimodal Models (LMMs)Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
IPA-CHILDES&G2P+: Feature-Rich Resources for Cross-Lingual Phonology and Phonemic Language Modeling
🤖 AI Summary
Existing grapheme-to-phoneme (G2P) tools produce phonemic representations lacking cross-linguistic consistency, while mainstream phonemic datasets suffer from insufficient multilingual coverage, absence of natural speech, and lack of child language data. Method: We introduce G2P+, a standardized G2P tool integrating the PHOIBLE phoneme inventory, and the IPA CHILDES dataset—the first large-scale, IPA-fine-grained, multilingual phonemic resource derived from spontaneous child speech in 31 languages. We further employ phoneme-level language modeling and feature probing to assess cross-linguistic learnability of core phonological features. Contribution/Results: Our work establishes the first standardized, multilingual G2P framework grounded in PHOIBLE; delivers the largest publicly available IPA-annotated child speech phonemic corpus; and demonstrates, across 11 languages, that articulatory place and phonation type—key phonological features—are robustly learnable cross-linguistically. These advances significantly enhance interpretability and cross-lingual generalization in phonological modeling.
Technology Category
Application Category
📝 Abstract
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.
Problem
Research questions and friction points this paper is trying to address.
Inconsistent phonemic vocabularies in grapheme-to-phoneme conversion tools
Lack of multilingual phonemic datasets with child-directed speech
Need for cross-lingual phonemic language modeling and phonological research
Innovation
Methods, ideas, or system contributions that make the work stand out.
G2P+ converts orthography to consistent phonemic representation
IPA CHILDES provides multilingual child-directed phonemic data
Leverages Phoible for accurate phonemic inventories
🔎 Similar Papers
No similar papers found.
