The Hebrew CHILDES corpus: transcription and morphological analysis

Area of Science:

• Linguistics
• Computational Linguistics
• Corpus Linguistics

Background:

• The CHILDES database provides transcribed spoken language corpora for numerous languages.
• Spoken Hebrew data requires specialized transcription and analysis methods.
• Existing resources for spoken Hebrew within CHILDES were limited.

Purpose of the Study:

• To develop a comprehensive, morphologically-annotated corpus of transcribed spoken Hebrew.
• To create a dedicated transcription scheme sensitive to Hebrew phonology and orthography.
• To build robust computational tools for analyzing spoken Hebrew corpora.

Main Methods:

• Compilation of a transcribed spoken Hebrew corpus from child-adult interactions.
• Development of a novel transcription system for spoken Hebrew.
• Creation of a morphological analyzer and disambiguation module tailored for the corpus.
• Evaluation of the analyzer's performance on the corpus and a new dataset.

Main Results:

• A high-quality, morphologically-annotated corpus of spoken Hebrew integrated into the CHILDES database.
• A transcription scheme accurately reflecting Hebrew phonology and orthography.
• A morphological analyzer with high coverage and accuracy for the Hebrew corpus.
• A disambiguation module enhancing the precision of morphological analysis.

Conclusions:

• The developed corpus and tools provide a valuable resource for spoken Hebrew research.
• The methodology can be extended to create similar annotated corpora for other languages.
• This work significantly advances the study of spoken Hebrew within a cross-linguistic framework.