Database of word-level statistics for Mandarin Chinese (DoWLS-MAN)

Area of Science:

• Linguistics
• Computational Linguistics
• Psycholinguistics

Background:

• Mandarin Chinese phonology presents challenges in consistent syllable segmentation.
• Existing databases lack comprehensive, multi-schema phonological data for Mandarin.
• Understanding word-level statistics is crucial for speech processing models.

Purpose of the Study:

• Introduce the Database of Word-Level Statistics for Mandarin Chinese (DoWLS-MAN).
• Provide researchers with a flexible resource for Mandarin speech processing and phonological segmentation studies.
• Facilitate hypothesis generation regarding phonological processing in Chinese orthography.

Main Methods:

• Compiled a database of over 92,000 words, 1600 characters, and 4300 nonwords.
• Included phonological features based on 16 schematic syllable representations (with and without tone).
• Categorized statistics into variant (e.g., frequency, density) and invariant (e.g., tone, transcription) characteristics.

Main Results:

• DoWLS-MAN offers variant and invariant lexical statistics across multiple segmentation schemas.
• The database supports selecting stimuli aligned with existing speech processing models or for testing segmentation hypotheses.
• An exploratory analysis demonstrated the utility of multiple schemas for hypothesis generation in Chinese reading.

Conclusions:

• DoWLS-MAN provides a valuable, publicly accessible resource for Mandarin speech and language research.
• The database's multi-schema approach enhances the study of phonological segmentation and processing.
• Researchers can utilize DoWLS-MAN to investigate various aspects of Mandarin psycholinguistics.