BUILDING A SPEECH CORPUS OF CHILDREN WITH COCHLEAR IMPLANTS VIA AN ENHANCED METADATA STRUCTURE

This paper introduces a pilot speech corpus of 8 children with cochlear implant (CI) and proposes an enhanced metadata structure. The corpus is annotated with transcriptions and speech error patterns. A metadata structure for the corpus is proposed based on a meta-analysis of patient-related information included in studies on children with CI and a survey administered to speech and language pathologists asking clinical importance of given information. Then the results of the meta-analysis and the survey are combined to prioritize the patient-related information. The results show that the following elements are highly recommended: chronological age, implantation age, gender, duration of CI use, post-operational hearing threshold, onset time of hearing loss, information about assisting devices, pre- operational hearing threshold, speech perception performance, information about multiple disorders, and information about hearing aid. Particularly, the first four are the elements expected to be included in the speech corpus of children with CI.

Automated summary

This paper introduces a speech corpus of 8 children with cochlear implants and proposes an enhanced metadata structure. The study determines key patient-related information for the corpus based on a meta-analysis and survey of speech and language pathologists. The results highlight important elements including age, implantation age, gender, and duration of CI use.