Development of the Mandarin matrix sentence test in noise.

Abstract

Speech audiometry is one of the most important tools for modern audiology and hearing research. Since speech audiometric tests are usually limited to a certain language, their formats and construction principles should yield a high comparability across languages (such as closed-set speech tests like the Digit Triplet Test or the Matrix test, see Kollmeier et al., IJA 54, 2015, pp. 3-16). One major challenge of developing test materials in Mandarin, more often referred to as Putonghua in Mainland China, is the diversity of dialects in spoken Mandarin due to the accent, the vocabulary, the educational status etc. This also calls for a closed-set format which makes the Matrix sentence recognition test a good candidate for standard sentence intelligibility measurements in noise for Mandarin. The Matrix test is already available in a compatible way for at least 16 languages thus allowing for multicenter and multilingual studies. The limited vocabulary and simple sentence format is suitable for potential broader Mandarin users in various regions of China, for both adults and children of school age. Furthermore, the test construction offers the opportunity of high repeatability within one subject. This study aims at developing a Mandarin Matrix test (CMNmatrix) according to the international standard procedure for both diagnostic and research purposes. The CMNmatrix consists of a corpus of phonologically balanced 50 words, with a construction of 10 names, 10 verbs, 10 numerals, 10 adjectives, and 10 nouns. All sentences have identical syntactical structure and unpredictable semantic content for minimizing the memorization effect. The details of the development of CMNmatrix will be presented including the test design, the speech material selection, the sentence, recording, the subsequent cutting and reconstruction procedure, the optimization and the preliminary multi-center evaluation. This work was funded by the DFG EXC 1077 hearing4all and EFRE-Project VIBHear.