A test word selection and optimization method for a new Swedish test of phonetic perception in noise

The present project aims to develop a highly reliable, and clinically suitable, Swedish language test of phonetic perception in background noise for adult persons with hearing loss. The test will be a multiple choice rhyme test based on test word groups consisting of real words with minimal phonemic contrast.
As previous literature clearly points out, for a speech perception test to be highly reliable, all contrasting test items need to be as equally difficult as possible. This is often achieved through a process of test stimuli optimization. The present study uses an optimization strategy based on test word perceptual probability. This approach is based on a thorough linguistic analysis in which a set of variables derived from measures previously proven to affect access to the mental lexicon, namely word frequency, phonetic neighborhood density, and phonotactic probability, were calculated for each word in a large corpus.
In order to accomplish this, a large word list of high quality phonetic transcriptions would be needed. Unfortunately, no such word list was readily available in Swedish. There exists, however, a large lexical database (NST), consisting of many, but unfortunately often erroneous, phonetic transcriptions. Using the NST-database as a starting point, a revised and updated word list containing around 800 000 words with phonetic transcriptions reflecting natural everyday pronunciation of the central Swedish dialect has been compiled. To this list, word frequency data from a corpus of Swedish internet blogs with a size around 500 000 000 tokens have been added. In addition, the word list also contains data on spelling regularity, special orthographic characters, homographs and homophones, abbreviations/acronyms, non-Swedish words, as well as word class information.
In order to select the most suitable test words, this new list were utilized in order to identify the test word groups that have the least intra-group variation in word frequency, phonetic neighborhood density and phonotactic probabilities.
Equal perceptibility within these selected test word groups will then be attained through an acoustic optimization process, in which the sound levels of each contrasting phoneme (rather than of the whole test word, which is common procedure) are experimentally adjusted until a predefined degree of intelligibility is reached.
Our hypothesis is that this careful test word selection process will prove to decrease the amount of acoustic optimization needed to achieve equal perceptibility within the test word groups, and thereby increase both the reliability and the ecological validity of the test material.