ISCA Archive ICSLP 1994
ISCA Archive ICSLP 1994

Word complexity measures in the context of speech intelligibility tests

Arnd Mariniak

This paper describes a study where the results of an intelligibility test (the Cluster Identification test) were compared to lexical information of words. Several parameters like word class, word frequency, cluster frequency and the transition probabilities of clusters were included, all obtained from a corpus of about 6000 monosyllabic German words. Different sets of words which were read aloud by a professional news reader were presented to subjects with and without masking noise. Investigations were carried out as to whether and to what extent the above mentioned parameters influence the word recognition process. The aim of the study was to gather information to clarify the as yet term 'complexity' so that it can be applied to characterize the degree of difficulty of a task in speech intelligibility tests. However, its understanding is extremely important for a valid interpretation and comparison of the results from different tests. It can be shown that the most influencing factors are word class and word frequency whereas such factors as cluster frequency, cluster transition probability or phoneme transition probability do not play an important role. The main influencing factors seem to dominate the others being processed in the background.


Cite as: Mariniak, A. (1994) Word complexity measures in the context of speech intelligibility tests. Proc. 3rd International Conference on Spoken Language Processing (ICSLP 1994), 1311-1314

@inproceedings{mariniak94_icslp,
  author={Arnd Mariniak},
  title={{Word complexity measures in the context of speech intelligibility tests}},
  year=1994,
  booktitle={Proc. 3rd International Conference on Spoken Language Processing (ICSLP 1994)},
  pages={1311--1314}
}