An alignment matching method to explore pseudosyllable properties across different corpora

In previous work we have defined a pseudosyllable unit for an English read speech recognition task. In this study we investigate the robustness of extraction of the pseudosyllable units and investigate how such units can be integrated into speech recognition systems. An evaluation method that maps hypothesis phonemes to reference phonemes is proposed. Analysis is performed on pseudosyllables extracted from two different sets of speech data. Mutual information is used to look at the relationship between different pseudosyllabic aspects and the error patterns of hypothesis phonemes. It is shown that the pseudosyllable extraction algorithm is robust and gives units with consistent statistics. Pseudosyllables which have a phone triplet structure tend to have lower insertion error. In temporal regions where pseudosyllables overlap with each other, more insertion errors may occur.