Relationship between perceptual accuracy and information measures: A cross-linguistic study

Abstract

Relationship between perceptual accuracy and information measures: A cross-linguistic study by Shinae Kang Doctor of Philosophy in LinguisticsUniversity of California, BerkeleyProfessor Keith A. Johnson, ChairThe current dissertation studies how the information conveyed by different speech elements of English, Japanese and Korean correlates with perceptual accuracy. Two wellestablished information measures are used: weighted negative contextual predictability (informativity) of a speech element; and the contribution of a speech element to syllable differentiation, or functional load. This dissertation finds that the correlation between information and perceptual accuracy differs depending on both the type of information measure and the language of the listener.To compute the information measures, Chapter 2 introduces a new corpus consisting of all the possible syllables for each of the three languages. The chapter shows that the two information measures are inversely correlated.In Chapters 3 and 4, two perception experiments in audio-only and audiovisual modalities, respectively, are described. The experiments adopt a forced-choice-identification paradigm. In both experiments, subjects listened to VC.CV stimuli composed of spliced VC and CV chunks, and they were asked to identify the CC sequence. Multiple statistical models are constructed to predict the perceptual accuracy of a CC stop cluster from the associated information measures of relevant speech elements in the listeners' languages.The estimated models show that high informativity has a generally negative effect on the perceptual accuracy of stops. Functional load shows less consistent correlations with perceptual accuracy across different model specifications, but generally has a positive effect on the perceptual accuracy.In addition, Japanese listeners show highly consistent results across the two experiments and different model specifications. This contrasts with less consistent results for English and Korean listeners.The current dissertation provides the first empirical evidence for a significant relationship between informativity and functional load, and between the information measures of speech elements and their perceptual accuracy. Furthermore, it reveals how listeners' native 2 languages affect that relationship, and how the cross-linguistic variation of that relationship may be related to the characteristics of individual languages such as their syllable structures.

Relationships Between Functional Load and Auditory Confusability Under Different Speech Environments

“…FL is one of the measures adopted to represent information in information-theoretic approaches. Although other measures, such as the information content in a word, have been used to explain behaviors such as stop deletion during production or phonetic reduction [7,8] (see also Section 1.2 in [9] for review), we chose functional load because it can capture the relationship between the structure of the lexicon and individual phonemes, rather than words, in a principled way.…”

“…Then, the overall FL of a phoneme x, expressed F L (x), is the sum of all such pair-wise FL F L(X, y), multiplied by 1/2 for normalization purposes [9];…”

“…FL has proven a useful tool for explaining sound-related linguistic phenomena. Sound mergers, for example, are more likely between phonemes that have lower pairwise FL [12], and people are more likely to correctly distinguish a pair of phonemes when the FL of that pair is higher [9]. The current project asks whether FL can also help explain the structure of the English lexicon.…”