很大一部份的語音學研究致力於發掘能決定性地區辨不同種音韻範疇（phonological category）的聲學信號。然而，大多數的研究著重在個別信號的特徵，而沒有去進一步地探討它們之間的互動和相對的權重。本研究針對兩個漢語方言：台灣華語和海陸客語當中的齒擦音（sibilant，又譯作咝音）進行了調查。決策樹和隨機森林被用於分析所得到的聲學特徵。此外，我們還利用二維的高斯混和模型（2D mixture of Gaussians (MOG) model）（Toscano and McMurray, 2010）運行了許多模擬分析來模擬齒擦音類別的感知學習過程。結果顯示一些對於分類有顯著效果的特徵，實際上可能會對於這樣一個無監督式學習模型的表現有所阻礙。我們認為只有少數的聲學信號對於齒擦音類別的感知是至關重要的，且這組信號有可能可以被不同語言的語者所感知到。我們還演示了一個MOG模型的應用：我們藉由使用橫跨兩個語言的輸入，模擬了所謂「非標準海陸客語」假設性的演變。這為MOG的架構開啟了新的可能性，同時也能讓我們對跨語言的分類學習有更好的理解。

A large part of phonetic studies have devoted to exploring the defining acoustic cues for distinguishing different classes of phonological categories. However, most studies have focused on the characteristics of individual cues without further exploring the interactions and relative weightings among them. Sibilants of two Sinitic languages, Taiwanese Mandarin and Hailu Hakka, were examined in this study. Decision trees and random forests were conducted to analyze the acoustic features obtained. Furthermore, we ran multiple simulations using the 2D mixture of Gaussians (MOG) model (Toscano and McMurray, 2010) to simulate the perceptual learning processes of sibilant categories. The results show that some features which are significant to the classification might in fact hinder the performance of such an unsupervised learning model. We argue that only a small number of acoustic cues are truly crucial to the perception of sibilant categories, and that this set of cues might be universally perceived by speakers across various languages. We also demonstrated an application for the MOG model where we simulated the hypothetical evolution of the so-called "Non-standard Hailu Hakka" by using inputs across two languages. This opens a new possibility for the MOG framework and might also provide us with a better understanding of categorical learning in a cross-linguistic context.

Bhattacharyya, A. (1943). On a measure of divergence between two statistical populations defined by their probability distributions. Bull. Calcutta Math. Soc., 35:99–109.
Boersma, P. and Hamann, S. (2008). The evolution of auditory dispersion in bidirectional constraint grammars. Phonology, 25(2):217–270.
Boersma, P. and Weenink, D. (2019). Praat: doing phonetics by computer [computer program], version 6.0.49. http://www.praat.org/.
Breiman, L., Friedman, J. H., Olshen, R. A., and Stone, C. J. (2017). Classification and regression trees. Routledge.
DiCanio, C. (2021). Spectral moments of fricative spectra script in praat, version 4.0. https://www.acsu.buffalo.edu/ ̃cdicanio/scripts.html/.
Duanmu, S. (2007). The phonology of standard Chinese. OUP Oxford.
Flemming, E. (1995). Auditory representations in phonology. PhD thesis, University of California, Los Angeles.
Flemming, E. (1996). Evidence for constraints on contrast: The dispersion theory of contrast. UCLA working papers in Phonology, 1:86–106.
Forrest, K., Weismer, G., Milenkovic, P., and Dougall, R. N. (1988). Statistical analysis of word-initial voiceless obstruents: Preliminary data. The Journal of the Acoustical Society of America, 84(1):115–123.
Getz, L. M., Nordeen, E. R., Vrabic, S. C., and Toscano, J. C. (2017). Modeling the development of audiovisual cue integration in speech perception. Brain sciences, 7(3):32.
Hakka Affairs Council (2017). National hakka population and language ability survey. https://www.hakka.gov.tw/File/Attach/37585/File_73865.pdf.
Hauser, I. (2022). Differential cue weighting in mandarin sibilant production.
James, G., Witten, D., Hastie, T., and Tibshirani, R. (2013). An introduction to statistical learning, volume 112. Springer.
Jeng, J.-Y. (2006). The acoustic spectral characteristics of retroflexed fricatives and affricates in taiwan mandarin. Journal of Humanistic Studies, 40(1):27–48.
Jongman, A., Wayland, R., and Wong, S. (2000). Acoustic characteristics of english fricatives. The Journal of the Acoustical Society of America, 108(3):1252–1263.
Ladefoged, P. and Wu, Z. (1984). Places of articulation: an investigation of pekingese fricatives and affricates. Journal of Phonetics, 12(3):267–278.
Lee, W.-S. and Zee, E. (2003). Standard chinese (beijing). Journal of the International Phonetic Association, 33(1):109–112.
Lee-Kim, S.-I. (2011). Spectral analysis of mandarin chinese sibilant fricatives. In ICPhS, pages 1178–1181.
Li, F., Edwards, J., and Beckman, M. (2007). Spectral measures for sibilant fricatives of english, japanese, and mandarin chinese. In Proceedings of the 16t International Congress of Phonetic Sciences, pages 917–920. Germany: Pirrot, Dudweiler.
Liaw, A. and Wiener, M. (2002). Classification and regression by randomforest. R News, 2(3):18–22.
Lin, Y.-H. (2007). The Sounds of Chinese with Audio CD, volume 1. Cambridge University Press.
Lo, Y.-c. (2021). An acoustic and articulatory study of sibilants in hailu hakka. Master’s thesis, National Tsing Hua University.
McAuliffe, M., Socolof, M., Mihuc, S., Wagner, M., and Sonderegger, M. (2017). Montreal forced aligner: trainable text-speech alignment using kaldi. In Proceedings of the 18th Conference of the International Speech Communication Association.
McMurray, B., Aslin, R. N., and Toscano, J. C. (2009). Statistical learning of phonetic categories: Insights from a computational approach. Developmental science, 12(3):369–378.
McMurray, B. and Jongman, A. (2011). What information is necessary for speech categorization? harnessing variability in the speech signal by integrating cues computed relative to expectations. Psychological review, 118(2):219.
Nowak, P. (2006). The role of vowel transitions and frication noise in the perception of polish sibilants. Journal of phonetics, 34(2):139–152.
Padgett, J. and ̇Zygis, M. (2007). The evolution of sibilants in polish and russian. Journal of Slavic linguistics, pages 291–324.
Python Software Foundation (2021). Python language reference, version 3.8.8. http://www.python.org/.
R Core Team (2021). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
Shadle, C. H. (2012). Acoustics and aerodynamics of fricatives. The Oxford handbook of laboratory phonology, pages 511–526.
Sussman, H. M., McCaffrey, H. A., and Matthews, S. A. (1991). An investigation of locus equations as a source of relational invariance for stop place categorization. The Journal of the Acoustical Society of America, 90(3):1309–1325.
Therneau, T. and Atkinson, B. (2019). rpart: Recursive Partitioning and Regression Trees. R package version 4.1-15.
Toscano, J. C. and McMurray, B. (2010). Cue integration with categories: Weighting acoustic cues in speech using unsupervised learning and distributional statistics. Cognitive science, 34(3):434–464.