pith. sign in

arxiv: 2405.07006 · v2 · pith:DVKENU4Tnew · submitted 2024-05-11 · 💻 cs.CL

Word-specific tonal realizations in Mandarin

Yu-Ying Chuang , Melanie J. Bell , Yu-Hsiang Tseng , R. Harald Baayen This is my paper

Pith reviewed 2026-05-24 01:21 UTC · model grok-4.3

classification 💻 cs.CL
keywords Mandarin tonestonal realizationword-specific effectsgeneralized additive modelscontextual embeddingsspontaneous conversationslexical semantics
0
0 comments X

The pith

Word type and contextual meaning shape Mandarin tonal realizations more strongly than form-related factors.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper demonstrates that the pitch contours of Mandarin two-character words are influenced by the specific word and its meaning in context, even after accounting for speaker, speech rate, co-articulation, and other form factors. Using a generalized additive regression model, it finds that word type alone predicts tonal patterns better than all those traditional predictors together. Incorporating meaning information from context improves the prediction further. Computational models then show that pitch contours can identify word types at 50 percent accuracy and that embeddings can predict contour shapes at 40 percent accuracy on new data, levels well above chance. This indicates that phonetic details of tones carry usable semantic information.

Core claim

Tonal realization is partially determined by words' meanings. After controlling for effects of speaker and context, word type is a stronger predictor of tonal realization than all the previously established word-form related predictors combined. The addition of information about meaning in context improves prediction accuracy even further. Token-specific pitch contours predict word type with 50% accuracy on held-out data, and context-sensitive, token-specific embeddings can predict the shape of pitch contours with 40% accuracy.

What carries the argument

Generalized additive regression model isolating word-type effects after controlling for form predictors, and bidirectional computational modeling with context-specific word embeddings.

If this is right

  • Lexical meaning directly affects the phonetic realization of tones in Mandarin.
  • The link between pitch contours and word meanings is strong enough to be potentially functional in language use.
  • Standard models of tonal production must be extended to include semantic factors.
  • Acoustic models for word recognition can leverage these tonal variations for better performance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Listeners might exploit word-specific tone variations to resolve ambiguities in conversation.
  • Speech technology for tonal languages could benefit from incorporating word identity into tone generation models.
  • Similar effects may be present in other tonal languages and warrant investigation.

Load-bearing premise

The generalized additive model successfully isolates the effects of word type by fully controlling for all word-form related predictors without any remaining confounding.

What would settle it

Finding that the prediction accuracy for word type from pitch contours drops to near chance level when tested on completely new speakers and contexts, or that adding word type does not improve the regression model fit after the form predictors.

Figures

Figures reproduced from arXiv: 2405.07006 by Melanie J. Bell, R. Harald Baayen, Yu-Hsiang Tseng, Yu-Ying Chuang.

Figure 2
Figure 2. Figure 2: The dotted line at y = 0 is a reference line: an adjustment curve for a given word that followed this line would indicate that no adjustment is needed and that this word’s pitch is identical to the population contour. Deviations above this reference line indicate an upward F0 adjustment, and deviations below it indicate a downward adjustment. The word 職業 zhi2ye4 ‘profession’, for example, represented by a … view at source ↗
Figure 14
Figure 14. Figure 14: For training data (left), accuracies are between 40% and 50%. The accuracies for [PITH_FULL_IMAGE:figures/full_fig_p029_14.png] view at source ↗
read the original abstract

The pitch contours of Mandarin two-character words are generally understood as being shaped by the underlying tones of the constituent single-character words, in interaction with articulatory constraints imposed by factors such as speech rate, co-articulation with adjacent tones, segmental make-up, and predictability. This study shows that tonal realization is also partially determined by words' meanings. We first show, on the basis of a corpus of Taiwan Mandarin spontaneous conversations, using a generalized additive regression model, and focusing on the rise-fall tone pattern, that after controlling for effects of speaker and context, word type is a stronger predictor of tonal realization than all the previously established word-form related predictors combined. Importantly, the addition of information about meaning in context improves prediction accuracy even further. We then proceed to show, using computational modeling with context-specific word embeddings, that token-specific pitch contours predict word type with 50% accuracy on held-out data, and that context-sensitive, token-specific embeddings can predict the shape of pitch contours with 40% accuracy. These accuracies, which are an order of magnitude above chance level, suggest that the relation between words' pitch contours and their meanings are sufficiently strong to be potentially functional for language users. The theoretical implications of these empirical findings are discussed.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper claims that tonal realizations of Mandarin two-character words (focusing on rise-fall patterns) in spontaneous Taiwan Mandarin speech are shaped not only by underlying tones and articulatory factors (speech rate, co-articulation, segmental makeup, predictability) but also by word-specific meanings. Using a generalized additive regression model on corpus data, it reports that word type is a stronger predictor than all word-form predictors combined after controlling for speaker and context; adding contextual meaning information further improves accuracy. Computational modeling with context-specific embeddings then shows token-specific pitch contours predict word type at 50% accuracy and embeddings predict pitch contour shape at 40% accuracy on held-out data—both well above chance—suggesting the pitch-meaning relation may be functionally relevant.

Significance. If the GAM controls adequately isolate word-type effects without residual confounding, the result would extend phonetic research on tone by demonstrating lexically specific, meaning-driven variation beyond established form-based predictors, with potential implications for models of speech production and perception. The held-out predictive modeling provides a quantitative check on effect strength and is a methodological strength. However, the central claim's defensibility depends on unverified aspects of model specification.

major comments (2)
  1. [generalized additive regression model (Abstract and modeling description)] The generalized additive regression model is presented as successfully isolating word-type effects after controlling for speaker, context, and word-form predictors, but the manuscript provides no concurvity diagnostics, variance inflation factors, or explicit checks for correlations between word type and unmodeled variables (e.g., prosodic boundary strength, syntactic role, or discourse status). This is load-bearing for the claim that word type outperforms the combined word-form predictors, as spontaneous-speech data make such correlations likely.
  2. [computational modeling with context-specific word embeddings (Abstract and results)] The reported 50% and 40% held-out accuracies for word-type prediction from pitch contours and embedding-to-pitch mapping lack details on exact controls, error estimation procedures, potential post-hoc model choices, or validation of the embedding-to-pitch mapping. These omissions directly affect assessment of whether the accuracies reflect genuine word-specific effects rather than artifacts of the modeling pipeline.
minor comments (1)
  1. [Abstract] The abstract and modeling sections would benefit from explicit statements of the number of observations, exact basis functions/smoothing parameters used in the GAM, and the precise definition of 'word type' versus 'word-form predictors' to improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight important aspects of methodological transparency. We address each major comment below and will revise the manuscript to incorporate additional details and diagnostics as outlined.

read point-by-point responses

  1. Referee: The generalized additive regression model is presented as successfully isolating word-type effects after controlling for speaker, context, and word-form predictors, but the manuscript provides no concurvity diagnostics, variance inflation factors, or explicit checks for correlations between word type and unmodeled variables (e.g., prosodic boundary strength, syntactic role, or discourse status). This is load-bearing for the claim that word type outperforms the combined word-form predictors, as spontaneous-speech data make such correlations likely.

    Authors: We agree that explicit reporting of concurvity diagnostics and variance inflation factors would strengthen the presentation. The current manuscript describes the GAM specification and the set of word-form controls but does not include these post-fit checks. In revision we will add concurvity scores for the smooth terms, VIF values for the parametric predictors, and a supplementary analysis examining correlations between word type and available corpus annotations for prosodic boundary strength and syntactic position. These additions will allow readers to verify that the reported word-type effects are not driven by residual confounding. revision: yes

  2. Referee: The reported 50% and 40% held-out accuracies for word-type prediction from pitch contours and embedding-to-pitch mapping lack details on exact controls, error estimation procedures, potential post-hoc model choices, or validation of the embedding-to-pitch mapping. These omissions directly affect assessment of whether the accuracies reflect genuine word-specific effects rather than artifacts of the modeling pipeline.

    Authors: The manuscript reports held-out accuracies well above chance but does not provide the full pipeline details requested. We will expand the computational modeling section to specify the exact cross-validation scheme, the number of random seeds used for error estimation, the absence of post-hoc hyperparameter tuning, and the precise procedure used to map embeddings to pitch-contour parameters. These clarifications will be added without altering the reported accuracy figures. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical corpus study with held-out prediction

full rationale

The paper reports GAM fits on spontaneous speech data controlling for speaker/context and word-form predictors, followed by held-out accuracies (50% word-type prediction from pitch contours; 40% pitch prediction from embeddings). These are standard out-of-sample evaluations with no self-definitional equations, fitted parameters renamed as predictions, or load-bearing self-citations. The central claims rest on independent data splits and external modeling rather than reduction to the same fitted quantities by construction. This is the expected non-finding for a predictive modeling study on held-out data.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claims rest on statistical modeling assumptions and the representational adequacy of embeddings rather than new physical postulates; free parameters are implicit in the GAM smooths and embedding training.

free parameters (2)
  • GAM smoothing parameters and basis functions
    Chosen to fit pitch contour data after controlling for listed predictors.
  • Embedding model hyperparameters and training objective
    Context-specific word embeddings trained on corpus data to capture meaning.
axioms (2)
  • domain assumption The Taiwan Mandarin conversation corpus is representative and annotations of tones and context are accurate.
    Data source for all regression and embedding analyses.
  • domain assumption Context-sensitive embeddings encode semantic information relevant to tonal choice.
    Invoked when using embeddings to predict or be predicted by pitch contours.

pith-pipeline@v0.9.0 · 5756 in / 1427 out tokens · 84796 ms · 2026-05-24T01:21:10.203390+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

105 extracted references · 105 canonical work pages

  1. [1]

    Akaike, H. (1998). Information theory and an extension of the maximum likelihood principle. In Selected papers of Hirotugu Akaike , pages 199--213. Springer

    work page 1998

  2. [2]

    H., Chuang, Y.-Y., Shafaei-Bajestan, E., and Blevins, J

    Baayen, R. H., Chuang, Y.-Y., Shafaei-Bajestan, E., and Blevins, J. P. (2019). The discriminative lexicon: A unified computational model for the lexicon and lexical processing in comprehension and production grounded not in (de)composition but in linear discriminative learning. Complexity , 2019:4895891

    work page 2019

  3. [3]

    H., Fasiolo, M., Wood, S., and Chuang, Y.-Y

    Baayen, R. H., Fasiolo, M., Wood, S., and Chuang, Y.-Y. (2022). A note on the modeling of the effects of experimental time in psycholinguistic experiments. The Mental Lexicon , 17(2):178--212

    work page 2022

  4. [4]

    Bell, A., Jurafsky, D., Fosler-Lussier, E., Girand, C., Gregory, M., and Gildea, D. (2003). Effects of disfluencies, predictability, and utterance position on word form variation in E nglish conversation. The Journal of the Acoustical Society of America , 113(2):1001--1024

    work page 2003

  5. [5]

    Bi, Y., Chen, Y., and Schiller, N. O. (2015). The effect of word frequency and neighbourhood density on tone merge. In Proceedings of the 18th International Congress of Phonetic Sciences , Glasgow, Scotland

    work page 2015

  6. [6]

    and Weenink, D

    Boersma, P. and Weenink, D. (2019). Praat: doing phonetics by computer [computer program]. http://www.praat.org/

    work page 2019

  7. [7]

    Bojanowski, P., Grave, E., Joulin, A., and Mikolov, T. (2017). Enriching word vectors with subword information. Transactions of the Association for Computational Linguistics , 5:135--146

    work page 2017

  8. [8]

    Bruni, E., Tran, N.-K., and Baroni, M. (2014). Multimodal distributional semantics. Journal of Artificial Intelligence Research , 49:1--47

    work page 2014

  9. [9]

    Chao, Y. R. (1968). A grammar of spoken Chinese . University of California Press

    work page 1968

  10. [10]

    Chen, Y. (2010). Post-focus f0 compression—now you see it, now you don’t. Journal of Phonetics , 38(4):517--525

    work page 2010

  11. [11]

    and Xu, Y

    Cheng, C. and Xu, Y. (2015). Mechanism of disyllabic tonal reduction in T aiwan M andarin. Language and Speech , 58(3):281--314

    work page 2015

  12. [12]

    and Baayen, R

    Chuang, Y.-Y. and Baayen, R. H. (2021). Discriminative learning and the lexicon: NDL and LDL . In Oxford Research Encyclopedia of Linguistics . Oxford University Press

    work page 2021

  13. [13]

    Chuang, Y.-Y., Huang, Y.-H., and Fon, J. (2007). The effect of incredulity and particle on the intonation of yes/no questions in T aiwan M andarin. In Proceedings of the 16th International Congress of Phonetic Sciences , pages 1261--1264, Saarbr\" u cken, Germany

    work page 2007

  14. [14]

    F., and Baayen, R

    Chuang, Y.-Y., Kang, M., Luo, X. F., and Baayen, R. H. (2023). Vector space morphology with linear discriminative learning. In Crepaldi, D., editor, Linguistic morphology in the mind and brain . Routledge

    work page 2023

  15. [15]

    Chung, K. S. (2006). Contraction and backgrounding in Taiwan Mandarin . Concentric: Studies in Linguistics , 32(1):69--88

    work page 2006

  16. [16]

    and Clifton Jr., C

    Cutler, A. and Clifton Jr., C. (1999). Comprehending spoken language: a blueprint of the listener. In Brown, C. and Hagoort, P., editors, The N eurocognition of L anguage , pages 123--166. Oxford U niversity P ress, Oxford

    work page 1999

  17. [17]

    Drager, K. K. (2011). Sociophonetic variation and the lemma. Journal of Phonetics , 39(4):694--707

    work page 2011

  18. [18]

    Duanmu, S. (2007). The phonology of standard Chinese . OUP Oxford

    work page 2007

  19. [19]

    Elman, J. L. (2009). On the meaning of words and dinosaur bones: Lexical knowledge without a lexicon. Cognitive Science , 33(4):547--582

    work page 2009

  20. [20]

    Ernestus, M. (2000). Voice assimilation and segment reduction in casual D utch. A corpus-based study of the phonology-phonetics interface . LOT, Utrecht

    work page 2000

  21. [21]

    Firth, J. R. (1968). Selected papers of J R Firth, 1952-59 . Indiana University Press

    work page 1968

  22. [22]

    Fon, J. (2004). A preliminary construction of T aiwan S outhern M in spontaneous speech corpus. Technical Report NSC-92-2411-H-003-050-, National Science Council, Taipei, Taiwan

    work page 2004

  23. [23]

    and Chiang, W.-Y

    Fon, J. and Chiang, W.-Y. (1999). What does Chao have to say about tones?-a case study of Taiwan Mandarin . Journal of Chinese Linguistics , 27(1):13--37

    work page 1999

  24. [24]

    and Hsu, H.-J

    Fon, J. and Hsu, H.-J. (2007). Positional and phonotactic effects on the realization of dipping tones in Taiwan Mandarin . In Gussenhoven, C. and Riad, T., editors, Phonology and Phonetics, Tones and Tunes: Vol. 2. Experimental Studies in Word and Sentence Prosody , pages 239--269. Mouton de Gruyter, Berlin

    work page 2007

  25. [25]

    Fu, J.-W. (1999). Chinese tonal variation and social network --- A case study in Tantzu Junior High School, Taichung, Taiwan . Master's thesis, Providence University

    work page 1999

  26. [26]

    Gahl, S. (2008). Time and thyme are not homophones: The effect of lemma frequency on word durations in spontaneous speech. Language , 84(3):474--496

    work page 2008

  27. [27]

    and Baayen, R

    Gahl, S. and Baayen, R. H. (2024). Time and thyme again: Connecting E nglish spoken word duration to models of the mental lexicon. Language , 100(4):623--670

    work page 2024

  28. [28]

    Gahl, S., Yao, Y., and Johnson, K. (2012). Why reduce? P honological neighborhood density and phonetic reduction in spontaneous speech. Journal of Memory and Language , 66(4):789--806

    work page 2012

  29. [29]

    G rding, E. (1987). Speech act and tonal pattern in Standard Chinese : constancy and variation. Phonetica , 44(1):13--29

    work page 1987

  30. [30]

    Goldman, J.-P. (2011). Easyalign: An automatic phonetic alignment tool under praat. In Interspeech , volume 12, pages 3233--3236

    work page 2011

  31. [31]

    G \"u nther, F., Rinaldi, L., and Marelli, M. (2019). Vector-space models of semantic representation from a cognitive perspective: A discussion of common misconceptions. Perspectives on Psychological Science , 14(6):1006--1033

    work page 2019

  32. [32]

    Harris, Z. S. (1954). Distributional structure. WORD , 10(2-3):146--162

    work page 1954

  33. [33]

    Heitmeier, M., Chuang, Y.-Y., and Baayen, R. H. (2021). Modeling morphology with linear discriminative learning: Considerations and design choices. Frontiers in Psychology , 12:720713

    work page 2021

  34. [34]

    Heitmeier, M., Chuang, Y.-Y., and Baayen, R. H. (2023). How trial-to-trial learning shapes mappings in the mental lexicon: Modelling lexical decision with linear discriminative learning. Cognitive Psychology , 146:101598

    work page 2023

  35. [35]

    Heitmeier, M., Chuang, Y.-Y., and Baayen, R. H. (2025). The Discriminative Lexicon: Theory and implementation in the julia package JudiLing . In preparation for Cambridge University Press

    work page 2025

  36. [36]

    Ho, A. T. (1976). The acoustic variation of M andarin tones. Phonetica , 33(5):353--367

    work page 1976

  37. [37]

    Howie, J. M. (1974). On the domain of tone in Mandarin . Phonetica , 30(3):129--148

    work page 1974

  38. [38]

    Hsieh, P.-j. (2013). Prosodic markings of semantic predictability in Taiwan Mandarin . In INTERSPEECH , pages 553--557

    work page 2013

  39. [39]

    and Tseng, Y.-H

    Hsieh, S.-K. and Tseng, Y.-H. (2020). Tutorial on sense-aware computing in chinese (version 0.1.6). In Paper presented in 32nd conference on Computational Linguistics and Speech Processing (ROCLING 2020)

    work page 2020

  40. [40]

    Huang, C.-R., Hsieh, S.-K., Hong, J.-F., Chen, Y.-Z., Su, I.-L., Chen, Y.-X., and Huang, S.-W. (2010). Constructing Chinese Wordnet: Design Principles and Implementation. (in Chinese) . Zhong-Guo-Yu-Wen , 24:2:169--186

    work page 2010

  41. [41]

    Huang, E., Socher, R., Manning, C., and Ng, A. (2012). Improving word representations via global context and multiple word prototypes. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages 873--882, Jeju Island, Korea. Association for Computational Linguistics

    work page 2012

  42. [42]

    Huang, J., Tang, D., Zhong, W., Lu, S., Shou, L., Gong, M., Jiang, D., and Duan, N. (2021). W hitening BERT : An easy unsupervised sentence embedding approach. In Moens, M.-F., Huang, X., Specia, L., and Yih, S. W.-t., editors, Findings of the Association for Computational Linguistics: EMNLP 2021 , pages 238--244, Punta Cana, Dominican Republic. Associati...

    work page 2021

  43. [43]

    and Chiu, C

    Huang, P.-H. and Chiu, C. (2023). Production and perception of coarticulated tones: The cases of Taiwan Mandarin and Taiwan Southern Min . Available at SSRN 4637487

    work page 2023

  44. [44]

    Huang, Y.-H. (2008). Dialectal variations on the realization of high tonal targets in Taiwan Mandarin . Master's thesis, National Taiwan University

    work page 2008

  45. [45]

    T., and Navigli, R

    Iacobacci, I., Pilehvar, M. T., and Navigli, R. (2015). S ens E mbed: Learning sense embeddings for word and relational similarity. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) , pages 95--105, Beijing, China. Ass...

    work page 2015

  46. [46]

    Johnson, K. (2004). Massive reduction in conversational A merican E nglish. In Spontaneous speech: data and analysis. Proceedings of the 1st session of the 10th international symposium , pages 29--54, Tokyo, J apan. The N ational I nternational I nstitute for J apanese L anguage

    work page 2004

  47. [47]

    Kendall, D. G. (1977). The diffusion of shape. Advances in Applied Probability , 9(3):428--430

    work page 1977

  48. [48]

    Kilgarriff, A. (2007). Word senses. In Agirre, E. and Edmonds, P., editors, Word Sense Disambiguation: Algorithms and Applications , pages 29--46. Springer

    work page 2007

  49. [49]

    Kuhn, M. (2013). Applied predictive modeling . Springer

    work page 2013

  50. [50]

    and Silverman, K

    Ladd, R. and Silverman, K. E. (1984). Vowel intrinsic pitch in connected speech. Phonetica , 41(1):31--40

    work page 1984

  51. [51]

    and Dumais, S

    Landauer, T. and Dumais, S. (1997). A solution to P lato's problem: T he latent semantic analysis theory of acquisition, induction and representation of knowledge. Psychological R eview , 104(2):211--240

    work page 1997

  52. [52]

    Lee, O. J. (2005). The prosody of questions in Beijing Mandarin . The Ohio State University

    work page 2005

  53. [53]

    J., Roelofs, A., and Meyer, A

    Levelt, W. J., Roelofs, A., and Meyer, A. S. (1999). A theory of lexical access in speech production. Behavioral and Brain Sciences , 22(1):1--38

    work page 1999

  54. [54]

    and Chen, Y

    Li, Q. and Chen, Y. (2016). An acoustic study of contextual tonal variation in Tianjin Mandarin . Journal of Phonetics , 54:123--150

    work page 2016

  55. [55]

    and Xu, Y

    Liu, F. and Xu, Y. (2005). Parallel encoding of focus and interrogative meaning in M andarin intonation. Phonetica , 62(2-4):70--87

    work page 2005

  56. [56]

    Lohmann, A. (2018). Cut (n) and cut (v) are not homophones: Lemma frequency affects the duration of noun--verb conversion pairs. Journal of Linguistics , 54(4):753--777

    work page 2018

  57. [57]

    and Chen, K.-J

    Ma, W.-Y. and Chen, K.-J. (2003). Introduction to CKIP C hinese word segmentation system for the first international C hinese word segmentation bakeoff. In Proceedings of the Second SIGHAN Workshop on C hinese Language Processing , pages 168--171, Sapporo, Japan. Association for Computational Linguistics

    work page 2003

  58. [58]

    Maaten, L. v. d. and Hinton, G. (2008). Visualizing data using t-SNE . Journal of Machine Learning Research , 9(11):2579--2605

    work page 2008

  59. [59]

    Marsolek, C. J. (2008). What antipriming reveals about priming. Trends in C ognitive S cience , 12(5):176--181

    work page 2008

  60. [60]

    Martinet, A. (1965). La Linguistique Synchronique: \'Etudes et Recherches . Presses Universitaires de France, Paris

    work page 1965

  61. [61]

    Moore, C. B. and Jongman, A. (1997). Speaker normalization in the perception of Mandarin Chinese tones. The Journal of the Acoustical Society of America , 102(3):1864--1877

    work page 1997

  62. [62]

    Neelakantan, A., Shankar, J., Passos, A., and McCallum, A. (2014). Efficient non-parametric estimation of multiple embeddings per word in vector space. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing ( EMNLP ) , pages 1059--1069, Doha, Qatar. Association for Computational Linguistics

    work page 2014

  63. [63]

    Nieder, J., Chuang, Y.-Y., van de Vijver, R., and Baayen, R. H. (2023). A discriminative lexicon approach to word comprehension, production, and processing: Maltese plurals. Language , 99(2)

    work page 2023

  64. [64]

    Ouyang, I. C. and Kaiser, E. (2015). Prosody and information structure in a tone language: an investigation of Mandarin Chinese . Language, Cognition and Neuroscience , 30(1-2):57--72

    work page 2015

  65. [65]

    and Hilpert, M

    Perek, F. and Hilpert, M. (2017). A distributional semantic approach to the periodization of change in the productivity of constructions. International Journal of Corpus Linguistics , 22(4):490--520

    work page 2017

  66. [66]

    Pilehvar, M. T. and Camacho-Collados, J. (2020). Embeddings in natural language processing: Theory and advances in vector representations of meaning . Morgan & Claypool Publishers

    work page 2020

  67. [67]

    Plag, I., Homann, J., and Kunter, G. (2017). Homophony and morphology: The acoustics of word-final S in E nglish. Journal of Linguistics , 53(1):181--216

    work page 2017

  68. [68]

    R: A Language and Environment for Statistical Computing

    R Core Team (2022). R: A Language and Environment for Statistical Computing . R Foundation for Statistical Computing, Vienna, Austria

    work page 2022

  69. [69]

    and Mooney, R

    Reisinger, J. and Mooney, R. J. (2010). Multi-prototype vector-space models of word meaning. In Human Language Technologies: The 2010 Annual Conference of the North A merican Chapter of the Association for Computational Linguistics , pages 109--117, Los Angeles, California. Association for Computational Linguistics

    work page 2010

  70. [70]

    Saito, M., Tomaschek, F., and Baayen, R. H. (2023). Articulatory effects of frequency modulated by inflectional meanings. In Schlechtweg, M., editor, Interfaces of Phonetics . De Gruyter

    work page 2023

  71. [71]

    Salton, G., Wong, A., and Yang, C. S. (1975). A vector space model for automatic indexing. Commun. ACM , 18(11):613–620

    work page 1975

  72. [72]

    Sampson, G. (2015). A chinese phonological enigma. Journal of Chinese Linguistics , 43(2):679--691

    work page 2015

  73. [73]

    Sampson, G. (2019). An unaddressed phonological contradiction. International Journal of Chinese Linguistics , 6(2):221--237

    work page 2019

  74. [74]

    Sch\" u tze, H. (1992). Word space. In Hanson, S., Cowan, J., and Giles, C., editors, Advances in Neural Information Processing Systems , volume 5. Morgan-Kaufmann

    work page 1992

  75. [75]

    Shen, X. S. (1989). Interplay of the four citation tones and intonation in Mandarin Chinese . Journal of Chinese Linguistics , 17(1):61--74

    work page 1989

  76. [76]

    Shen, X. S. (1990a). The prosody of Mandarin Chinese , volume 118. University of California Press

    work page

  77. [77]

    Shen, X. S. (1990b). Tonal coarticulation in M andarin. Journal of Phonetics , 18(2):281--295

    work page

  78. [78]

    Shen, X. S. and Lin, M. (1991). A perceptual study of M andarin tones 2 and 3. Language and Speech , 34(2):145--156

    work page 1991

  79. [79]

    and Zhang, J

    Shi, B. and Zhang, J. (1987). Vowel intrinsic pitch in S tandard C hinese. In Proceedings of the 11th International Congress of Phonetic Sciences , pages 142--145

    work page 1987

  80. [80]

    Shih, C. (1988). Tone and intonation in M andarin. Working Papers, Cornell Phonetics Laboratory , 3:83--109

    work page 1988

Showing first 80 references.