1 / 28

Yao Yao @ LSA 2010-1-7

Separating speaker- and listener-oriented forces in speech – Evidence from phonological neighborhood density. Yao Yao @ LSA 2010-1-7. Introduction | Methodology | Linear mixed-effects model | Discussion. phonetic variation. Widely exists in spontaneous speech Duration

nell
Download Presentation

Yao Yao @ LSA 2010-1-7

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Separating speaker- and listener-oriented forces in speech–Evidence from phonological neighborhooddensity Yao Yao @ LSA 2010-1-7

  2. Introduction| Methodology | Linear mixed-effects model | Discussion phonetic variation • Widely exists in spontaneous speech • Duration • Segmental realization • Pitch • Why?

  3. Introduction| Methodology | Linear mixed-effects model | Discussion explaining variation Listener-oriented Talker-oriented Result of ease or difficulty of production Examples Shortening and reduction in High-frequency or high-predictability forms “articulatory routinization” (Bybee, 2001) • Response to different models of listener’s needs • Result of ease or difficulty of comprehension (modeled by the speaker) • Examples • Foreigner- and child-directed speech • Speech under noise • Shortening and reduction in • High-frequency or high-predictability forms Many word properties have the same predictions for comprehension and production…

  4. Introduction| Methodology | Linear mixed-effects model | Discussion general research question • Is it possible to tease apart talker- and listener-oriented forces in variation at the word level? Any word property with different predictions for comprehension and production? Yes!

  5. Introduction| Methodology | Linear mixed-effects model | Discussion phonological neighborhood density High-density words are hard for perception but easy for production (Dell & Gordon, 2003)

  6. Introduction| Methodology | Linear mixed-effects model | Discussion phonological neighborhood • Concept • Similar-sounding words are connected to each other and form phonological neighborhoods • Neighborhood density: number of phonological neighbors each word has • One-phoneme difference rule (Luce & Pisoni 1998, etc) add fad Additional factors: neighborhood freq. cap cat fat coat fight kite

  7. Introduction| Methodology | Linear mixed-effects model | Discussion phonological neighbors and word perception • Inhibition • Similar-sounding primes inhibit auditory word recognition (Goldinger & Pisoni 1989) • Slower (and less accurate) responses for words from dense neighborhoods in perceptual tasks (Luce & Pisoni 1998) • Perceptual identification, lexical decision and word naming tasks

  8. Introduction| Methodology | Linear mixed-effects model | Discussion phonological neighbors and word production • Facilitation • Words from dense neighborhoods induce fewer speech errors and have shorter latency times in picture naming tasks (Vitevitch 2002)

  9. Introduction| Methodology | Linear mixed-effects model | Discussion phonological neighbors and phonetic variation • Phonological neighbors • Both compete with and bring more activation to the target word • Either impede or facilitate the processing of the target word • How does neighborhood density tease apart the two accounts of variation? perception production

  10. Introduction| Methodology | Linear mixed-effects model | Discussion predictions • Talker-oriented • High-density words are easy to produce  shortening and reduction • Listener-oriented • High-density words are hard to perceive  lengthening and vowel dispersion • High-density words have more expanded vowel space (Wright 1997, Munson & Solomon 2004) and more nasalized vowels (Scarborough 2004)

  11. Introduction| Methodology | Linear mixed-effects model | Discussion keywords of current study • Spontaneous speech • Aspects of production • Word duration • Vowel production • High-density words are shorter  talker-oriented

  12. Introduction| Methodology | Linear mixed-effects model | Discussion data • Buckeye corpus (Pitt et al 2007) • 40 speakers, ~300,000 words • Target words • CVC • Monomorphemic • Content words • 414 word types / 13,858 tokens

  13. Introduction| Methodology | Linear mixed-effects model | Discussion neighborhood measures • Two separate variables (from Hoosier Mental Lexicon; Nusbaum et al, 1984) • Neighborhood density (i.e. # of neighbors) • Using the 1-phoneme difference rule • Average neighbor frequency

  14. Introduction| Methodology | Linear mixed-effects model | Discussion coding variables • Outcome variable • Word token duration • Control variables • Baseline duration • Speaker characteristics • sex, age • Other lexical properties • word freq, length (in letters), familiarity, imageability, POS, phonotactic probability • Contextual factors • pre/fw predictability, pre/fw speech rate, disfluency, pre mentions

  15. Introduction| Methodology | Linear mixed-effects model | Discussion linear mixed-effects model • Fixed effects • All predictors • Neighborhood measures • Control variables • Random effects • Speaker • Word

  16. Introduction| Methodology | Linear mixed-effects model | Discussion modeling results • Neighborhood density • A significant negative effect • More neighbors  shorter duration • Facilitation • Neighbor frequency • Insignificant

  17. Introduction| Methodology | Linear mixed-effects model | Discussion partial effect of neighborhood density Effect confirmed by model evaluation.

  18. Introduction| Methodology | Linear mixed-effects model | Discussion confounding factor? • Phonotactic probability • The frequency with which a phonological segment, […] and a sequence of phonological segments, […] occur in a given position in a word (Jusczyk et al, 1994) • Correlated with neighborhood density (r = 0.46) • Phonotactic probability is never significant in the model, with or without neighborhood measures • The facilitative effect is at the lexical level, not the sublexical level

  19. Introduction| Methodology | Linear mixed-effects model | Discussion implications • Evidence for talker-oriented account • Talker-oriented: High-density words are easy to produce  shortening and reduction • Listener-oriented: High density words are hard to perceive  lengthening and vowel dispersion Ease of articulation? Fast lexical access? Not really… Probably… Synchrony between planning and articulation (Bell et al, 2009)

  20. Introduction| Methodology | Linear mixed-effects model | Discussion looking back… • Conflict with previous experimental results? • Wright (1997) and Munson & Solomon (2004): Vowel dispersion in high-density words • Shorter but more expanded vowels? • Differences in the type of speech? • Maybe it’s not density, but neighbor frequency… • Preliminary results in the current dataset: NO effect of density, but words with high-frequency neighbors have more expanded vowel space • Previous results can also be explained by neighbor frequency

  21. Introduction| Methodology | Linear mixed-effects model | Discussion conclusion • Facilitative effect of neighorhood density on word duration • Unambiguous evidence for the talker-oriented account of phonetic variation • Ongoing work: effect of phonological neighborhoods on vowel production

  22. Introduction| Methodology | Linear mixed-effects model | Discussion The end…

  23. selected references • Dell & Gordon(2003). Neighbors in the lexicon: Friends or foes? In N.O. Schiller and A.S. Meyer (eds.), Phonetics and phonology in language comprehension and production: Differences and similarities. New York: Mouton. • Luce & Pisoni (1998) Recognizing spoken words: the Neighborhood Activation Model. Ear & Hearing, 19, 1-36. • Munson & Solomon (2004) The effect of phonological neighborhood on vowel articulation. JSLHR, 47, 1048-1058. • Pitt et al (2007Buckeye Corpus of Conversational Speech (2nd release) [www.buckeyecorpus.osu.edu] Columbus, OH: Department of Psychology, Ohio State University (Distributor). • Scarborough (2004). Lexical confusability and degree of coarticulation. Proceedings of the 29th Annual Meeting of the • Berkeley Linguistics Society. • Vitevitch (2002) The influence of phonological similarity neighborhoods on speech production. J. of Experimental Psychology: Learning, Memory and Cognition, 28, 735-747. • Wright (1997) Lexical competition and reduction in speech: A preliminary report. . Research on Spoken Language Processing Progress Report. 21, 471-485. Indiana University

  24. Thanks to… • Prof. Susanne Gahl and Prof. Keith Johnson for helpful discussion • Anonymous subjects in Buckeye • Buckeye corpus developers

  25. Introduction| Methodology | Linear mixed-effects model | Discussion Perception & Production Dell & Gordon (2003) Perception Production add fad cap cat fat coat fight kite

  26. Introduction| Methodology | Linear mixed-effects model | Discussion model evaluation • Confirms the robustness of the results • Testing t-values • Model comparison • Cross-validation

  27. Introduction| Methodology | Linear mixed-effects model | Discussion Individual differences Having one more neighbor decreases duration by 0.4%

  28. Introduction| Methodology | Linear mixed-effects model | Discussion Distribution of neighborhood density and neighbor frequency

More Related