Morphological Information and Fine Phonetic Detail in English Irregular Verbs
30
Morphological Information and Fine Phonetic Detail in English Irregular Verbs Michelle Sims, Benjamin V. Tucker, Harald Baayen University of Alberta Mental Lexicon Conference - Windsor, Ontario - 1 July 2010
-
Upload
independent -
Category
Documents
-
view
1 -
download
0
Transcript of Morphological Information and Fine Phonetic Detail in English Irregular Verbs
Morphological Information andFine Phonetic Detail in English
Irregular Verbs
Michelle Sims, Benjamin V. Tucker, Harald BaayenUniversity of Alberta
Mental Lexicon Conference - Windsor, Ontario - 1 July 2010
IntroductionSpeech VariationIrregular English VerbsSpeech Reduction Theories
MethodologyPredictors
Results & DiscussionMorphologyFrequencyGang SizeImageability
DiscussionInterpretation
Conclusions
Speech Variationeveryday speech is not careful butinstead contains a great amount of
phonetic reduction and variation
this acoustic signal is thefingerprint of speech processing
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results & Discussion
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
1
Irregular English Verbsdifference between the past andpresent tense is the alternation ofa single phonetic segment (vowel)
sing / sangrun / ran
hold / held
morphological effects on vocalicvariation in duration
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results & Discussion
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
2
Speech Reduction TheoriesLexical Confusability: the denserthe gang, the less likelihood forreduction (Scarborough 2004)
Smooth Signal Hypothesis: lexicalfrequency will negatively correlatewith vocalic duration (Aylett & Turk 2004)
Paradigmatic Signal EnhancementHypothesis: the more common thevocalic alternation, the longer thevowel (Kuperman et al 2007)
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results & Discussion
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
3
MethodologyBuckeye Corpus of Spoken English
Irregular English Verbs48 itemsmonosyllabicvocalic alternation between
past/present forms
7410 tokens total
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results & Discussion
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
4
Methodology: Predictorslinear mixed effects regression model
dependent variable:vocalic duration
independent variables:morphologylog frequencylog gang sizemean imageability rating
random effects factor:word lemma
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results & Discussion
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
5
Methodology: Predictorsindependent variable: log gang size
large gangs:/o/ to /u/
blow to blewthrow to threwknow to knew
small gangs:/ɔ/ to /ɛ/fall to fell
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results & Discussion
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
6
- longer durations in the past tense
Results: MorphologyIntroduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
7
- higher frequency correlates with longervocalic duration in the past tense
- no significant effect in the present tense
Results: FrequencyIntroduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
8
- bigger gangs correlate with shortervocalic durations in the past tense
- no significant effect in the present tense
Results: Gang SizeIntroduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
9
- more imageable words correlate withlonger vocal durations in the past tense
- no significant effect in the present tense
Results: ImageabilityIntroduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
10
Discussion
It appears that speech production issubject to morphology, frequency,
gang size and imageability...
though their effects do not followfrom current theories.
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
11
DiscussionSmooth Signal Hypothesis
predicted the more frequent a wordis, the less information it carries &the less information a word carries,the shorter it will be
This Data
past tense vowels are longer whenword frequency is high
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
12
DiscussionLexical Confusability
predicted that words belonging todenser neighborhoods will havelonger durations
This Data
past tense vowels are shorter whenthe gang they belong to is denser
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
13
DiscussionParadigmatic Signal Enhancement
predicted that words following morecommon paradigms will have longerdurations (increased acousticsalience)
This Data
past tense vowels are shorter whenthe vocalic alternation pattern theyfollow is more common
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
14
DiscussionImageability
predicted that more imageablewords have longer naming latencies
This Data
past tense vowels are longer whenthe word is more imageable
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
14
Discussion: Interpretationcompetition with regularization
1. less pressure to regularize if in alarge gang
2. hyperarticulation of high frequencywords is a consequence ofresisting regularization
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
15
ConclusionsMorphology plays a significant
role in acoustic realization.
Lexical frequency and paradigmaticalternation also play a role in
acoustic realization, though theireffects are in surprising directions.
Introduction
Speech Variation
Irreg Eng Verbs
Reduction Theories
Methodology
Predictors
Results
Morphology
Frequency
Gang Size
Imageability
Discussion
Interpretation
Conclusions
16
ConclusionsFurther Investigations:
regular verbsvowel qualityprosodic featuresprobabilistic syntactic factorssociolinguistic featuresother measurements of vowel
reduction
Irregulars
Regulars
Irregulars & Regulars
Vowel Quality
Vowel Quality
45961929tense3301749lax
presentpast
Methodology: Predictorsindependent variable: imageability
Prado & Ullman (2009) found averb’s imageability to co-determineits naming latency
mean ratings of imageability weregathered through a short study
Methodology: Predictorsstudy to gather imageability ratings:
112 participants87 female, 25 males17-63 years (71.4% 18-21 years)native English speakers
rate 171 verbs (48 irregular, 123regular) on how readily the verbproduces a mental image
1 5least imageable most imageable
Methodology: CorpusBuckeye Corpus of Spoken English
300,000 words
40 speakers20 male/20 female20 < 30 years/ 20 > 40 yearsresidents of Columbus, OH
speech from informal interviews
ReferencesAylett, M. and Turk, A. (2004) The Smooth Signal Redundancy
Hypothesis: A Functional Explanation for Relationshipsbetween Redundancy, Prosodic Prominence, and DurationSpontaneous Speech. Language and Speech 47:1, 31-56.
Kuperman, V., Pluymaekers, M., Ernestus, M. and Baayen, H.(2007) Morphological predictability and acoustic duration ofinterfixes in Dutch compounds. JASA, 121:4, 2261-2271.
Pitt, M.A., Dilley, L., Johnson, K., Kiesling, S., Raymond, W.,Hume, E. and Fosler-Lussier, E. (2007) Buckeye Corpus ofConversational Speech (2nd release)[www.buckeyecorpus.osu.edu] Columbus, OH: Department ofPsychology, Ohio State University (Distributor).
Prado, E.L and Ullman, M.T. (2009) Can Imageability Help UsDraw the Line between Storage and Composition? JEP: LMC,35:4, 849-866.
Scarborough, R.A. (2004) Coarticulation and the Structure of theLexicon. Doctoral dissertation, University of California, LosAngeles.
