MIT 24 910 - Effects of the lexicon and context on speech production

Unformatted text preview:

Effects of the lexicon and context on speech productionWord recognitionEffects of lexical statistics on productionListener-Oriented SpeakersProduction and perceptionAre speakers altruistic or selfish?Frequency and Neighborhood Density: Munson & Solomon (2004)Experiment 1Experiment 2Selfish speakers and neighborhood densitySelfish speakers and neighborhood densityLexical access and neighborhood densityMunson (2004) resultsContextual predictability and speech productionContextual predictability and speech productionRepetition and speech productionRepetition and speech productionRepetition and speech productionSelf-centered behavior in production of repetitionsSelf-centered behavior in production of repetitionsSelf-centered behavior in production of repetitionsSelf-centered behavior in production of repetitionsListener-oriented behavior in speech productionReferencesReferencesReferencesReferencesA Bayesian model of the listener - context effectsMIT OpenCourseWare http://ocw.mit.edu24.910 Topics in Linguistic Theory: Laboratory PhonologySpring 2007 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.Effects of the lexicon and context on speech productionWord recognitionThe speed an accuracy of word recognition depends on:• Word frequency• Neighborhood density– and frequency of neighbors• Contextual predictability• Speech production is also affected by these factors.Effects of lexical statistics on production• Wright (2004) found that neighborhood density/relative frequency affects pronunciation of isolated words.• ‘Hard words’ - low frequency, high neighborhood density• ‘Easy words’ - high frequency, low neighborhood density• Vowels in hard words are more dispersed from each other in F1*F2 space than vowels in easy words.Figure by MIT OpenCourseWare.F2 Bark15 14 13 12 11 10 9 823456789F1 Barki+iuI+e+eu+Iaa++c+c+ooListener-Oriented Speakers• It has been hypothesized that the production and perception effects are linked:– Words that are more difficult to recognize are pronounced more clearly.Production and perceptionBroad outlines of an explanatory model of the effects of frequency, neighborhood density etc on production (Wright 2004, Scarborough 2004, 2006):• Speaker wishes to be understood.• Speaker wishes to minimize the effort involved in speech production.• Reduced effort tends to lead to reduced identifiability of words.• Optimal strategy: reduce effort more where clarity is less important, i.e. where top-down evidence makes it easier for a listener to identify a word - Hyper- & Hypoarticulation theory (Lindblom 1990).Are speakers altruistic or selfish?• Implication of H&H account of pronunciation variability: Speakers estimate listener difficulty moment to moment and adjust clarity of speech accordingly.– Speaker has a model of the listener (altruistic speaker).• Alternative line of analysis: Speakers are selfish.– Speakers do not track listener difficulty.– Pronunciation variation is related to speaker difficulty with lexical access for production:• Slower lexical access results in clearer speech.– To the extent that ease of lexical access is similar for speaker and hearer, similar results are predicted (but there are differences).– But why does slow lexical access result in clearer speech?Frequency and Neighborhood Density:Munson & Solomon (2004)• Point out that Wright (a) confounded frequency and neighborhood density, (b) didn’t measure duration, so we can’t be sure if neighborhood density affects vowel formants directly or via vowel duration.Figure by MIT OpenCourseWare.Vowel Lexically easy Lexically difficultaaaaæææIIIiiooujobshopwashwatchgaspathgiveshipthingpeaceteethbothvotefoodjackcodcotknobwadhackpathickkinkitbeadweedgoatmoathoophashExperiment 1• Words read in isolation by 10 subjects.• Vowel space is expanded in Hard words (mean Euclidian distance of vowels from the mean F1, F2 of all vowels).• Vowels were shorter in hard words (222 ms vs. 232 ms).Image removed due to copyright restrictions.Please see Figures 1 and 2 in Munson, B., and N. P. Solomon. "The Effect of Phonological Neighborhood Density on Vowel Articulation."Journal of Speech, Language, and Hearing Research 47 (2004): 1048-1058.Experiment 2• 4 classes of words, crossing:– high vs. low frequency– high vs. low neighborhood density• 20 words /class• Duration:– High frequency words had shorter vowels (205 ms vs. 211 ms)– No effect of neighborhood density.• Vowel space expansion:– Less expanded in high frequency words.– More expanded in high neighborhood density words.– No significant interaction.Image removed due to copyright restrictions.Please see Figures 3 in Munson, B., and N. P. Solomon."The Effect of Phonological Neighborhood Density onVowel Articulation." Journal of Speech, Language, andHearing Research 47 (2004): 1048-1058.Selfish speakers and neighborhood density• The H&H account: Words from dense neighborhoods/low frequency words are pronounced more clearly because the speaker knows they are likely to be more difficult for listenersto recognize.• Speaker-oriented account (e.g. Pierrehumbert 2002): – Speakers have to perform lexical access in speech production.– Hypothesize that high neighborhood density impedes lexical access in production.– Slower lexical access results in clearer pronunciation.Selfish speakers and neighborhood density• But high neighborhood density can actually speed lexical access in production: Pictures are named more quickly when their names are in dense neighborhoods (Vitevitch 2002).• Lexical access in production starts from meaning, so there is no problem of competition based on phonetic similarity.• Vitevitch offers two explanations for the facilitatory effect of dense neighborhoods:• In an ‘interactive activation model’: activation spreads between phonetically similar words. In dense neighborhoods more activation ‘reverberates’ back to the target word.• OR: words in dense neighborhoods generally involve more common sound sequences - perhaps the motor plans for frequent sound sequences are easier to access/assemble.Lexical access and neighborhood density• Munson (2004) directly tested the ‘lexical access’ account of neighborhood density effects:• Four classes of words crossing:– High vs. low frequency– High vs. low neighborhood density• Subjects read words in two conditions:– Read word


View Full Document

MIT 24 910 - Effects of the lexicon and context on speech production

Download Effects of the lexicon and context on speech production
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Effects of the lexicon and context on speech production and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Effects of the lexicon and context on speech production 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?