What are the 4 elements of speech recognition?
Segmentation
Lexical selection
Access to meaning
Context effects
What is speech processing?
progressively extracting invariant and discrete representations from a variable, continuous input
What are some challenges of speech processing?
- Continuous
- Distributed in time
- Fast-fading
- Variable
What is the word segmentation problem?
Female and male talker amplitude, frequency and time should be identical
Brain has to understand the same word but with different properties
Bottom-up process
What is the metrical segmentation strategy?
Culter & Norris (1988)
In English:
- stressed syllables are likely onset of words
- continuous speech is segmented at stressed syllables
What evidence is there in support of the metrical segmentation strategy?
74% of stressed syllables in english corresponds to sole or initial syllable of a content word
(Culter & Carter, 1987)
What is a weakness of the MSS?
Listeners still need other sources of information to segment successfully
Its language specfic
What is a strength of the MSS?
It solves the child’s paradox (how children can segment a word if they don’t know the word)
What are Matty et al’s (2005) three tiers of segmentation?
Tier one - Lexical
- sentential context + lexical knowledge
Tier two- Segmental
- phonotactics + acousitc-phonetics
Tier three- Metrical prosody
- Word stresss
What is lexical selection?
Input from a segmented stream
The search process determines the best fit between the input and abstract lexical representations
its fast and can finish before the word has been fully said
How quick can words be recognised?
Within 175-200ms of their onset
What evidence from shadowing is there to support lexical selection?
Marslen-Wilson (1975)
- Participants listen to a sentence and repeat aloud what they hear
Results:
- Fast at repeating and before the end of target words
- Correct misspelled words when repeating them
What evidence from gating is there to support lexical selection?
Tyler & Wessels (1983)
- Participants are given a word to listen
- Word is chopped into different fragments (gates) of different durations
- Task is to guess the word
Results:
- Listeners consider multiple word candidates that are consistent with the incoming speech
What is the cohort model?
Marslen-Wlson & Welsh (1978)
Access- actvation of initial set of candidates based on word-initial cohort
Selection- words that mismatch the incoming signal removed from the cohort
Integration- their syntactic and semantic properties integrated with context
What is the recognition point when understanding words?
A word is recognised at the point where it is the only world still consistent with the input
What is a strength of lexical selection?
Its a highly efficient system: maximally effective use of incoming signals and the word will be recognised as soon as info is available
How do we identify meaning in speech?
Context
What did Marslen-Wilson, Brown & Tyler (1988) find about effects in monitoring?
Multiple types of contextual information are integrated during spoken word recognition
What brain areas are activated in understanding speech?
Bilateral activity in Heschl’s gyrus, STG and MTG for mapping sound to meaning
Left-lateralised activation in dorsal stream
What is the evolutionary explanation for brain activation areas?
Gil-da-costa et al (2006)
- Non-human primates also communicate by exchanging meaningful calls
- triggers bilateral activity in the brain of a macaque
- continuity of the bilateral system supports mapping from sound to meaning
What are syntactic rules in sentence processing?
rules that govern how words can be combined
- the allow permissable sentences and rule out illegal sentences
What is the garden-path model?
Frazier & Rayner (1982)
Initial parsing is syntactic, meaning is not involved
What are constraint-based theories?
MacDonald et al (1994)
- intial interpretation depends on all available sources of information
What are unrestricted race-model theories?
Van Gompel et al (2000)
- all sources of information used to identify a syntactic structure
