Please enable JavaScript.

Coggle requires JavaScript to display documents.

language processing (the common experimental techniques involved in the…

- - - - longer time = more processing
    - - more common = faster access
  - - - if they are under the same kind, it is easy to make decision
  - - - sounding out the word
      - Suggests two parallel processes
- - - - the process of speech chain
        
        comprehension of speech signal
        
        how do we recognize particular speech sound when they occur in different contexts and when spoken by different people
        
        Normalisation
        
        how do we find word in the speech streams?
        
        Segmentation
        
        Word recognition (lexical access)
        
        how do we carve up the continuous speech signal into meaningful unit?
        
        Perception depends on what you know
        
        perception of the speech signal
        
        if u cant receive it u cannot understand it
        
        based on hearing
  - - - Begins with small units (phones)
        
        agent verb theme
        subject verb object
        [NP] [VP[NP]]
        noun verb noun
        I love linguistics
        aɪ lʌv lɪŋˈgwɪstɪks
        aɪlʌvlɪŋgwɪstɪks
      - Proceed step by step to increasingly larger units (words and phrases)
      - Finish when entire input is processed (often ending in a complete sentence and semantic interpretation)
    - - Analyse into successively smaller constituents (phrases, words, morphemes , etc.)
      - Compare with sensory or acoustic data to validate the analysis
      - Begin with assumption that a large syntactic unit (sentence) is present
      - If analysis not validated, procedure returns to previously validated point and resumes.
      - the evidence
        
        When shadowing, words often anticipated
        
        how u can anticipate when somebody speaks? "expectation"
        
        When phonemes deleted, listener still hears them
        
        spoken word identified even with background noise interfence
        
        like in crowded party.. u still understand the speakers because u have "the expectation"
        
        Ambiguous words segmented correctly
        
        we know what we expect
      - why it is
        
        if u compute the meaning, you can guess ahead
        
        Knowledge of syntactic and semantic relations allows you to limit future possibilities
        
        Top-down processing can override bottom-up processing
        
        basic answer: it is more efficient, ur brain is lazy, so u assume it instead of going several stages to understand the speaker
- - - - computational
        
        semantics
        
        Computational morphology
        
        pragmatics
        
        phonetics and phonology
        
        syntax
      - computational processes
        
        Speech synthesis
        
        format vs concatenate
        
        Identify roots & affixes
        
        irregular forms
        
        Parsing
        
        ambiguity
        
        Speech understanding
        
        lexical limit
        
        Semantic network
        
        logical limit
        
        Speech recognition
        
        noise , variation
        
        context
        
        world knowledge
    - - parallel processing
      - statistical analysis
      - machine learning
      - non-human systems
    - - concordances
      - collocations
      - frquency and speech analysis (adaaa data) 47
      - computational lexicography
      - information retrieval and summarisation
      - spellcheckers
      - machine translation
    - - trademarks
      - intepreting legal terms
      - speaker identification