Word Processing Flashcards
LOGOGEN (1st gen.)
- Bottom-up, AI-style account
- Word are represented by LOGOGENS: evidence collecting device that fires (via context and auditory signals) when it accumulates evidence drives it above the threshold and information about the word is accessed
- Each word is a separate “entry”
FOBS (frequency ordered serial bin, 1st gen.)
- Bottom-up
- Auditory cues drive long-term memory search; self-terminates once you find the word you need
- Search through “bins” organized according to root morphemes
TRACE (2nd gen.)
- Interactive, bottom-up input and top-down feedback –> cascaded activation
- Good for word superiority effect and degraded input
- Predicts increased competition with increased cohort size
COHORT (2nd gen.)
- Developed to explain lexical access for spoken words
- Processes: Activation (bottom-up, autonomous), selection (sort through activated forms of a word), integration (evaluated for fit based on context)
- Incremental processing: recognition point at ~200ms into each word
- Cohort size does not predict increased competition with increased cohort size
SRN (Simple Recurrent Network, 3rd gen.)
- Adding context units to TRACE
- Responds to current input units as well as context units
DCM (distributed cohort, 3rd gen.)
- Input = phonetic features
- Uses output of hidden units to activate semantic and phonological units
- Affected by bottom-up (auditory) input
Localizationist Hypothesis
Word meanings are represented based on category-specific semantic deficits
Distributed Representations
- Word meanings represented as coordinated patterns of neural activity
- Correlated features approach: knowledge about objects spreads through the brain –> concepts have different kinds of features (correlated vs. distinctive)
Lexical semantics (meaning)
- Sense (dictionary definition of the target object) vs. reference (words uses to refer to the target object)
Definition Hypothesis
- Word meaning = list of necessary/core features of a word
- Problems: some words do not have consistent features, word meanings change across contexts
Word-Association/Semantic Network Approach (alternative to the Definition Hypothesis)
- Word meaning = whatever comes to mind when you think of a word
- Collections of associated concepts (a set of links and nodes)
- Conserves memory resources via spreading activation when one node of a Semantic Network is activated –> automatic and decreases further it travels in network (mediated priming)
- Allows semantic priming to occur
Associative vs. semantic priming
- Driven by different mechanisms (e.g. each break down separately in Alzheimer’s disease
- Co-occurring words become connected in semantic network
- Associative: easier for priming to occur, causes faster responses (decreases N400 semantic reaction time effect) –> associative relations are encoded in lexical/semantic representation of word
- Semantic: slower responses
Associative Word Meaning Models
2 ways a word’s meaning is determined:
1. Hyperspace Analog to Language (HAL): target word’s meaning depends on another word it appears with
2. Latent Semantic Analysis (LSA): relationship between target word’s meaning and context
- Both predict similarity judgments, vocabulary developments, text-quality –> describing mapping between words
Embodied Semantics Model
- Words tied to representations outside the linguistic system, built using the perceptual system
- Determined by interaction of perceptual abilities –> affordances
- Indexical Hypothesis (steps to determine word meaning): (1) Word must be died to actual object, (2) perceptual signals provide affordances of a word, (3) affordances provided by different words are combined
Word-Action Compatibility Effects (evidence for semantic/perceptual/motor link)
- Reaction time faster when motor response is in same direction as action in the sentence
- Precision vs. power grip
- Word-action interference & facilitation effects depend on timing of stimulus presented
- Neuroimaging data: motor strip activated when word relates to body parts, faster lexical access to certain words when TMS stimulation is applied
- Problems: motor stimulation may be optional and may not apply to all kinds of language
