Exam 3 Semantic Networks and Connectionism

Exam 3 Semantic Networks and Connectionism - Semantic...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Semantic Networks & Connectionism Semantic Network Approach Categories and concepts are arranged in networks that represent the way categories and concepts are organized in the mind. – Collins and Quillian (1969) Collins & Quillian (1969) A node represents a category/concept concepts are connected by links and are arranged hierarchically Related Relevant principles: – Properties associated with each concept are indicated at the _____. Nodes are connected by links and represents one thing, connected to you can access different properties. -Properties associated with each concept are indicated at the nodes -Cognitive economy- shared properties are only stored at higher level nodes -Exceptions are stored at lower level nodes – ___________________: shared properties are only stored at higher-level nodes – _________ are stored at lower-level nodes Semantic Network Model: To get information about canary you start at canary and move up the levels. Captures that we think things specific to canaries but also know more about it because we know things that are connected to it. Ex: Know it can fly because it is in the bird property 1 Semantic Networks & Connectionism Collins & Quillian (1969) Asked participants to verify sentences: – Manipulated distance between concept and a property “A “A Measured sentence verification times between concept and property. Predicted that the further distance you to travel or more levels you have to cross, the longer the response time. Would take longer to identify the canary as an animal than a bird canary is a bird” canary is an animal” Collins & Quillian (1969) Results: – The time to retrieve information about a concept was determined by the distance that must be traveled through the network. Collins & Quillian (1969) Activation spreads across any link connected to an activated node – Concepts that receive this activation become “primed” and are accessed more easily from memory – Evidence for spreading activation: Any time the name of a concept is presented the node representing that concept is activation and spreads to any other node to which it is connected. Priming occurs from spreading activation from concept to concept & Schvaneveldt (1971) – semantic priming in lexical decision tasks Demo Meyer 2 Semantic Networks & Connectionism Lexical Decision Meyer & Schvaneveldt (1971) Bread and butter are strongly associated. Manipulated associations between wordseither strongly associated or no associations – “YES” if both are words, “NO” otherwise BREAD BREAD BUTTER DOCTOR Results: – Faster to say “yes” to BREAD-BUTTER than to BREAD-DOCTOR = semantic priming Problems with Collins & Quillian model Some sentence verifications were inconsistent with model’s predictions: – The ________________ occurs, even when comparing properties that are the same # of links from the concept: The typicality effect occurs, even when comparing properties that are the same # of links from concept: is faster slower than “A canary is a bird” is ________ than “an ostrich is a bird” – Some closer nodes took longer to verify: “A salmon is an animal” is _______ than “a salmon is a living thing” Collins & Loftus (1975) No hierarchical structure – based instead on person’s experience Length of link determines relatedness – Concepts that are more closely related are connected by short links 3 Semantic Networks & Connectionism Semantic Networks in the Real World http://www.visualthesaurus.com http://www.aclib.us/ The Connectionist Approach Creating computer models for representing concepts and their properties based on characteristics of the brain – PDP models PDP Models Relevant terms: – Unit: metaphor for neurons Input units: Activated by stimuli in the environment Hidden units: receive signals from input units Output units: contain the final output of the network 4 Semantic Networks & Connectionism PDP Models Relevant terms: – Connection weight: metaphor for synaptic transmission determines the degree to which signals sent from one unit either increase or decrease the activity of the next unit How are concepts represented? A concept is represented by the pattern of activity distributed throughout a set of nodes – Connections weights are adjusted through learning PDP Models Support: – Operation of connectionist networks is not totally disrupted by damage Graceful degradation – Connectionist networks can explain generalization of learning 5 Semantic Networks & Connectionism Categories in the Brain Different areas of the brain may be specialized to process information about different categories – e.g., fusiform face area (FFA) – BUT … Categories in the Brain Categories are represented by distributed activity: – Category-specific neurons in monkeys (in book) – Patients with category-specific knowledge impairment – Brain imaging studies 6 ...
View Full Document

Ask a homework question - tutors are online