UNIT II: Memory & Language

1)Differentiate between episodic and semantic memory with examples.

Difference between Episodic and Semantic Memory
Episodic and semantic memory are two types of long-term declarative memory—they both store
facts and information but differ in content and context.

Aspect
Episodic Memory
Semantic
Memory
Definition Memory of personal experiences Memory of general facts, concepts, and
and specific events knowledge
Content "What", "where", and "when" of Meaning, facts, ideas, and rules
events
Time and Place Always
and place
linked to a specific time Not tied to a specific event or moment
Conscious Recall Often involves vivid recollection
("mental time travel") Retrieved without needing context
Emotional Usually emotional and personal Typically neutral and impersonal
Involvement
Example "I had biryani with my friends last "Biryani is a rice-based Indian dish."
Friday night."
Brain Areas Hippocampus and medial Lateral temporal lobes and prefrontal
Involved temporal lobe cortex

Examples:
Episodic Memory:
• Your first day at college.
• A vacation trip to Goa in 2022.
• What you wore to your graduation ceremony.
Semantic Memory:
• Delhi is the capital of India.
1
 • Water boils at 100°C.
• The formula for calculating area of a circle is πr².
2)Explain computational modeling of human memory.
Computational Modeling of Human Memory
Computational modeling of human memory refers to the use of computer-based algorithms and
systems to simulate how humans encode, store, and retrieve information. These models aim to
replicate the structure and functioning of human memory for applications in cognitive science, AI, and
cognitive computing.
Objectives of Computational Memory Models:
• Understand how memory processes work in the brain
• Predict memory behavior under different conditions
• Design artificial systems that mimic human memory
• Improve machine learning models by integrating cognitive principles
Types of Human Memory Modeled:
1. Sensory Memory – Stores information for a few milliseconds (e.g., iconic, echoic memory)
2. Short-Term Memory (STM) / Working Memory – Holds limited information briefly for
manipulation
3. Long-Term Memory (LTM)
• Episodic Memory (events and experiences)
• Semantic Memory (facts and meanings)
• Procedural Memory (skills and tasks)
Key Computational Memory Models:
1. Atkinson-Shiffrin Model (Multi-Store Model)
• Memory is modeled in three stages: Sensory → Short-Term → Long-Term
• Includes processes like attention, rehearsal, and encoding
2. ACT-R (Adaptive Control of Thought – Rational)
• A cognitive architecture that models declarative memory (facts) and procedural memory
(skills)
• Items in memory are retrieved based on activation values that reflect recency and frequency
3. SOAR Cognitive Architecture
• Integrates learning, memory, and decision-making
• Focuses on problem-solving and chunking (grouping of information)
4. Connectionist Models (Neural Networks)
• Mimic the brain using nodes (neurons) and weights (synaptic strength)
• Represent distributed memory—information is stored in the pattern of connections
• Good at generalizing from incomplete or noisy inputs
5. Bayesian Memory Models
• Model memory as probabilistic inference
• Capture how people update memories and predict future events using past data

2
Processes Modeled:
Process
Modeling Approach
Encoding Attention filters and feature extraction mechanisms
Storage Long-term potentiation or weight adjustments in networks
Retrieval Activation-based search (ACT-R) or pattern completion
Forgetting Decay, interference, or noise-based degradation

Applications:
• AI systems that mimic human-like memory and learning
• Educational tools to enhance learning and recall
• Robotics that adapt to experience like humans
• Medical simulations for disorders like Alzheimer’s
3)Discuss the role of language in cognitive models.
Role of Language in Cognitive Models
Language plays a central role in cognitive models because it is one of the most complex and
essential cognitive functions of the human mind. It is not only a medium of communication but
also a tool for thought, reasoning, memory, and learning.
1. Language as a Cognitive Function
• Language is tightly integrated with other cognitive processes such as:
◦ Perception
◦ Memory (semantic and episodic)
◦ Attention
◦ Problem-solving
◦ Learning and decision-making
• Cognitive models often treat language as a symbolic system for encoding, storing, and retrieving
information.
2. Key Areas Where Language Appears in Cognitive Models

3
 Aspect
Role of Language
Representation of Words and grammar represent concepts, categories, and relationships.
Knowledge
Memory Models Language is used to encode and retrieve semantic/episodic memories.
Reasoning & Logical reasoning often relies on structured language patterns.
Inference
Learning Models Language provides input (e.g., instructions, feedback) and output (e.g.,
explanations).
Interaction with Language is used to describe, explain, and manipulate the world
Environment cognitively.

3. Examples in Cognitive Modeling:

a. Symbolic Models (e.g., ACT-R, SOAR)
• Use language-like symbols to represent mental constructs.
• ACT-R includes modules that simulate language comprehension and production.
b. Connectionist Models (Neural Networks)
• Capture language patterns such as syntax and semantics through layers and weighted
connections.
• Used in speech recognition, machine translation, and semantic analysis.
c. Bayesian Models
• Handle language ambiguity and learning by updating probability estimates based on context
and prior knowledge.
4. Language and Thought
• Language influences how we think (linguistic relativity hypothesis).
• It helps in abstract thinking, planning, and self-reflection.
• Inner speech (talking to oneself) is an internal form of cognitive processing in many models.
5. Application in Cognitive Systems
• Chatbots & Virtual Assistants: Model dialogue using natural language processing.
• Language Learning Systems: Simulate how humans acquire and use language.
• Human-Robot Interaction: Robots interpret and respond to commands using language models.
4)What is psycholinguistics? How is it modeled computationally?
What is Psycholinguistics?

4
Psycholinguistics is the interdisciplinary study of how humans acquire, comprehend, produce, and
represent language in the mind. It combines concepts from psychology, linguistics, neuroscience,
and cognitive science to understand the mental processes behind language use.
Core Areas of Psycholinguistics:
1. Language Acquisition – How humans, especially children, learn language.
2. Language Comprehension – How we understand spoken and written language.
3. Language Production – How we form and speak or write sentences.
4. Language Representation – How language is stored and organized in the brain.
5. Lexical Access – How we retrieve words from memory during communication.
Computational Modeling of Psycholinguistics
Computational models simulate the cognitive processes involved in language. They help test theories
and build systems that mimic human language abilities.
Key Computational Models in Psycholinguistics:
1. Connectionist Models (Neural Networks)
• Mimic the brain's structure with nodes and connections.
• Learn language patterns from exposure to large datasets.
• Used for phonological processing, morphological analysis, and semantic comprehension.
Example: Parallel Distributed Processing (PDP) models simulate how children might acquire past
tense forms like “go → went.”

2. Symbolic Models
• Use rules and symbolic structures (syntax trees, grammar rules).
• Model how people parse sentences or form grammatically correct phrases.
• Focus on formal grammar and syntax-semantic mapping.
Example: ACT-R (Adaptive Control of Thought–Rational) simulates sentence comprehension and
production.

3. Bayesian Models
• Treat language processing as probabilistic inference.
• Useful for modeling ambiguity resolution, word prediction, and semantic inference.
• Help explain how humans anticipate words in context and disambiguate meanings.
4. Cognitive Architectures (e.g., SOAR, OpenCog)
• Integrate language as one module among many (e.g., memory, decision-making).
• Simulate natural conversations and learning from feedback.
Applications of Computational Psycholinguistics:
• Natural Language Processing (NLP): Text understanding, translation, summarization.
• Speech Recognition & Generation: Mimicking how humans understand and produce speech.
5
 • Language Disorders: Modeling aphasia or dyslexia for diagnostic tools.
• Education: Language tutoring systems that adapt to cognitive models of learners.
5)Describe a model that simulates memory retrieval in cognitive systems.
Model that Simulates Memory Retrieval in Cognitive Systems: ACT-R
One of the most well-known cognitive architectures that simulate memory retrieval is the ACT-R
model (Adaptive Control of Thought—Rational), developed by John R. Anderson. It is a modular
cognitive architecture that aims to simulate human thought processes, including memory retrieval,
decision-making, language processing, and learning.
Key Concepts in ACT-R for Memory Retrieval
ACT-R divides memory into two main components:
1. Declarative Memory (What you know):
• Stores facts and information (e.g., "Paris is the capital of France").
• Each piece of knowledge is stored as a chunk.
2. Procedural Memory (How you act):
• Stores rules for decision-making (if–then rules).
Memory Retrieval Process in ACT-R:
1. Activation of Chunks
• Each chunk has an activation value that determines how easily it can be retrieved.
• Activation depends on:
◦ Frequency of use (how often the chunk has been retrieved).
◦ Recency of use (how recently the chunk was retrieved).
◦ Contextual relevance (how well it matches the current context).
2. Spreading Activation
• When a question or problem is presented, related concepts (chunks) spread activation to
associated chunks.
• This simulates associative recall in human memory.
3. Retrieval Threshold
• Only chunks with activation above a certain threshold are retrieved.
• If no chunk exceeds the threshold, retrieval fails or takes longer.
4. Retrieval Time
• The higher the activation, the faster the retrieval.
• This models reaction time in human memory tasks.
Formula (Simplified)
Activation = Base-Level Activation + Associative Activation

Where:
• Base-Level Activation is based on frequency and recency of use.
• Associative Activation is influenced by relevance to current context.

6
Example Simulation:
Let’s say the system is asked: "What is the capital of France?"
• The input cue triggers related memory chunks: "France", "Capital", etc.
• These cues spread activation to connected chunks like "Paris", "Berlin", etc.
• If “Paris” has the highest activation, it is retrieved.
• Response: "Paris" is returned, simulating human recall.
Applications of ACT-R Memory Model
• Human-Computer Interaction: Predicting how users remember and retrieve information.
• Educational Systems: Simulating student memory for adaptive learning.
• Cognitive Psychology: Modeling memory tasks like recall, recognition, and forgetting.
• AI Systems: Embedding human-like memory behavior in intelligent agents.