Upload
sage-turkel
View
214
Download
0
Embed Size (px)
Citation preview
Application of Knowledge Based Systems in Education
Dr Priti Srinivas SajjaDepartment of Computer ScienceSardar Patel UniversityVallabh Vidyanagar, Gujarat
Dr. Priti Srinivas Sajja6-7 February, 2009
Introduction and Contact Information• Speaker: Dr Priti Srinivas Sajja• Communication:
• Email : [email protected]• Mobile : 9824926020• URL : priti.sajja.info
• Academic qualifications : Ph. D in Computer Science• Thesis title: Knowledge-Based Systems for Socio-Economic Rural Development
• Subject area of specialization : Knowledge Based Systems
• Publications : 84 in International and National Books, Chapters and Papers
• Academic position : Associate Professor Department of Computer Science Sardar Patel University Vallabh Vidyanagar 388120
Dr. Priti Srinivas Sajja6-7 February, 2009
Lecture Plan
Knowledge Based Systems• Introduction to Knowledge Based Systems• Categories and Structures of KBS• Applications of KBS
KBS in Education• Symbolic Approach
• Parichay: Adult Literacy System for Leaning Gujarati Language• Multi Agent KBS fro e-Learning Accessing Distributed Databases on Grid • Multi-tier KBS Accessing LOR through Fuzzy XML
• Connectionist Approach • Symbolic verses Connectionist Approach• Soft Computing • Neuro-fuzzy System for Course Selection • Fuzzy-genetic System for Evolving Rule Bases to Measure Multiple
Intelligence • Acknowledgement, References and Contact
Dr. Priti Srinivas Sajja6-7 February, 2009
Artificial Intelligence
• “Artificial Intelligence(AI) is the study of how to make computers do things at which, at the moment, people are better”
• -Elaine Rich, Artificial Intelligence, Mcgraw Hill
Publications, 1986
Dr. Priti Srinivas Sajja6-7 February, 2009
Knowledge Based Systems
K
Knowledge Based Systems (KBS) are
Productive Artificial Intelligence Tools working in a narrow domain.
Dr. Priti Srinivas Sajja6-7 February, 2009
How Knowledge is organized?
Volume Complexity & Sophistication
Wisdom(experience)
Knowledge(synthesis)
Information(analysis)
Data
Data PyramidSource: Tuthill & Leavy, modified
Dr. Priti Srinivas Sajja6-7 February, 2009
Data
• Raw Observation• Stand alone numbers and symbols that do possess little value• Data are symbols that represent properties of objects, events and
their environments. • ANYTHING numbers, words, sentences, records, assumptions• Example BMI, 10, (smith, 50)
Dr. Priti Srinivas Sajja6-7 February, 2009
Information
• Processed Data• Smith weight is 50 Kg. • Information has usually got some meaning and
purpose
Dr. Priti Srinivas Sajja6-7 February, 2009
Knowledge
• Information can be processed further with the operations such as • Synthesis• Filtering• Comparing etc.
to get generalized knowledge
Dr. Priti Srinivas Sajja6-7 February, 2009
Wisdom
• Knowledge of concepts and models lead to higher level of knowledge called wisdom.
• One needs to apply morals, principles and expertise to gain and utilize wisdom.
• This takes time and requires a kind of maturity that comes with the age and experience.
Dr. Priti Srinivas Sajja6-7 February, 2009
Data Pyramid and Computer Based Systems
Basic transactions by operational staff using data processing
Middle management uses reports/info. generated through analysis and acts accordingly
Higher management generates knowledge By Synthesizing information
Strategy makers apply morals, principles and experience for generating policies
Wisdom (Experience)
Knowledge (Synthesis)
Information (Analysis)
Data (Raw Observations Processing)
Volume Sophistication and complexity
TPS
DSS, MIS
KBS
WBS
IS
Dr. Priti Srinivas Sajja6-7 February, 2009
Computer Based Systems Tree
MIS
DSS
EES*
ESS
ES
EIS
TPS
OAS
Figure 1.8: CBIS Tree (Sajja & Patel 1995)
1990
1970
1950
Hardware Base/Technology
Users Requirement
IS
Intelligent Systems: 21st Century Challenge
EES:Executive Expert System, which is hybridization of Expert System , Executive Information System and Decision Support System.
S/W Resources
Dr. Priti Srinivas Sajja6-7 February, 2009
Structure of KBS
Knowledge Base
Inference Engine
User Interface
Explanation/ Reasoning
Self Learning
Provides explanation
and reasoning facilitates
Knowledge base is a repository of domain knowledge and meta
knowledge. Inference Engine is a software
program, which infers the knowledge available in the
knowledge base
Friendly interface to
users working in their native language
Enriches the system with self learning capabilities
Dr. Priti Srinivas Sajja6-7 February, 2009
Categories of the KBS
According to Tuthill & Levy (1991), KBS can be mainly classified into 5 types:
Expert SystemsThe Expert Systems (ES) are the most popular and historically pioneer knowledge based systems, which replace one/more experts for problem solving.
Linked Systems The Hypermedia systems like hyper-text, hyper-audio, hyper-video are considered as linked knowledge based systems.
CASE Based SystemsThese systems guide in information/intelligent systems’ development for better quality and effectiveness.
Database in conjunction with an Intelligent User InterfaceAn intelligent user interface can enhance the use of the content available in the traditional format.
Intelligent Tutoring SystemsThe knowledge based systems are also used to train and guide the different level of students, trainers and practitioners in specific area. These systems are also useful to evaluate students’ skills, prepare documentation of subject material and manage the question bank for the subject.
Dr. Priti Srinivas Sajja6-7 February, 2009
Major Advantages of KBS
• Increased effectiveness with efficiency
• Documentation of knowledge for future use
• Add powers of self learning
• Provides justifications for the decisions made
• Deals with partial and uncertain information
• Friendly interface
Dr. Priti Srinivas Sajja6-7 February, 2009
Difficulties with the KBS
• Nature of knowledge • Large Size of knowledge base• Slow Learning and Execution • Little methodological support from typical life cycle
models• Acquisition of knowledge• Representation of knowledge
Dr. Priti Srinivas Sajja6-7 February, 2009
KBS Applications
HealthDevelopment
PhysicalDevelopment
EconomicalDevelopment
SocialDevelopment
NR
HRLA
NR: Natural ResourcesHR: Human ResourcesLA: Live stock and Agricultural Resources
Physical CommunicationPlanning & AdministrationForestry, Energy, Agriculture etc.
HealthNutrition, SanitationCommunity Health etc.
EconomicalSmall Scale IndustryAgri-Business & Co-operativeetc.
SocialEducation & TrainingSocial Awareness Programme etc.
Dr. Priti Srinivas Sajja6-7 February, 2009
Technology and Education
Technology Education
Technology helps in learning
Education helps in development of technology
Dr. Priti Srinivas Sajja6-7 February, 2009
Objectives of Educational SolutionDifferent Model like Class room education, Distance learning
and Virtual learning / E-Learning etc. have some common objectives as follows:
• Support learning objectives and goals• Facility to publish, update and access learning material and
announcements• Friendly interface for non-computer professionals and
students for communication• Evaluation of learners and feedback mechanism• Administrative and documentation support• Meets standards and security aspects
Dr. Priti Srinivas Sajja6-7 February, 2009
Content Service
Technology
· Information Retrieval· Assistance· Learning System
Management· Evaluation · Documentation etc.
· Accessibility (Internet)
· User friendliness· Security· Communication· Inference and self
learning etc.
· Domain knowledge
· Supporting databases and documents etc.
Subject Experts
Media developers, Editors, Instructors
Web Designers, Technical Experts
Dr. Priti Srinivas Sajja6-7 February, 2009
Symbolic KBS: Some Examples
Parichay: Adult Literacy System for Leaning Gujarati Language
This is a Single PC based system where knowledge based contains set of rules in if…then…else form.
This system has been developed as an agent to help adults to learn regional language, Gujarati.
Dr. Priti Srinivas Sajja6-7 February, 2009
Some results from ‘Parichay’The system gives training to adult users in multi media to speak and write Gujarati alphabets, words, sentences and numbers.
The package of ‘parichay’ is accommodated in CD with auto-run facility.
The touch screen facility helps even an illiterate person to identify icons and choose appropriate actions.
Dr. Priti Srinivas Sajja6-7 February, 2009
The frequent continuous development of a letter helps users to see the exact motion to write the letter.
At the end of the full letter generation, the picture representing use of the letter and pronunciation is represented to the user.
Dr. Priti Srinivas Sajja6-7 February, 2009
With a notepad facility given, user may practice any letter.
That letter written by the user is matched with the correct letter by measuring shapes and angles in terms of percentages.
If the degree of matching is low then user may ask to redraw/rewrite the letter.
Dr. Priti Srinivas Sajja6-7 February, 2009
Limitations of the System
• ‘Parichay’ is limited to single user system only.
• It can be used only for elementary Gujarati learning (reading and writing) such as simple alphabets, numbers and sentences.
Dr. Priti Srinivas Sajja6-7 February, 2009
Multi Agent KBS for e-Learning Accessing
Distributed Databases on Grid
• e-Learning is supported by a knowledge based systems to improve quality.
• e-Learning emphasis on on-line delivery, management and learning of educational material.
• The following aspects are given importance for such learning:• Easy access of material in user friendly way• Anytime and anywhere learning• Better control and administration of material and users• Quick results and reporting
Dr. Priti Srinivas Sajja6-7 February, 2009
• System considers different databases which may be available in distributed fashion.
• At many places the learning material and supporting information like students, courses and infrastructure are available in electronic form.
• The idea is to access the available data sources in knowledge based way.
• e-Learning is a big job encompasses different activities hence multiple independent agents have been considered.
Dr. Priti Srinivas Sajja6-7 February, 2009
Architecture of the system
Users
Experts
Use
r Inte
rface
Ag
ent
Agents
Learning Mgt.
Drills and Quizzes
Explanation
Semantic Search
E-mail & Chat
Resource Management
Question/Answer
Tutorial Path
Documentation
Distrib
uted
D
atabases
Local Data-Bases
Resources
Knowledge Mgt.
Meta knowledge
Conceptual system
Content knowledge
Learner’s ontology
Documents
Knowledge Discovery
Knowledge Utilization
Knowledge Management
Dr. Priti Srinivas Sajja6-7 February, 2009
Communication between Agents
• Agents developed here are communicating with a tool named KQML.
• Knowledge based Query Management Language.
(register
: sender agent_Lerning_Mgt
: receiver agent_Tutorail-Path
: reply-with message
: language common_language
: ontology common_ontology
: content “content.data”
)
Action intended for the message
Agents name sharing message
Action intended for the message
Context-specific information describing the specifics of this message
Ontology of both the agents
Language of both agents
Dr. Priti Srinivas Sajja6-7 February, 2009
Some results form the System
Dr. Priti Srinivas Sajja6-7 February, 2009
Some results from the System
Dr. Priti Srinivas Sajja6-7 February, 2009
Some results from the System
Dr. Priti Srinivas Sajja6-7 February, 2009
Some results from the System
Dr. Priti Srinivas Sajja6-7 February, 2009
New architecture on Grid Environment
Future extension
Users
Experts
Use
r Inte
rface
Ag
ent
Agents
Learning Mgt.
Drills and Quizzes
Explanation
Semantic Search
E-mail & Chat
Resource Management
Question/Answer
Tutorial Path
Documentation
Intern
et
Grid
Mid
dlew
are Services
Resource Management
(Grid Resource Allocation
Protocol-GRAM)
and
Grid FTP Replica-LocationServices
Information Discovery Services
Security Services
Distributed databases
Middleware Services and
Protocols
Local Data-Bases
Resources
Knowledge Mgt.
Meta knowledge
Conceptual system
Content knowledge
Learner’s ontology
Documents
Knowledge Discovery
Knowledge UtilizationKnowledge Management
Dr. Priti Srinivas Sajja6-7 February, 2009
Towards reusable component library logic
• Learning Object Repository (LOR)
Dr. Priti Srinivas Sajja6-7 February, 2009
Multi-tier KBS Accessing LOR through Fuzzy XML
Dr. Priti Srinivas Sajja6-7 February, 2009
Neural Nets
Knowledge Representation
Fuzzy Logic
Trainability
Implicit, the system cannot be easily interpreted or modified (-)
Trains itself by learning from data sets (+++)
Explicit, verification and optimization easy and efficient (+++)
None, you have to define everything explicitly (-)
Get “best of both worlds”:Explicit Knowledge Representation from Fuzzy Logic with Training Algorithms from Neural Nets
Combining Neural and Fuzzy
Dr. Priti Srinivas Sajja6-7 February, 2009
Neuro-fuzzy System for Course Selection
• Critical decision + limited time period • Parents and students are not exposed to the
opportunities though educated• All alternatives are not available at one place• Continuously changing data• Changing job opportunities• Too many choices Vs. shortfall in specific stream
industry gap imbalance in trained personnel
Dr. Priti Srinivas Sajja6-7 February, 2009
Current scenario
• Available systems: • Local with limited scope,
• biased and
• manual systems are available
• Static information system
• Work on Database and explicit documentation required
• Lacks knowledge orientation
• No justification of the decisions
• No self learning about new opportunities and courses
‘Course Selector’, University of Edinburg, UK
‘Course Advisor Expert System’ is developed at the Griffith University
Dr. Priti Srinivas Sajja6-7 February, 2009
Requirements• Timely decision
• Uniform Information availability at one place
• Management of large amount of data
• Effective and knowledge oriented personalized decision support
• Justification (explanation and reasoning)
• Adaptive to new courses
• Friendly user interface working in natural fashion
Dr. Priti Srinivas Sajja6-7 February, 2009
Users
• Students• Parents• Institutes and Universities • Professional consultants, if allowed• Researchers and policy makers
Dr. Priti Srinivas Sajja6-7 February, 2009
Critical Parameter categories• Institute and course information:
• Institute name, registration number, preliminary information, courses, seats, reservation, placement, history etc.
• Users academic qualification/marks:• Name, location, degree/exam, marks, year, board etc.
• Users personal preferences:• Institute & course preference, hostel accommodation,
foreign chances etc.
• Family background:• Parents business, economical conditions etc.
Dr. Priti Srinivas Sajja6-7 February, 2009
Methodology
Fu
zzy Interface
Structure of the Neuro-fuzzy System
Fuzzy interface
Linguistic fuzzy
interface
Fuzzy rule base and membership
functions
Workspace
Crisp Normalized
values
Decision support
Users choice and needs
Decision support
Underlaying ANN
P1
P2
P3
P4
Implicit learning
& self learning by ANN
Friendly interface and
Explicit justification, documentatio
n
Dr. Priti Srinivas Sajja6-7 February, 2009
Students Information Collection:Name, Location, Score, Subject wise marks, Board Name, State information, etc.
Family Background Information: Economical conditions, parents profession etc.
Aptitude and Preference Seeking Questions: Choice of institute, course, homesickness, etc.
Institute list with Courses, seats, accredition, faculty, resources, history, placement, cut-off marks etc.
ANNNormalized Student Info + Reference Ids *
* generated from Institute +Courses + Scheme etc.
Input Layer
Hidden Layers
Output Layer
Array of alternatives in Sorted order with default three best suitable alternatives
Available in Knowledge Base
Collected from User(s) through Input Screens
Fully Connected Feed Forward Multi-Layer
Back-Propagation ANN
Can be changed according to
users demand
Users
PI & PF
Conversion into crisp normalized values by Fuzzy Interface
Fuzzy Interface
User
Dr. Priti Srinivas Sajja6-7 February, 2009
An Example Prototype
Elective Course Selection system:• Objective: To test feasibility of the proposed project
• Place: Department of Computer Science, S P University
• Tools: .Net 2005 + ANN simulator (JavaNNS)
• Training set: 100 records
• Users :Final Year MCA Students at the S P University (220 app.)
Dr. Priti Srinivas Sajja6-7 February, 2009
Interface Screen to collect training data
Dr. Priti Srinivas Sajja6-7 February, 2009
Fuzzification of the parameters resulting in normalized values…Linguistic Distance :[very far way, far away, away, near, very near etc.]
Distance :[ 50, 100, 150 km]
0.1 0.4 0.6 0.8 1.0 thousand KM
Linguistic variable ‘Distance’
1.0
0.5
0
Membership degree
Very Near
Near
Away
Far Away
Too Far
Dr. Priti Srinivas Sajja6-7 February, 2009
Network Structure
Input Layer
Output Layer
Hidden Layers
Availability of expertise
Availability of hardware/based technology
Content /length of the course
Degree of assistance required[[[
Knowledge level required for the course/ depth of the course
Market trend towards technology/course
Personal interest
Success history if any (last few years result in%)
Time taken to complete (revision)
Bio-Informatics
suggested decision for Current Trends
Wireless Tech.
Dr. Priti Srinivas Sajja6-7 February, 2009
Advantages• Quick and effective decision support• Ease of cloning and documentation • Knowledge Based
• Dual advantages through explicit and implicit representation
• Self learning• Manages vague parameters in fuzzy way• Explanation and reasoning• Management of large amount of data & dynamic
• Object oriented• Platform independent• Easy to use with fuzzy interface
Dr. Priti Srinivas Sajja6-7 February, 2009
Fuzzy-genetic System for Evolving Rule Bases to Measure Multiple Intelligence
• Fuzzy genetic hybridization
• The paper will be presented by Ms. Kunjal Mankad, ISTAR
Dr. Priti Srinivas Sajja6-7 February, 2009