Computrised Paper Evaluation Using Neural Network

COMPUTERISED PAPER EVALUATION USING NEURAL NETWORK Presented by, ANU.T.A R7 A No:7

1. INTRODUCTION Exam paper evaluation using neural network.  Adaptive real time learning.  Computers will be connected to a Knowledge Server. 



The exam is adaptive.

2. Conventional Evaluation System Students write their answers for the questions asked.  Sent for correction.  The evaluator may be internal or external.  Uses the key to correct the paper.  Marks are awarded. 

2.1 Demerits of Conventional Evaluation System Evaluator’s biasness.  Improper evaluation.  Appearance of the paper.  Time delay. 



No opportunity to present student’s ideas.

3. Proposed System  Basis :

Computerised evaluation system.  Application of neural network.  Software is built on top of the neural net layers.  Features all the requirements of a regular answer sheet. 

3.1 Neural Network - Basics 

Composed of a large number of highly interconnected processing elements (neurons).

Fig: A Simple Neuron

3.2 Artificial Neural Network An Artificial Neural Network (ANN) is an information processing paradigm that is inspired by the way biological nervous systems, such as the brain, process information.  Three layers : 

  

Input layer Output layer Hidden layer Fig:ANN-Basic Structure

Neural networks learn by example.  Two types of learning process 1. Supervised learning- Technique for deducing a function from training data. 2.Unsupervised learning- It is a class of problems in which one seeks to determine how data are organized. Eg:-SOFM 

3.3 Basic Structure  The examination system can be divided basically into three groups:  Primary education  Secondary education -learning process  Higher secondary education -specialization -adaptive testing

3.4 Organization of the reference sites  Organized for a particular institution or a group of institutions.  Specific weightage for each point.  Intelligent evaluation. 3.5 Requirement of a new grammar  Restricted to a new grammar.  Eg:-If one is to negate a sentence it is compulsory to write the ‘not’ before verb.

3.6 Question Pattern & Answering Depends on the subject.  Eg:- If a question is put up in operating systems then the student starts answering it point by point. 

3.7 Role of Neural Network 

Tasks cut out for the neural network:  Analyze the sentence written by the student.  Extract the major components of each sentence.  Search the reference for the concerned information.









Compare the points and allot marks according to the weightage of that point. Maintain a file regarding the positives and negatives of the student. Ask further questions to the student in a topic he is more clear off. If it feels of ambiguity in sentences then set that answer apart and continue with other answers and ability to deal that separately with the aid of a staff.

3.7.1 Analysis of Language by Neural Network 1. Perceptron learning :  Used for learning past tenses of English verbs. 2. Prediction of Words :  Back propagation algorithm - Elman  Present a training sample to network.  Compare the networks output to the desired output from that sample.Calculate the error in each output neuron.  For each neuron,calculate what the output should have been,how much lower or higher the output must be adjusted to match the desired output.This is local error.  Adjust the weight of each neuron to lower the local error.

Network architecture for word prediction

3. Self Organizing Feature Map(SOFM):  Invented by Kohonen.  Unsupervised learning algorithm that forms a topographic map of input data.  Represent the multidimensional data in much lower dimensions.  Vector quantisation.

A Self Organizing Feature Map

3.7.2 Training The training involves a team of experienced  Subject masters train the net to have a general idea of paper evaluation.  The language masters give specific training to the net to expect for various kinds of sentences.  The psychology masters train the net for various levels of error acceptance in semantics.

4. Merits Effective distant education programmes.  Competitive exams to become more realistic.  Evaluators biasness,handwriting-not really an issue.  Freedom of ideas.  Specialisation. 

5. Demerits Student has to learn few basic changes in grammar.  The computer cannot be cent percent error free.  Reasoning type questions cannot be evaluated.  Subjects like Mathematics,English cannot be evaluated. 

6. Conclusion The computing world has a lot to gain fron neural networks.  Their ability to learn by example makes them very flexible and powerful.  Easily integrated into a working model.  Does a lot of good for students.  Change the educational system. 

Thank You !!