Please let me know how to create a knowledge base for question answer system on a text.
I have been trying to understand how to go about it, but I am clueless so far. Please help me .
Your suggestions are much appreciated, thanks in advance
First you read the text, highlighting possible questions and their possible answers.
You will also notice that there are possible questions with no answers given. You will need to note the answers in the margins.
You will also be stimulated to think of questions that do not appear in the text, but which could be useful questions to ask. You will find the answers and note these Q and A's inside the back cover of the book. Up to now, you have been behaving like a student trying to anticipate what questions might be asked in an examination.
Next, you transcribe the questions into your knowledge base. I suggest that you set up different files each with different structures, depending on the most suitable type of question: True/False Multiple Choice, Open ended, Essay question, Drawing question, as appropriate. You will need to populate the Multiple choice question answers with four wrong alternative that are look equally likely at first glance. Test for understanding, not memorisation.
Lastly, to generate a test from the text bank, you start off the students test with 5 T/F, progress to 5 MQ's etc. If you have a big enough bank, you can give different students different questions to ask, but I would only do this after many years of debugging the answers.
thank you so much Dr.Ian Kennedy,
I would like to build a knowledge base by parsing the text given to me so that when a question is asked i should be able generate an answer from the text. Could you please let me know how that can be done ?
OK. So you wish to take a question, and inspect the text to find the answer to that question.
Let us say that the text to inspect is “First you read the text, highlighting possible questions and their possible answers.”
Possible Questions are
1. “Why is the word 'First' in bold?”
2. “What must you do first?”
3. “What is meant by highlighting?
4. “What is meant by the word 'possible?' ”
5. “Why did the author use the word 'possible' twice in the same sentence?”
6. “Would it have been better if the author used a synonym for the second 'possible'?”
Possible Answers are
1. “The writer wanted to emphasise the order of the steps.”
2. “You have to read the text first.”
3. “Highlighting is the use of coloured pens to emphasize passages.”
4. “It signals uncertainty.”
5. “It was necessary to convey the sense of what the author wanted to say.”
6. “No.”
Now, how are you going to get from Qn to An? Only for the case of Q2 is there any hint of what A2 should be.
In terms of automation, you should define the tool first. Then, develop your KB using the tool's knowledge terms. For instance, if you wish to use Prolog you then you need to develop the predicates...... it could be general predicates to represents all types of questions/ answers...
Generally, to develop the KB you should first develop the Knowledge vocabularies ...
All the best
Thank you so much :) I found about prolog and i have been trying to make it work.
This is very nice tool for Prolog
http://www.swi-prolog.org/
Try it and I am sure you will enjoy it
thank you so much prof.Osman Elfaki :)
I will try my hands on the tool
Hello,
Another possible and recent approach is to use combination of semantic and syntactic approach. You should build conceptual model for the underlying domain (ontology) and then use this ontology to undeestand the question and answer it
you can check my profile for a paper called content related feedback using ontologyg
hopefully this useful for you
Thank you so much Safwan Shantnawi, this is what exactly i am looking for. Thank you so much.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages