In a situation where we want to build a subject course ontology using textbooks, notes or other teaching resources; general ontology learning algorithm generate poor quality ontology.
I'm researching this issue and I captured set of possible reasons. Can anyone suggest some reasons based on either authentic research results or his/her experience of this domain?
Hi Safwan, this can be based on the quality of text. If you using well-written text books as your knowledge source, you will be able to get reasonably good results. I have done a research using 'lecture slides' as the knowledge source. This research was not on ontology learning, but concept map mining (extracting concept maps from lecture slides). My main issue was the element of noise in text, particularly semantically and syntactically missing and ambiguous text in slides.
So far, what have you found out?
Hi Thushari,
Currently I'm researching the reasons, I have some possible reasons but I need more investigations to validate it.
I argue that even having good text books as source for ontology building, the available general learning ontology algorithms will not generate acceptable subject course ontology.
I came up with few researches to build an ontology for a subject course. However, they did not discuss the quality of the generated ontology.
I tried TextToOnto and Neon tools to create an ontology for "Introduction to Database Design and Management course". I got many noisy terms and relationships.
I guess that the first step which is extracting the ontology terms generate noisy/extra terms because usually text books contains real life multi-domains examples to explain the theories in the book. these examples responsible for generating extra terms and relationships. Also, usually, the terms in the textbooks have different semantic meaning than the general context meaning.
I hope you and other researchers enrich the discussion to better understand this problem and then we can proceed to design subject -knowledge ontology learning algorithm.
Regards
Hi Safwan,
I agree with you in this aspect that not all the extracted terms and relations from these tools are relevant terminologies or domain concepts.
Even though there has been lot of works done in the area of refining/ranking extracted knowledge such as utilising taxonomies/previously developed ontologies/manually annotated knowledge to train algorithms, my opinion is the problem is still open when applying in general.
I have been using IR measures and graph-metrics to rank these terms, however, I agree that it is still not beneficial in general.
If you carrying out this as a research, I would like to collaborate to design subject-knowledge ontology learning algorithm.
My email: [email protected]
Cheers!
Hi Safwn and Atapattu
what do you mean when you say "subject course content knowledge"? If you mean domain specific ontology, then It is obvious that general algorithms will not work in a sufficient manner. Knowledge is a very wide term. You can find different forms of written text in the same chapter. I am not fully aware of general ontology learning algorithms but I am sure that there are a lot of literature exist. Try to find out some surveys or critical reviews, I am sure they will save a lot of your time.
Wish you all the best
"Subject Course content" is the course textbook, slides, or notes for a given subject. i.e If we want to represent the knowledge for a given subject using Ontology representation.
In the educational field there exist few researches that tackled this issue and they were for a specific subject.
Many thanks for your input.
- Badges
- Science method
- Similar topics
- Teaching