Week 3: Semantic Web, Linked Data, & Intelligent Curriculum

How will (could) linked data and the semantic web impact learning?

How will (could) linked data and the semantic web impact learning?

by George Siemens -
Number of replies: 15
The semantic web has received far more hype over the last five years than practical results on our daily lives would suggest. However, there are some fascinating developments underway with linked data that may have a dramatic impact in how we encounter information. The task is enormous - remove the ambiguity of terms/concepts/words. But, as Peter Norvig's video last week suggested - a large enough data set, combined with probabilistic computing, can provide a huge leap forward. What happens when we tag big data and combine it with linked data? That's the topic this week - and after you've had time to review a few readings/videos, please share your thoughts about how this might impact education, learning, and training and development...
In reply to George Siemens

Re: How will (could) linked data and the semantic web impact learning?

by Juanma Dodero -
Linked data (LD) and the semantic web (SW) are an opportunity for learning analytics to streamline the analysis. It can be helpful to perform a more in-detail analysis of concepts and subjects dealt with in a coure. For example, you can extend beyond 'moodle posts' and find within them some instances of a concept that has to be treated in the course. If you have first the posts tagged/annotated or linked with meta-data, and then you do a SNAPP on the forum, you may find out that many of the interactions can be non-sense for the purpose of the forum, that people in those posts were actually greeting each other or writing about their holidays ;)

An old threat to the SW/LD opportunity is that first there must be someone/something annotating the posts with the concepts. If this has to be done manually, the opportunities of exploiting the SW an LD can decrease. A classical and critical text about that: http://bit.ly/1Dpk5. Fortunately, there exist great efforts to obtain metadata automatically, such as FreeBase or DBPedia.

Another issue is that we need to formalize the concepts and linkage relationships, i.e. the knowledge about educational concepts, so that they can be computer processed. We have to agree on the 'model' underlying that knowledge. But some people can be reluctant to an uniform, machine-understandable way of formalizing and expliciting the knowledge on education and educational research.
In reply to Juanma Dodero

Re: How will (could) linked data and the semantic web impact learning?

by Thieme Hennis -
Juanma:
Another issue is that we need to formalize the concepts and linkage relationships, i.e. the knowledge about educational concepts, so that they can be computer processed. We have to agree on the 'model' underlying that knowledge. But some people can be reluctant to an uniform, machine-understandable way of formalizing and expliciting the knowledge on education and educational research.

Formalization requires agreement about definitions. Not only about education, but epistemology as a whole. What can we know? What is expertise? How should we define and structure knowledge? There is a bottom-up approach to this which becomes more feasible if the definitions and the structures are making sense and are already adopted by the community using them. An imposed ontology is less likely to become accepted than an emerged ontology. That leads to the question: how can we influence the emergence of ontologies? We have seen that folksonomies tend to expand and therefore more difficult to manage. If three different words in a folksonomy have the same meaning, how can we make people want to use only one?

My assumption is that reputation is an instrument that can be used for this matter: Suppose that people's reputation in a community is expressed using the keywords/definitions of the community ontology. Then these keywords all have a relative value that decreases with the number of keywords that mean the same. It is therefore wise for people who find their reputation important to agree upon a set of keywords that express their knowledge, and not use too many keywords for that.

PS. I am doing research in this direction, and wrote an algorithm that supports this reputation mechanism.
In reply to Thieme Hennis

Re: How will (could) linked data and the semantic web impact learning?

by Thieme Hennis -
the first paragraph (above) is a quote, but the it did not render it as such. anyway, I repeated the last paragraph of Juanma's response to reply to it.
In reply to George Siemens

Re: How will (could) linked data and the semantic web impact learning?

by merce galan -
Hello, I have little experience to know about the semantic web through this learning platform GNOSS, is being scheduled this way.
Leave the link in English that will explain better than me.
Greetings
Mercè Galán

http://www.gnoss.com/acerca-de-gnoss/2
In reply to George Siemens

Re: How will (could) linked data and the semantic web impact learning?

by Apostolos Koutropoulos -
Quite a huge topic to chew on!
I've been working in an academic Library for the past 5 years and the Semantic Web has been something that my fellow techie librarians have been mentioning. Honestly, nothing much has panned out on that front. I count myself as a sceptic when it comes to the SW but I am open to the idea and the conversation.

Linked Data is quite interesting to me. The library is full of linked data. If you think of LCSH (library of congress subject headings), book arrangements, book co-locations and other types of "old school" tagging, it's not hard to see that some linked data already exists. We could go much further, but at least there is some existing system in place. The main hurdle that I see with Linked Data in education can be seen now: how do you get a new crop of students to realize that the data that the OPAC (library catalog) provides you is in its own way linked, even though there aren't always hotspots that you can always click on to get related info.

Part of it, I think, is a systems design and UI design issue (making our systems more accessible), and part of it is training students to be information hounds so that they don't just take the first result, but they are encouraged to traverse the links to find something that is more useful to them.
In reply to Apostolos Koutropoulos

Re: How will (could) linked data and the semantic web impact learning?

by Laurence Cuffe -
I think its a wonderful project, but perhaps a little too ambitious, and I'll tell you why. Humans love to re-purpose things in unexpected ways, and I cant see a set of tags which would be broad enough to encompass all possible uses, yet retain enough specificity to be useful. As an example:
Once upon a time there was an astronomer who was looking at galaxies, and after two years of looking at galaxies everyday they asked for help:
http://en.wikipedia.org/wiki/Galaxy_Zoo
and currently
http://www.zooniverse.org/projects
Whats interesting in all of this is that:
1) These are collections of big data which still need many thousands of people to generate knowledge, and
2) A specific tool for a specific task, (deciding whether galaxies were spirals or not) is now looking for space junk on the moon, examining solar storms, and reading old ships logs to find out about global warming.
This is a very human thing to do, other examples, a website set up for selling Pez dispensers turned into Ebay, and a piece of software set up to help an academic keep track of references, turned into Google.
This is human.
At our best we look above and beyond.
The second problem is the one alluded to in the old joke about the British and the Americans being two nations divided by a common language.
As a chemist I once studied a set of small molecules each containing 4 atoms.
I was talking to another computational chemist and they recommended a particular piece of software as being good for small molecules.
I tried it, and it failed.
When I met them again, I asked about it. It turned out that they were doing computational biochemistry, and small for them was around 10,000 or so atoms big.
Even in very closely related fields terms can have very different meanings.
My 2c.
Laurence Cuffe



In reply to Laurence Cuffe

Re: How will (could) linked data and the semantic web impact learning?

by Dianne Rees -
I think an interesting view of differences in perceptions when it comes to tagging can come from a "game" like Google Image labeler where people are paired up with partners and asked to tag (separately) the same image and then share results. You receive points when your tags match. http://images.google.com/imagelabeler/

It would be interesting to see whether winning the game would cause you to start overgeneralizing or to get more specific (arguably, a combination of both types of tags are needed).

In reply to Apostolos Koutropoulos

Re: How will (could) linked data and the semantic web impact learning?

by Inge Ignatia de Waard -
It is indeed true that not everybody finds relevant information in the same way. Indeed there are many students that still miss the skill to come up with relevant tags, screen options that pop-up. The difficulty might be related to understanding what makes meaning or what will deliver personal, relevant results.

Maybe the way students use tags to describe learning objects, and to search through them, can give an idea on how the first steps the semantic web are taken? At first tags and things do not really match sensitive differences in meaning, but as time goes by and linguistic semantics are fine-tuned, the tags linked to object become more precise.

I would think that this precise tagging and finding the most relevant key words related to a certain subject, is linked to the novice or expert state that a person is in. The more expertise, the easier it is to provide specific and accurate key words or phrases, the more one is a novice, the more those choices might be colored by popular nouns or hyped words.

@Laurence: I also really like your idea that human beings have the tendency to go above and beyond. If this urge to move beyond could be put into an algorithm, learning analytics would provide us with cognitive or other laws that we would not even have imagined (or proven).
In reply to Inge Ignatia de Waard

Re: How will (could) linked data and the semantic web impact learning?

by Apostolos Koutropoulos -
"I would think that this precise tagging and finding the most relevant key words related to a certain subject, is linked to the novice or expert state that a person is in. The more expertise, the easier it is to provide specific and accurate key words or phrases, the more one is a novice, the more those choices might be colored by popular nouns or hyped words. " - IIdW

I agree with you 100% and, anecdotally of course, I've seen this in my own personal information organizing behavior. Back in the early days of Del.icio.us use, I didn't really pick tags well, as a result some of my bookmarks were not so easy to find later on. As I used the system more, I actually ended up coming up with terms that were general enough to point a category out, but specific enough so that not everything was "web2.0" (Now if only I had the time to put in an abstract/description for each URL :-) )
In reply to Apostolos Koutropoulos

Re: How will (could) linked data and the semantic web impact learning?

by Inge Ignatia de Waard -
oh, I so follow you on the 'adding an abstract/description' into delicious :-D
And I must confess, it also feels like an effort to take out those tags that I only used once (similar in my blog, I should clean house, but I keep putting it forward to ... later). I wonder if there would be people willing to clean up the messes that are made, or maybe a cleaning algorithm?
In reply to Inge Ignatia de Waard

Re: How will (could) linked data and the semantic web impact learning?

by Apostolos Koutropoulos -
A cleaning algorithm would be useful...but...I wouldn't trust it :-)

a year or so ago I undertook a project to clean up my music, re-rip it into iTunes, scan cover art and have appropriate meta-data. I was almost 90% when my hard drive died. Haven't gone back to it yet :-)

I think it would be worthwhile for me to go back through my delicious bookmarks and retag and add abstracts. I "only" have about 2000 bookmarks. I think that what I will find is that about 500 of those links are dead, which makes it a good excuse to clean things up ;-)

As far as going back to old blog posts. I have considered it, but with about 3000 posts (across 3 blogs), that could be an issue. For blogs, I think it doesn't matter as much (the retagging) since one could do a forensic analysis on our blogs and could tell things about the evolution of our thinking. I write blog posts to outwardly communicate with others. Bookmarks on the other hand, I keep because I find them useful for my own personal use, the outward sharing is just a nice side-effect :-)
In reply to George Siemens

Re: How will (could) linked data and the semantic web impact learning?

by Gillian Palmer -
From me, not so much answers as a vision: Designing for adult learners as individuals Defining the datasets needed, aggregating them and analysing them not only for prescribed courses but for whole-life use (especially when developing career paths) will take a while but BIG DATA deserves BIG VISION.
In reply to Gillian Palmer

Re: How will (could) linked data and the semantic web impact learning?

by Kelly Edmonds -

Gillian, I enjoyed your blog post about designing learning for adults, also my background. I liked the idea of learners collecting their own data based on their learning patterns supplied through learning analytic means. This way they can understand how they work with content, etc, and become cognitive of their learning patterns, thus making decisions for themselves on what and how to engage.

I was struck by a notion given by Marissa Mayer of Google in a video provided in Week 3 of this course. Instead of being influenced by the content that is found through a general search online, our context (lives, desires, ideas, etc) would become the influence of that search and drive it. I can see this extending to learning. When we post an idea, etc. it would be unique to have our context come forward to shape and influence that post, such as images, sound, colour, data, etc. Who I am would come with me when I engage with content and others.

This collective self would change and morph over time presenting who I am at that time. To create a specified ontology ahead of time (as discussed in this thread) would negate that personal development and the emergence of new ideas to share with others. I believe we need to make way for evolution.

And, I don't fear that the unfettered development of ideas would become eccentric and chaotic creating misunderstood content. We are humans after all, and we do have limited processing powers that rests on previous knowledge and experiences. In essence, I don't think we would lose control but instead allow room for the expansion of thoughts and expressions.

In reply to Kelly Edmonds

Re: How will (could) linked data and the semantic web impact learning?

by Gillian Palmer -
Kelly, It's getting late here for a Friday with no dinner yet but I shall get back to you as I appeciate your thinking.
Best,
Gillian
In reply to Kelly Edmonds

Re: How will (could) linked data and the semantic web impact learning?

by Viplav Baxi -

"This collective self would change and morph over time presenting who I am at that time. To create a specified ontology ahead of time (as discussed in this thread) would negate that personal development and the emergence of new ideas to share with others. I believe we need to make way for evolution. "

Kelly, this is spot on. There is no one conceptualization of a learner. Its like the Hiesenberg Uncertainty Principle ("the more precisely one property is measured, the less precisely the other can be measured.") can be so aptly applied when we talk of describing the "collective self".

This is why personalization is so very tough. And it is precisely why BIG data is deficient because it makes for aggregate personalization (stereotypes).

It is also where RDF/Linked Data could be potentially most useful if only it could also *model* relationships (impact of changes) instead of just *describing* them. That is where Connectionists feel they can contribute because they do attempt to *model* the relationships.

I still feel uncomfortable about the difference in the way connective knowledge is defined by Stephen and knowledge is defined in the Semantic Web discussions. For the semantic web, knowledge is propositional and symbolic. Connective knowledge is neither (to the best of my understanding).