Thesis on Interactive Information Retrieval and Linked Data


As the title states, I’m doing my masters thesis on Linked Data/ Semantic Web and how we can access it in a human-friendly way. I’m humbly asking you - the domain experts - for some thoughts and ideas.

My initial thoughts are that an information retrieval system like this should; abstract SPARQL away, be knowledge base agnostic, and have an interface that makes both exploration and exploitation of data on sources like DBpedia easy.

If you have any good ideas about what something like this should look like, be able to do, or what it shouldn’t do, a comment would be much appreciated!

If you’ve used something similar (like Sparklis, what was your experience?

Hi @matsjs,

note that we have a challenge for this going on here at GFS

We made a rough prototype here:

The trick here is that users get a view over merged instances from multiple sources. We will add MusicBrainz, Geonames and GND now to this. So you have a global view. It can also be build ad-hoc based on Linked Data and the sameAs Links and ontology mappings. Would you be interested in something like this?

1 Like

Hi @matsjs

We also wrote a kick-off note a few months ago which summarizes the GFS project nicely, in case you want to read some more about it.

1 Like

Thanks for your reply @kurzum @tina_s. Here are some unorganized thoughts;

I really like the GFS data browser, it’s got some features I think are really good and necessary. Like listing all available predicates for a given URI.

I would like to work on something similar - definitely a very simple interface, even at the cost of losing some SPARQL functionality. I think there should be a plain text search for URI’s, preferrably avoiding URI’s in favor of english words (for display purposes). I think a great use-case for linked data is datasets on demand, which I’ve used myself. For example: “Get me a list of all currently operating airports, their ICAO codes, and their geolocation”. I would love to create an interface that makes it easy for people to compile these datasets from open linked data.

What endpoint does the GFS data browser use, and could I access that?

Should I keep you posted on what I’m planning for the thesis, and how it turns out?

We use a mongodb here: it is open and you can read-access it.
There is also the file we used to load it, but we are making a newer one now.
We only needed lookup for GFS so there is no SPARQL.

Check out this Thesis

I hope it will you in solving your problem or you can search Quora

Hello @matsjs,

You are completely right, there is a lack of people working in Human-Computer Interaction at the Semantic Web community. Gladly, some folks already start to realize that [1].
The main problem, in my opinion, is that SW community (mostly leaders and conference reviewers) are focusing on the back-end part of the problem, so they are not really the best people to evaluate such work.
Maybe in 5 to 10 years with a new generation of reviewers ;-).
However, must of the things you are going to find are software engineering with no profound discussion.
I suggest you have a look at the work of Lohmann [2], he is doing some interesting work in ontology visualization for Semantic Web non-experts.

If you are looking for interface patterns and how you can improve search interfaces, have a look here [3].
I do not know the status of your thesis, but I will be glad to give you some feedback or perhaps collaborate at a project level.

If you are working on such a problem, I suggest you submit in two venues: (1) ICSC Resource track [4] and (2) Natural Language Interfaces for the Web of Data Workshop [5].