STELLAR May update

We’re now well into our Post-enhancement and evaluation stage as the project starts to draw to a close and this post will provide an update on the work over the last few weeks.

The STELLAR proof-of-concept tool
The project has created a semantic web tool to search the linked data representations of the digital library.   The tool has been created by Dr Mathieu d’Aquin from the Knowledge Media Institute and builds on previous work including the linked data materials created for the Jisc-funded LucSTELLAR smenatic web proof of concept screenshotero project that established data.open.ac.uk, and a tool called DiscOU.

Essentially the tool (shown above) allows you to paste in (or type in) either a whole section of text or just a few search terms into a search box.  Ideally you would put in a rich piece of text rather than just a few simple search terms.   The text is then passed through a semantic meaning engine and concepts are matched against the concepts contained within the digital library dataset.  A selection of the closest matches are then displayed, along with a thumbnail image of the item.  These link through to the object in the Fedora digital library.

The tool lets the user revise their query by using a set of ‘sliders’ on the left of the screen to choose which concepts most closely match what they want.  STELLAR concept featureSo it’s possible to decide that you don’t want to see so much material that talks about ‘Ice’ or ‘Snow’ for example, but you do want to see more material on ‘Atmosphere’.  Changing the sliders and clicking on Discover gives you a new selection of material to choose from.

The difference with a semantic web tool is that you aren’t just matching a string of text, but are analysing the meaning of those words and finding related material.  So you could put in a search for a particular thing, say Mt St Helens, and the semantic search system will recognise that material about other volcanoes might also be relevant.

As well as retrieving material from the digital library the tool can also show related material from other datasets in data.open.ac.uk, currently this includes material from OpenLearn.

STELLAR item displayEach item from the digital library is shown in a separate block, with the title of the item, a description of what it is and which course it comes from, alongside a thumbnail of the item itself.  For the proof-of-concept we just picked a selection of material from three OU modules (S100, D251 and L550) to include in the system to show how the system works, but items are indexed at a page level so you can get directly to the appropriate page.  Clicking on the title or thumbnail takes you directly to that item in the digital library and opens that up in a separate tab. Clicking on the course code takes you to a data.open.ac.uk page showing the extensive metadata about that course or element of the course (e.g. http://data.open.ac.uk/page/book/0851413080).  Finally the tool allows you to deselect items if they aren’t relevant to you.

Post-enhancement and evaluation stage
Current activity is mainly focusing on a series of evaluation activities such as focus groups to talk to stakeholders who were originally involved in the pre-enhancement stage.  From that survey we had a pool of some 170 people who were happy to be contacted again about STELLAR and we’ve drawn on that group for a series of workshops.  The workshops give them a chance to use the tool and comment on how they find it, how they might use it and what’s its value would be.  This will be followed up with some surveying.

Next steps
As we reach the last stage of the project we’ve also been starting to work on the final report and putting together our thoughts about the sustainability of semantic web technology in our digital library.

This entry was posted in Enhancement, Evaluation, Linked Data, Post Enhancement Survey, Updates and tagged , , . Bookmark the permalink.

One Response to STELLAR May update

  1. Pingback: Digital Library update May 2013 | OU Digital Library Development Project

Leave a Reply

Your email address will not be published. Required fields are marked *