Streamline was a JISC funded project that completed in March 2008. This is a presentation I gave about the Automatic Metadata Generator that was developed during that project. at the HEA-ICS conference in Durham 2010
1. Facilitating reuse of learning resources: a tool to support self-deposit and automatic metadata generation Dawn Wood & Janet Finlay Leeds Metropolitan University d.a.wood@leedsmet.ac.ukj.finlay@leedsmet.ac.uk http://streamlinenews.wordpress.com/
2. Introduction JISC funded project Streamline Completed in 2009 Examine the process of Repository: Deposit Retrieval The crux of the problem: Metadata (data about data) Learning Object Metadata (LOM) standard Present one possible solution for the generation of KEYWORDS within LOM This was tested on a series of Podcasts produced by CETL ALiC
3. First … What KEYWORDS would you use to Search for this image?
4. Purpose of the photograph Description of Mac Chair Desk Window Location Hotel room Manchester File information Time and date (already present) File type Context Hospitality training How to maintain guest rooms. Checking for left key cards. Furniture design Photography Metadata is only from the authors perspective ... The LO may be relevant in many other scenarios
5. Observations of metadata creation When entering metadata for a learning object observations show that: Phrases rather than single words are used Repeated content is often used Very few keywords are entered if any The keywords used were either very specific or very generic, rarely a mixture of both This creates a problem for searchers of information and limits the possibility of the learning object being found within a repository
6. Creation Workflows Examination of workflows showed that: Creation of multimedia learning objects revolved around a script Course and module documentation was referenced during creation Contains common content that specifically defines what the learning content is about Contains the context of the learning object There is also the learning objects written content, if any.
7. Assumptions Documentation describes the learning object More keywords are better This is reliant on good search techniques Not starting from a blank slate is easier Generate a selection list Author still feels in control
9. Example Process To test Keyword generation of the system with multi media. The aim was to produce: more words than the author all the author words relevant words For a Podcast on Accessibility produced by CETL ALiC One of the authors used the following key words and phrases: HCI Accessibility User centred design Disability Six words in total after splitting phrases and removing duplicates The most obvious keyword is missing: Podcast
10. Results from the Generator Remember that different users will find different aspects of your interface difficult. For exampleusers who are colour blind will not see colours as you intend and may not be able to distinguish between colours on the screen. You can test how your colours will look using an online simulator. Extract from script accessibility blind colours disability example information interface podcasts potential produced reader screen test text user accessible users Missing: HCI centred design From 6 to 14 Key Users keywords Similar words removed Appropriate keywords Extracted words
11. Finally Generated also tested with: Text based learning objects With users within a deposit workflow The generator increases the number of keywords assigned to a learning object Selection enables author control
12. Thank You. Questions? Dawn Wood & Janet Finlay Leeds Metropolitan University d.a.wood@leedsmet.ac.ukj.finlay@leedsmet.ac.uk http://streamlinenews.wordpress.com/