ENTAG

Presenter: Kora Golub, UKOLN
Feb-12-08, Birkbeck, London
http://www.ukoln.ac.uk/projects/enhanced-tagging/

ENTAG is about enhancing user-generated vocabularies with controlled vocabularies.

Objectives (managed to note 2 of 3 main project objectives):
- investigate indexing aspects when using only social tagging versus when using social tagging in combination with a controlled vocabulary
- Investigate the above in 2 different contexts: tagging by users/readers and tagging by authors.
- 2 demonstrators: Intute (reader tagging) and STFC – Science and Technology Facilities Council digital repository (author tagging)

Next step: evaluate and compare “basic” system vs “advanced” system, attending to factors including:
- Indexing: perspective, exhaustivity, specificity
- Linguistics etc

Proposed user sample for testing: 50 graduate students in political sciences
Students will be asked to tag 100 documents, covering 4 topics of relevance
Data collection: logging, pre and post-questionnaire
Results by: August 2008

Expectations: precision and recall to be improved when controlled vocab used to enhance tagging.

Users would be advised on what the team thinks would be the best way to discover resources. Not yet decided how to instruct users on what to search for. At the same time, don’t want instructional burden to be too heavy – team is wary of asking too much of testbed users.

The hope is that answers might ‘fall out’ through observing user behaviours.
Team will be logging series of ‘required’ searches – not just observing users’ normal behaviour.

Discussion

Discussion focused on battling problems of controlled context and learned expertise versus observation of behaviour ‘in the wild’ – on a time-limited project

CH comment: It would be interesting to compare the results of this study with analysis of information search/resource discovery behaviours in VLEs (as ‘closed systems’ with clearly defined user communities). There is clear potential to use VLE log analysis to observe user behaviour in multiple group / subgroup contexts (and to use these observations, in turn, to inform future use case development). Another parallel exists with TRAILS and similar projects looking at social dimensions/drivers of information/resource discovery (e. g. looking at the implications of making ‘hot spots’ visible / transparent to users - what happens when you make it transparent what other people are looking at/searching for; how does this influence information resource consumption).