Transcript MS - ICBL - Heriot

Transforming Current
Awareness Through RSS
Lisa Rogers
Research Associate
Institute for Computer Based Learning
Heriot-Watt University, Edinburgh, Scotland
ticTOCs and Gold Dust Projects
Can RSS really transform current awareness?
• RSS Current Situation
• ticTOCs Project
• Gold Dust Project
• The way forward?
Current Situation?
•
•
•
•
RSS for Sharing Information.
What are the uses in research?
RSS adoption.
What should information professionals
do?
• Information Overload: Is RSS
contributing or easing the situation?
ticTOCs Project
• Journal Tables of Contents Service
• Aggregates TOC feeds.
• Find, Save, Display and Export latest
Tables of Contents from over 12,000
journals ~430 publishers
• Accommodates both users of RSS and
non users
• http://www.tictocs.ac.uk
ticTOCs Demonstration
Guidelines for Publishers of TOCs
• Use RSS 1.0 specification
• Use RSS 1.0 Modules (dc, prism, content)
• Don’t include HTML in the standard RSS
elements
• Use the RSS Content Module to present HTML
marked up content.
• Ensure feeds are valid
• Include abstracts
• Understand the purpose of each feed
• Do not restrict access to TOC RSS feeds.
• Provide up-to-date OPML file(s)
ticTOCs Data Set
• tiCTOCS text file
http://www.tictocs.ac.uk/text.php
• Tab Delimmited File with ticTOCs ID,
Journal Title, Feed URL, ISSN, eISSN
1 Nature http://www.nature.com/nature/current_issue/rss 0028-0836 1476-4679
2 Nature Biotechnology http://www.nature.com/nbt/current_issue/rss 1087-0156
1546-1696
3 19th-Century Music
http://caliber.ucpress.net/action/showFeed?ui=0&mi=0&ai=s2&jc=ncm&type=etoc&f
eed=rss 0148-2076 1533-8606
Using tiCTOCs data set
Gold Dust
Tracking
• ticTOCs usage:
Articles viewed,
exported or
clicked on in
ticTOCs are
collected.
• User Submitted Documents:
Journal articles written by or
of interest to user.
Profiling
• Collated Articles are fed
into NaCTeMs TerMine
Web Service
• Personal Interest Profiles
(PIPs) are produced
• Also trialled ExtMiner:
Open Source tool
combining structured
search and document
clustering techniques.
compressive strength, 3
crack initiation, 6
shear stress, 4
critical shear stress, 9.50978
threshold value, 4
microcrack initiation, 2
martensitic steel, 2
impact response, 3
composite laminate, 3
structural variation, 2
health monitoring, 2
composite structure, 2
failure mode, 2
damage detection, 3
unified approach, 2
Matching
• The Users profile is
matched to items from
the various categories
of RSS Feeds
Calls for Papers
New Items in Institutional Repositories and
Subject Repositories
Funding Opportunity News
Patents
Press Releases
Professional Society News
Engineering News Feeds
Component Announcements
Teaching and Learning Resources
Forthcoming Conferences and Events
Theses and Dissertations
News from JISC Services and Projects
Suppliers
New Book Announcements
Standards
Others
Journal Articles
Delivery
Results
•
•
•
•
Item considered ‘Gold Dust’ if rated 8-10
Tested 4 Methods in two Iterations
2nd Iteration was better 15% and 14%
Best Categories were:
–
–
–
–
Journal Articles
Items from IR and SRs,
Theses and Dissertations
Engineering News Feeds
• Best Results for a User was 63% ‘Gold Dust’
Observations
• Need more and better initial usage data
• Require method of stopping generic
terms
• Matching against items of a similar style
to input data gives better results
• If a user’s research area is more
specific results are likely to be better
Conclusions
• What should Information Professionals
do?
• What about RSS feed providers?
• What about using RSS and text mining
as a recommender system?
Questions?
• Questions?
• Email [email protected]
• Twitter @lisajrogers