jcdl

JCDL 2005 Using Collection Descriptions to Enhance an Aggregation of Harvested Item-Level Metadata: 

JCDL 2005 Using Collection Descriptions to Enhance an Aggregation of Harvested Item-Level Metadata Muriel Foulonneau (mfoulonn@uiuc.edu), Tim W. Cole (t-cole3@uiuc.edu) Thomas G. Habing (thabing@uiuc.edu) Sarah L. Shreeves (sshreeve@uiuc.edu) Grainger Engineering Library University of Illinois at Urbana-Champaign June 2005

The CIC metadata portal: 

The CIC metadata portal 12 partner institutions – 500.000 metadata records testing multiple interfaces on top of aggregated metadata http://cicharvest.grainger.uiuc.edu/

Aggregated collections : the loss of context: 

Aggregated collections : the loss of context Missing information to Retrieve records Select from a list of results Co-locate records

Collection-enabled functions: 

Collection-enabled functions Search granularity Search collections only Search items with collection information Selecting relevant search results Interpret item description Co-locating resources Grouping results Browsing collections Filtering results

Filtering / selecting: 

Filtering / selecting

Searching / selecting source / co-locating: 

Searching / selecting source / co-locating

Suggesting complementary resources: 

Suggesting complementary resources

The DC Collection standard: 

The DC Collection standard http://dublincore.org/collectionwiki/InstanceTemplate

Approaches to collection –item retrieval: 

Approaches to collection –item retrieval A matching record is in a collection A collection description has a match

Match cases for multiple terms queries (MT): 

Match cases for multiple terms queries (MT)

Test with 2287 real-life queries (1638 MT): 

Test with 2287 real-life queries (1638 MT) Partial match Rest of match Full match No match

Impact on application results: 

Impact on application results Ranking results AND in item level records AND in combination of item and parent collection records OR in item level records Suggested collections Results grouped by collection

Harriet Beecher Stowe: 

Harriet Beecher Stowe “Harriet Beecher Stowe” appears in 40 item records Q1 = Poetry of Harriet Beecher Stowe 0 item records contain poetry & Harriet Beecher Stowe But 6 item records contain Harriet Beecher Stowe AND appear in collections having to do with poetry Q2 = Harriet Beecher Stowe in Antebellum History 0 item records contain antebellum & Harriet Beecher Stowe But 14 items in Harriet Beecher Stowe AND appear in collections having to do with antebellum history

Match in combination Item-Collection : 

Match in combination Item-Collection Search “Harriet Beecher Stowe” appears in 40 item records Q1 = “Poetry of Harriet Beecher Stowe” 0 item records contain all words 6 item records contain Harriet Beecher Stowe AND appear in collections having to do with poetry Q2 = “Harriet Beecher Stowe in Antebellum History” 0 item records contain all words 14 items contain Harriet Beecher Stowe AND appear in collections having to do with antebellum history

Conclusions: collection info has an impact: 

Conclusions: collection info has an impact Potential influence on Result ranking when limited number of results matching the full query Co-location – suggested related resources through collection access Property inheritability Always -> no added value of combined operator Never -> no interest in collection / item retrieval Sometimes -> A non matching record is in a collection where other records match, therefore …. User expected granularity level Future user testing Other pieces contain context resources Data providers expose more CLD What is the best way to share them? Metadata services Collections Resources

Collection information impact: 

Collection information impact Potential influence on Result ranking when limited number of results matching the full query Co-location – suggested related resources through collection access Property inheritability How often? Metadata practices and their practical consequences User expected granularity level Future user testing

Further explorations of digital library systems: 

Further explorations of digital library systems Data providers expose more CLD What is the best way to share them? Other pieces of information contain context Metadata Interpretation / reference services Collections Resources