Telstar

Resource Lists and Collective Intelligence

Chris Clarke from Talis asks “How can we use collective intelligence to improve resource lists?”

Talis have product Talis Aspire – hosted ‘resource list management’ service. Talis found resource list management was a collaboration between Academics, Library and Students. Need all three involved – and academic engagement is key. Libraries also need to be able to get and manage stock. Talis wanted to avoid overheads – rekeying data etc. Also wanted to give students v high quality experience.

‘Collective Intelligence’ – aggregating information across many users/uses to find patterns of use etc. and use this to generate information. E.g. ‘Which items are frequently referenced together by the experts?’; ‘If we by this book, will learners actually use it?’; What items to learners substitute for when the guided resource are not available?’; ‘Can we guess the loan strategy upfront, instead of waiting for an item to be heavily borrowed?’

Can only do this across largish datasets – and Talis is able to aggregate over Talis Aspire customers who contribute their data. At the moment have a trial dataset made up over 4 customers – but still millions of transactions.

Chris mentions Talis use MapReduce to process large quantities of data (this approach was developed by Google, although now there are open source implementations (Hadoop) and Amazon provide an elastic MapReduce service).

Four prototype APIs (all REST based):

• “Appears with” recommendations
  • Based on co-occurrences of items on resource lists. Academics who reference this, also reference…
• “Borrowed with” recommendations
  • Based on the patterns of what students actually borrow. Learners who borrowed this also borrowed…
• Loans
  • Show me how popular this item has been over time, across all institutions
• Holdings
  • Which institutions actually have this item

Chris has posted further description of the API functions to the Reading List Solutions JISCMail list.