Streamline News

Integrating Repository Function with Work Practice

Search method proposals

Posted by Dawn on September 18, 2008

Mark and I meet earlier this month to discuss where we wanted to go regarding the search prototype. We decided that we would utilise the work already done by Mike on the SRU interface for the research aspects of the repository. This would provide us with an Input Interface for querying the repository.

Mark would come up with some ingenious methods of extending the basic search process. These will incorporate Elizabeth’s ideas about bookshelf browsing without relying on the latent semantic analysis (LSA) she is developing. The LSA current version is reliant on libraries not compatible with a web based apps and will take some time to translate. In the meantime we can put something together that gives the visual impression for testing with users.

The output from this would then be displayed in some visually interesting way, see interface designs here, possibly using PHP, JavaScript, Java servlets or the Google web toolkit. This is the bit I have together put together, after exploring potential technologies.

Today I received four ideas for the processing element form Mark. For those of you not so techy 😉 I have done a few diagrams to give you a better idea.

Method 1 – simple extension
Submit multiple searches, by getting search info. Then creating alternative versions using thesaurus. Rather than doing a single search do multiple searches using this information then produce results as a visualisation.

Method 2 – Iterative result reuse
Getting initial results from a search, extract keywords (or other metadata) of returned document, then use this to find related documents. Extracted keywords and those found in second pass should be standardised using thesaurus. For each set of search results, submit new search, then follow same process recursively (obviously removing common returned elements). Each iteration should result in documents less related to initial document.

Method 3 – Collaborative search via profile matching
Identify all searches performed by a user in a particular session i.e. build a profile of a particular users searches. X-reference this to searches performed by other users in a particular session. Identify different searches are commonly used within the same sessions. This should give a list of related articles that may be of interest to users who search for the dame type of things (even if the keywords, content or metadata of the documents is completely unrelated). The more sessions that a set of related searches are performed, the more likelihood that they are related in some way.

Method 4 – Collaborative search via document matching

Anytime a search is done cache the returned documents. When new searches are done get a returned resource and find all caches that also contain that document, then return the contents of each cache as secondary results. Again this could be done recursively.


One Response to “Search method proposals”

  1. […] and Mike have built the input and processing elements of the Learning Object search tool, see post here. My job now is to take Mark’s output and do something visually funky with it. The GWT […]

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: