Search Brainstorm

    • Saved Searches
      • Allow users to save their queries
    • Logging
      • Format
        • Simple string first
        • Develop query AST and store stringified AST
          • Allows differently ordered by identical searches to be aggregated together
          • Allows partial search keyword analytics
      • Assign unique id to each query
        • markup query results with search id to track selected solution
        • markup search box uri to correlate a refined search
      • Log what page was chosen for result (and how far down it was)
      • Log if another search was started instead of picking a result
      • Log date when search was done
    • Additional Analytics
      • Mine google analytics ( or create ajax callback for simple built-in tracking) to create popularity ratings for pages
      • Add community and curator ratings to search index to influence user influence order
        • if values can't be used to affect rating quality rating, use ratings at deki side to re-order results
    • Reports
      • Popular searches
        • Allow digging into list of most popular selected results for popular searches
        • Allow digging into popular alternative search terms (from subsequent search tracking)
        • Show top 5 results for each most popular search
        • Show most popular searches by time range (today, this week, this month)
      • Filtered
        • show all searches that contain a keyword
    • Search result improvements
      • Suggest alternative terms for search terms using subsequent search history
      • Influence result result order based on selection popularity of items
        • Use both popularity as it relates to the current term and popularity in general
      • Show sidebar of items that were tagged with search:<keyword> matching the search keywords
        • search tagging should be limited to admins.. So maybe thsi should be properties?
    • Boost settings
      • allow customization of boost settings (e.g. don't consider tags as important)
      • boost priority:
        • title, segment name, tags, content
        • not path

    Division of Concerns

    • Configuration
    • Logging
    • Analytics (what kind of reports can be generated)
    • Search Experience
    • Reporting Experience (how reports are generated)
      • what kind of views are defined
      • how are these views parametrized

    Search result optimizations

    • Lucene
      • Lucene should no longer include content in the index itself
      • The default result set should not do paging and include only the vital information, likely the following:
        • content id
        • content type id
        • ranking
        • private flag
        • dates (created, modified)
    • API
      • Retrieves condensed search set
      • Filters by "private" pages for requesting user
      • Augments ranking with
        • Demotion
        • rating
        • search analytics quality data
      • sorts by ranking/date criteria
      • stores entire augment set in cache for some period
      • Each page only hits the database for the content Id's in the page's view into the full set
      • Paging does not return to lucene or do any other DB query other than content fetching
    Tag page
    Viewing 1 of 1 comments: view all
    @SteveB Something more:
    - Bost define:tag at first result if tag it's on search criteria
    - Allow multiple searchers, this way interal users just use wiki search, I mean:
    * Results from wiki
    * Results from google
    - Deterministic return, on getSearch, If I ask for 100 results give me 100 results, don't
    bother me with filtered by security effects.
    - Take care of RoyK page rating: http://developer.mindtouch.com/en/docs/MindTouch/Specs/Content_Rating

    Thanks. edited 22:22, 11 Mar 2010
    Posted 22:20, 11 Mar 2010
    Viewing 1 of 1 comments: view all
    You must login to post a comment.

    Copyright © 2011 MindTouch, Inc. Powered by