Clustering

The new results clustering option that was detailed when we rolled it out for testing is now implemented on mojeek.com and is available in the API. Thank you all, for the feedback here and elsewhere. We show an example below.

The number of indented results can be controlled with a parameter ‘clufmt’. The default is a maximum number of 5 (main result + 4 as shown below). if you wanted to reduce that to 3, say, using &clufmt=3 in the URL. If you are using Focus the default for ‘clufmt’ is 2.

Most people seem to like it a lot, and the only negative mentioned was that some people like the old (one result per domain) default option. That can still be obtained by using URL parameter (set &clufmt=0) - full details and other options here posted by @mike. We had to remove the number of pages per site option from the Preferences menu, but will consider putting it back in; it’s not simple or we would have done it already.

Some detailed notes:

  • The new clustering is more dynamic than the old method and will only show more than one result from a host (technically FQDN) if it’s relevancy is determined it to be on that page.
  • Clustered hosts can be repeated on subsequent result pages, unlike the current clustering.
  • This clustering method is closer to other search engines than our current clustering.
  • The default clustering has a maximum number of 5 results (main result + 4 as shown below).
  • A value between 1-5 means max 1-5 results for the very first result only, and a max of 1-2 results for every other host (per page).
  • This is a completely new clustering system/algorithm and not compitable with the current ‘si’ setting.
  • The ‘si’ functionality will remain in the API, but will require ‘clufmt=0’ once the new clustering is the default.
1 Like