Similar content module wrapup

Submitted by Owen Barton on March 24, 2008 - 11:32pm.

Here is an (incomplete) rundown of the various 'similar/related node' type modules:

Manually entered Parsed links Search/analysis of title/body Taxonomy Same content type
Related Links ☺ ☺ ☺ ☺
Related Block ☺
Similar by Terms ☺
Similar Entries ☺
Relevant Content ☺ ☺

It should be pointed out that while related links is one of the oldest of the bunch, and has most features, it is not quite as flexible as people are looking for (which I think is the main reason for the other modules), and the UI is not always easy to understand.

Here is what I would like to see - a module that (on node creation/edition and also regularly refreshed on cron) provides a hook to allow various modules to add in some numeric factors for how that node relates to other nodes. These modules could have their own configuration (e.g. to only pull from certain vocabularies), and the central API would allow you to weight one module/factor against another (e.g. taxonomy similarity v.s. word similarity v.s. destination node votes or freshness).

It would basically then just dump this information into a table and provide it to views as a sort option (that would need to load a node from an NID passed in an argument to provide a 'relevant' sort). This would allow you to display it per content type or whatever. For bonus points there would be a way for Content Recommendation Engine or other modules with user specific input to add their influence to the final list.

A couple of
Submitted by nicholasThompson on March 25, 2008 - 8:42am.

A couple of questions/points/issues/'things' I'd like to raise...

I am the maintainer of Relevant Content and am interested in rising to this challenge...

  1. This "hook" you would like. How exactly would this influence the listings? Would it override? How about this; RC can call a hook of something like "hook_rc_pre_list" and then append its own list, then call "
  2. My module (Relevant Content) lists other items simply by finding the number of similar terms (also filtering on node type, etc). This would be quite hard to achieve using a view as there are a number of joins and group by's which I don't think you could achieve with Views.
  3. I'll have a look into the Content Recommendation Engine. Sounds interesting! Although not 100% sure how it can help (due to the requirement of the VotingAPI and 1 vote-min)

    4. Interesting post and a VERY good idea to compare!

    http://www.thingy-ma-jig.co.uk

Collaboration and a note on Views
Submitted by Benjamin Melançon on April 8, 2008 - 11:31pm.

Pivots is a heavy module but probably closest to providing a central API.

On using Views to show the lists, it may be possible to provide your own query.

I am making yet another one of these modules, focusing on human-defined related content and ease of use, and would like to help be glue between modules at the same time.

Anyone doing anything on this please keep me posted.

benjamin, Agaric Design Collective