[Tagdb] Tag Relevance

[Thijs Terlouw]

I am currently working on the same problem and came up with the same solutions:

1. Just rank based on how many times a tag is used with a resource
(link) and then perhaps rank by date descending.

or

2. Rank by how many tags, but take into account the wisdom of the
crowd. I am leaning towards this approach, since tagging is all about
social wisdom.

There are some things I have not worked out yet though:
* Resources with same tag-count but different co-tags. Like in your
example R_a{20,100} and R_b{20,0} Then there is more consensus in the
crowd that the resource B (R_b) is more definately about that
particular tag. But the problem might be that some new website does
not yet have those 20 tags, but only has 19: R_c{19,0} for example.
Should resource C now be scored higher than resource A or not? Perhaps
need some ranking for this, which takes creation date of the resource
into account as well.

Perhaps this brings us full circle back to Lucene-based implementations?