Present a relevant list of tags
Use cases
- Enrico is navigating the Debian archive and to use tags to further refine the search. However, he does not want to waste time going through a long list of tags that are meaningless to his search.
Implementations
Most frequent tags
- Take the list of currently displayed packages
- Take the list of their tags
- Display only those tags that appear most frequently.
Example implementation: Adept 2.0 smart search.
Tags most relevant to a full-text search
- Take the entire list of packages
- Rank their tags by frequency
- Take the list of packages resulting from a keyword search or some other kind of search
- Rank their tags by frequency
- For each tag, compute how many places it advanced in the ranking after the search
- Display only those tags that advanced the most
This is a very accurate way to find, given a package search, the tags that lead in the same direction of the search.
It can also be used to convert a keyword search to a tag search.
Example implementation: debtags smart search, from still unreleased debtags (http://svn.debian.org/wsvn/debtags/debtags/1.6.0)
Most discriminant tags
Given a list of packages find a list of tags each of which, if selected, would lead to a list of packages which is half as long as the previous one.
This method presents the user with a list of tags each of which embeds a relevant step forward in the selection.
Example implementation: none known.