More than once I have met articles that the tag cloud is either too unreadable or, on the contrary, does not carry useful information, and the developers are trying to find the perfect threshold at which keywords will suffice, but the cloud will not overload the user as a whole. At the same time, one has to choose between readability and informational content.
1. Contextual cloud
')
The idea is to create a new contextual cloud consisting of “friendly” tags when clicking on the main cloud tag. The “friendship” between tags is determined by their presence in one article. In simple language - when clicking on “Google”, we get a cloud from tags from all articles that have a “Google” tag (most likely the main tag in the new cloud will be “Gmail”)
This makes it possible for a few clicks to view several different in content, non-overloaded clouds and get to specific tags, and therefore articles.
2. Live cloud
Another approach, in which there is a filter above the tag block, and the formation is carried out dynamically. By typing "network" you can load a new cloud from tags in which there is the word "Network".
I like the second option less, but it is quite possible that the combination of methods will lead to the best result. It will be great if someone checks the idea of ​​a contextual cloud on real data and writes the result (problems / effectiveness).
updated3. Tag cloud can be made more contextually. If you take Habr, then you can take into account the blogs in which the user is and if he reads only games in which google almost does not appear, then the total tag weight is reduced.