Let's start by describing the output : An easy to understand, yet comprehensive guide to tagging everything. The guide must contain : 1. Descriptions of all the tags. 2. An index (preferred) or search page that emphasizes the most popular tags and deemphasizes the unpopular tags.
The world wide web is arranged in the same way with the index being generated by Page Rank or substitute. Our algorithm only has a few possible inputs per tag : Tagwatch, vote count, software support etc. We just need to decide what linear weights to use. Since voting requires very little effort it shouldn't have a weight higher than tagwatch. >From the web we learn a few things : 1. Manipulation of results esp. when the algorithm and weights are publicly known. 2. Computer generated indexes often contain duplicates and irrelevant results. So our current Map Features page performs quite well, given that the algorithm is performed by hand. We may need to make the algorithm a bit more formal, e.g. by requiring that tags must have a score higher than N before they are included.
_______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk