Package Tags

I read somewhere about the tags assigned to packages, but I can't find it... so I will ask.

One of my packages that I am looking at is the mysql to mysqli package. The only tag assigned to it is 'sharing'. Now it seems to me that it should be with the 'mysql' and also 'mysqli' packages to make finding it easier.

How can I make that happen?

Dave

Yes the tag assignment is automatic.

It is based on the site search results and the top most searched keywords.

Every day the site computes the top most searched keywords. Then it performs searches with the internal search engine. The top packages that appear for each of the top keywords, are assigned to these keywords.

This is not ideal because the way the internal search engine sorts the results is according to its concept of relevance of keywords in the package pages.

It is a bit weird that your package is not one of the top relevant packages about mysqli. It is related to sharing because somebody posted a forum comment thanking you for sharing.

This is a just a sign that the internal search engine needs to be replaced with one that evaluates better the relevance of keywords.

That is something that is on my long to do list for quite some time. With so many things to do, it is even hard to figure an estimate of when it can be done. :-(

Would it help if I added a readme with relevant keywords?

Very likely yes. I am not sure how the ranking score is computed but it seems to be related with the relevance of the keywords in the page as well backlinks to that page.

Same problem with my package. Tags completely unrelated to my package.

No matter what I did, mine never did correct itself either. Something Manuel will have to look into.

Dave

Well, I already explained above but let me give more details so you can understand where those keywords come from.

Basically if you search for the keywords your package appears in the top (10 if I am not mistaken) results.

The site uses the old htdig program to crawl the package page and index it. So for some reason your package page is matching those keywords. Those may not be the best keywords for your package but those are the ones that the users may use in the site search to find your packages.

Anyway, the site search engine needs to be replaced for a better one. That is something that is on my to do list but I never found time to implement it.

Meanwhile, nowadays the tags do not serve any big purpose. In the past they used to be useful to help search engines rank package pages better according to those keywords. Since Google started penalizing sites for keyword stuffing, those tags were removed for non-logged users, as well Google crawling bot.

So, if some users do not agree with them, it may as well be better to remove these tags from the pages permanently. What do you think?