Archive for October, 2007

How much weight does categories derived by tags have?

Sunday, October 14th, 2007

As a result of the questions “what category does this tag belong to?” on encyclopedias knowledge we obtain a list of categories. But in what extent does this count in the interestprofile? The Formula beneath trys to calculate some weight.weight of categories derived by tags

c stands for count, g for weight. the formula reads as follows: ” The weight of a category arises by its invers count of appearance multiplied with the sum of all tags in all services weighted by the count of appearance of one tag in a service, the weight of the tag an the weight of the service”.

Weigth of service measures if tags are rarely used in a serive (e.g. delicius vs. flickr).
Weight of tag measures how important ist a tag, but this needs to be improved, it is not exactly clear how to detect important tags. Maybe you got any ideas?

What nature can tags be of?

Friday, October 12th, 2007

The approach to use only tags for interest determination brings some new questions.

Not only that even tags that won’t get useable data from encyclopedias are maybe good to use in other context, e.g. all the geo-information provided via tag can be used to get data about users locations when providing data.

But what nature can tags be?

  • numbers e.g. years
  • ratings
  • recommenddations for other users beginning  with “for:” (delicious)
  • reminders like toRead, toDo (and similar noptations 2Read)
  • opinions e.g. fun, funny, humor
  • geography data starting with “geo:” or “ge:”
  • mechanical tags with prefixes like “filetype:”

Is this all? Are there more then this? Does anyone bring it into order? Maybe Metatagging, will do but do you know?