Social Bookmarks are useful for finding information on the web. They have the ability to traverse the web where machines cannot go. this is because they rely on the user rather than algorithms. Users can tag and look up tags of interest. Thus creating communities of like-minded tags and individuals. Although tagging can create some problems:
1. Tagging does not produce what is known as "synonym control". Basically, if someone looks for the tag "fruit", they'll miss tags like "apple" and "orange" (which are obviously also fruit).
2. Tagging has a problem with polysemous terms. Are you looking for orange the fruit, or orange the color?
3. Many search engines only search the tags, thus missing many relevant articles.
These problems naturally occur because people perceive things differently. These problems can be marginalized by doing the following:
1. Provide them with a list of popular and recommended terms. This will make users more likely to use them.
2. Clustering. Take tags that already exist to see if the tag is referring to the same thing.
3. Collaborative filtering. When a user tags an item, create an algorithm that can link people together with like-minded tags.
It would be hard to fix these problems, because people will always think differently. What we can do though, is cure the way it is organized and project it to the user.
I just finished reading your post about the trouble with polysemes. I remember doing a database for 202 and working with the same ideas. According to the readings you are right on the money with the answers to the problems, although there are areas of compatability that I think will never be addressed. I remember reading a study about emergency room language definition where there were found to be something like 20 different descriptions of pain in the stomach. I think or wonder anyway if there will need to be a larger thesarus for every tag.
ReplyDelete