January 21, 2008

The Meaning Of A Tag

I am very excited about the MOAT Project, an effort to map user-generated tags to the semantic web. This is a super-important effort as web site operators often might want to utilize and contribute to the semantic web but cannot invest too much valuable brain power into doing so. It also helps with the marketing of meaningful tagging, that there is more to be had from tags than simply tallying them up like votes for bad candidates.

Here is an example of how to assert two different meanings for a tag:

<moat:Tag rdf:about="http://tags.moat-project.org/tag/paris">
<moat:name><![CDATA[paris]]></moat:name>
  <moat:hasMeaning>
    <moat:Meaning>
      <moat:meaningURI rdf:resource="http://sws.geonames.org/2988507/"/>
      <foaf:maker rdf:resource="http://example.org/user/foaf/1"/>
    </moat:Meaning>
  </moat:has_meaning>
  <moat:hasMeaning>
    <moat:Meaning>
      <moat:meaningURI rdf:resource="http://sws.geonames.org/4402452/"/>
      <foaf:maker rdf:resource="http://example.org/user/bob/foaf"/>
      <foaf:maker rdf:resource="http://somwhere.net/myblog/foaf.rdf#me"/>
    </moat:Meaning>
  </moat:has_meaning>
</moat:Tag>

Source: MOAT ontology

I have always believed that site operators and end-users should interact with the semantic web in a seamless, familiar manner. What MOAT must be answered with is UI that encourages meaningful tagging. In its current state, though, tagging appears to have hit an innovation plateau because it is difficult for users to add more than shallow, impressionistic meaning to a subject. There is enough in tagging to generate interesting visualizations such as tag clouds and (of course) to improve the findability of site resources. But there needs to be a way to add one more dimension to the subject-tag relationship.

Two recent experiments that I have developed in meaningful tagging are to be found in the Memecat and Listgasm projects.

In order to add the third "predicate" dimension to the tagging of a subject, I provide cues as to what the tagging context is when a user enters tags. To allow for creativity, we chose to allow the users to freely enter whatever tag they wished, but still provide "object" suggestions. On the storage side, each user assertion is stored as a triple: the item, the predicate, and the tag entered by a user.

When entering a tag to "keep" a list item in Listgasm, the user is presented with a dropdown of suggested tags. The same approach is used to "lose" a list item, which represents a different predicate for the same subject. In Memecat, a video meme is similarly tagged for pattern, culture, and what it contains. A major benefit to the tagging and storage approach used in these projects is to allow a user to not only tag different aspects of an item, but to also enter more than one tag in doing so.

What I haven't yet done is connect the data that we've collected in these projects to the semantic web. MOAT could be the best way to quickly do this.

January 1, 2008

Hashtags

Hashtags is an effort to add loose metadata to twitterings. Currently, users are referred-to with a preceding "@" in Twitter messages. In the case of hashtags, a twitterer "follows" the Hashtags user and then tags words by prefixing them with a "#". This keyword is then tracked by hashtags.org.

This could get messy as hashtags participants are already placing the "#" in front of too many words that they deem relevant. Perhaps hashtaggers should only place their metadata at the end of the twittering so as not to interfere with the human-readable part of the message.

More ideas for metadata in short messages:
- hashtags could encourage participants to use a set of common prefixes and then connect their community into the greater semweb.
- comma-delimited tags after the "#"
- generate threads of related twitterings using a "re:" and a keyword

Spotter: Everything Is Miscellaneous