TaxonomyForContext
From MashupCamp
Discuss this topic in an online threaded discussion forum
Contents |
Content Taxonomy and Semantic Search
How best to organize content from mashups? Traditionally, directories, tagging, and keyword-based searches are the most popular. The problem with directories is that people don't necessarily agree on what the best directories should be. On the other hand, tagging and keyword-based searches produces results that are oblivious of the contextual relationships between different categories.
I propose that content are tagged by two different kinds of tags: semantic tags and syntactic tags. Semantic tags annotate the underline semantic concepts the particular content deal with, whereas syntactic tags annotate linguistic expressions of that concept. For example, the sport "SOCCER" (a semantic tag) is normally called "soccer" (a syntactic tag) in North America but is also called "football" (another syntactic tag) in other parts of the world. In North America, something else (the sport "FOOTBALL") also has the syntactic tag "football". Searching for "football" should return all content dealing with both the sport "SOCCER" and the sport "FOOTBALL" (the search engine does not yet know if the user is interested in "SOCCER" or "FOOTBALL" or both). The user should be given the choice of zooming in, for example, on the sport "SOCCER" without paging through the list dealing with "FOOTBALL". In addition, the user should be able to zoom out to the concept of "SPORT" so that she can see all content in the "SPORT" category. It can be done by organizing the semantic tags in a multifacted taxonomy. Unlike simple directories, multifaceted taxonomies are multiple directories overlaid on top of each other. The different directories organize the content in different facets (attributes), such as the type, subject, location, time, author, and the publisher of the content (pretty much following the Dublin Core Initiative). Therefore, knowing one or two atributes of the content, such as the type, subject, or location, can quickly narrow the scope of the search. Searching and browsing are combined.
I can give a demo of a work-in-progress (http://www.xinlogic.com) to illustrate the ideas, provided that connection to the internet is available at the meeting. More importlandly, I would like to connect with people who are interested in content taxonomies and semantic search engines. Suggested areas of work include coming up with standards in taxonomy definitions and APIs.
Allen Lin, Xinlogic
I would be interested in this -- and if it makes sense I can demo a work in progress also (http://www.metaxio.com)
Summary
We had a good discussion. Around 20 people participated in the discussion. We introduced ourselves. Talked about why we are or might be interested in taxonomies. Gave some examples of taxonomy uses. Discussed some of the technologies for taxonomies. At the end, we agreed that we would like to pursue the topic futuher and we might want to go through an example of taxonomy construction, if we meet again in the future.
Due to time constraint, the demos were given during Geeks Speak instead of in the session.
Discussion Points
Some of the reasons for taxonomy in content mashups:
- Foundation for understanding.
- Shared data types.
- Filtering.
- Contextual relations.
- Easier to find things.
- Making sense of large amount of data and information.
Examples of taxonomy use:
- IT/feature Configuration
- Application mashups (in addition to data mashups)
Challeges in taxonomy:
- Transformation among taxonomies
- Background differences
Other suggestions and questions:
- Mashup Markup Language
- How does taxonomy help application sequencing?
