Use SKOS for Search Term Suggestion

We have added a new feature in gnizr 2.3.0 that suggests search terms based on user-defined SKOS tag relation hierarchy. Search is an important feature in gnizr, which provides a means for the users to find bookmarks that are created by others in the community. To help users to find information, we must solve two problems: 1) allow users to find information beyond resources that exist within a gnizr community, 2) provide search assistance to users when navigating through a massive amount of information.

We attempt to solve the first problem by embracing OpenSearch technology, which is a powerful way to mashup search results from multiple search engines. To solve the second problem, we are experimenting a Semantic Web approach that exploits relationships between tags.

This approach works as the follows. When a gnizr web application is deployed for a specific community, the system administrator defines a domain ontology for the community in terms of tags and tag relations. In gnizr this is achieved by editing tag relations. When a user conducts search, the search system analyzes the input search term (treating it as a list of tags) and recommends new search terms (i.e., tags) that are either related-to, is narrower than or is broader than the input search term.

Here is a set of screenshots of an actual use case.

NOTE: when you're at the flickr pages, click on the "All Sizes" icon to see a high-resolution image of the screenshot.

Domain Ontology for 2008 US President Election

Let's assume that we are deploying a gnizr web application for sharing bookmarks related to the 2008 US President Election. The system administrator 'gnizr' defines a simple domain ontology in terms of tags and tag relations.

democratic
related to: republican, government, liberal
narrower: clinton, edwards
broader: politics

republican
related to: conservative, moderate
narrower: anti:clinton, pro:Giuliani, pro:Romney

clinton:
related to: anti:cliton, edwards, election
broader: person

Screenshots:

• Create/Edit tag relations
• Defining the narrower relations for tag 'democratic'
• Defining the related-to relations for tag 'democratic'

User searches bookmarks on 'democratic'

Gnizr user "Harry Chen" does a search on the term 'democratic'. Using the OpenSerach interface, results from multiple search engines are mashed-up in a single page. On the top of search result, he finds a set of system suggested search terms -- i.e., terms related-to, is narrower than and is broader than 'democratic'.



Screenshots:

• Following the suggestion, Harry finds bookmarks related to 'republican'
• Looks up other resources on the Yahoo! Web Search
• All 'republican' -- OpenSearch result mashup
  

What's Next?

The described feature is still in its early stage of development. It will be available for evaluation in our next milestone release gnizr-2.3.0-M2. Let's us know what you think.