BlueBlog: The 5 Types of Semantic Applications: Part 1

June 3rd, 2008

The 5 Types of Semantic Applications: Part 1 - Semantic Search

by Fraser

Last week I spoke at the TechCocktail Conference in Chicago on Understanding the Semantic Web. It was a successful event and my session was well received, accomplishing the goal of communicating a general understanding in a simple way.

Over the coming days I’ll summarize the elements from one section of the presentation: The 5 Types of Semantic Applicaitons.

The semantic web is an emerging market that’s the focus of great interest, hype, and confusion. In an attempt to clarify what’s happening in the space it’s helpful to examine the five basic application types that exist and how companies are entering the market using either the top-down or bottom-up approach to semantics.

Part 1 - Semantic Search:

The semantic search application type is the best known - and most misunderstood - semantic application. Today there are a number of differing approaches working towards the common goal of providing a search experience that benefits from meaning and context.

Bottom-up approach:

The level of hype and misaligned expectations (around timing and performance) for semantic search applications using the bottom-up approach has created a confused early-adopter base.

While semantic search engines provide improved results with complex queries that require inference and reasoning, results will not be improved when a basic query can be addressed via statistical frequency (”Which bacteria is resistant to antibiotics?” vs. “Who directed Pulp Fiction?”). Sadly, semantic search engines will not solve queries that require more than a semantic understanding.

Powerset and Hakia are examples of bottom-up semantic search companies. Currently Hakia allows you to search the entire web but Powerset constrains the search to Wikipedia articles.

Another benefit of semantic search is the ability to introduce context to the user interface. Existing search engines already do this for specific queries:

However, semantic search engines provide a more contextually relevant search experience, introducing a contextual UI to search results and follow-on interactions with the search engine itself.

In summary, bottom-up semantic search engines hold a lot of promise and the benefits can be - simplistically - summarized as: (i) providing results for complex queries that require inference and reasoning; (ii) introducing a contextual UI to search results and interaction with the search engine.

Note: this section of the session benefited significantly from Alex’s article on Semantic Search, which provides a deeper examination of the space.

Top-down approach:

There are a number search engines using a top-down approach to provide an improved search experience around a specific vertical.

Because the context is fixed by the vertical of interest the companies can parse a subset of information on the web to provide relevant data and results. By fixing the context via the vertical it also becomes easier to provide a UI that delivers a contextually better experience for the query.

Spock, a vertical search engine for people, is a good example of a top-down semantic search company and highlights the benefits of vertical search. Compare the results of a search on Spock vs. a search on Google for the same person.

Not only is the result better - the actual person vs. online approximations for the individual, but Spock’s UI also provides a better experience for consuming the information. Spock’s results include a photo and bio of the person as well as basic information (age, sex, location) and a contextual UI for refining search results (filter by news, related people, etc.).

Vertical search engines that leverage top-down semantics provide an easier path towards semantic search. By restraining context via the specific vertical of interest, and parsing specific data that exists across the web, they’re able to provide a compelling contextual search experience. Similarly to bottom-up semantic search engines, the ability for vertical search companies to infer meaning in the query enables them to provide a better search experience by returning results within a contextual UI.

Subscribe to this blog