There have been numerous startups in the semantic search space. Microsoft acquired one -- Powerset, which now powers Bing. Meet Pragmatech, founded in 2009 as a wholly owned subsidiary of the United Development Company (UDC), based in Doha, Qatar. It has since opened a research and development branch in Jounieh, Lebanon, just outside the capital, Beirut.
The company's engineers define semantic technology as a set of techniques, tools, and algorithms intended to enable intelligent filtering, querying, searching and understanding of unstructured context by meaning rather than keywords. For example, having a "blast" at David's birthday party is quite different than a powerful "blast" ripping through a building. A semantic search engine can ferret out the difference in contextual meanings.
With a semantic search engine as the basis for a set of tools, Pragmatech engineers plan to build and place application layers on top of the semantic search engine. The first, a news-filtering feature called ctrl news, allows people to sign up to receive information about specific topics based on context rather than keywords. The free service allows users to extract key topics from documents when searching for metadata, and automatically generates a summary because it understands the content of the document. It can identify organizations, brands, products, and more.
On the back end, documents get processed, and turned from words into meanings into compound concepts that are analyzed to determine the key topics in order of importance. The concepts in articles match other topics, regardless of the words being used in the content.
This means the engine supports topic-based retrieval -- not just words or concepts, but rather composite meanings categorized as objects, according to Walid Saba, who leads the core research and development team. The engine also produces a summary of the content because it understands the context. The engine concentrates on meaning, topics and key topics to understand the true meaning.
ctrl news allows users to input words to track specific article topics. A list of stories arrives in an email each morning describing "Today's relevant news to your selected subject(s) (selected from 2520 articles collected during the past 24 hrs)" The technology automatically generates a summary for each article, performs entity extraction for structured queries, and retrieves documents based on related topics.
Aside from the semantic news aggregator, Pragmatech will launch a test site for advertisers. Saba says the advertiser will give the company an ad, along with a variety of URLs. The technology will identify the best publisher sites to run specific ads based on the content. The technology will match the Web site with the ad content based on the contextual relevance of the advertisement, even if the ad and Web page are in two different languages.
For instance, if someone places an ad with content in English on a French language Web site, the ad will serve up because it translates the content both on the page and in the ad.
Saba focuses on semantic technology, with expertise in artificial intelligence for the past 20 years. He gathered a team of experts a few years ago, and has been working for the past two years to develop a semantic search engine.
Led by Saba, the core research and development (R&D) team selected a group of IT professionals with graduate degrees in computer science and experience in research, academia and industry in Europe, Canada, U.S. and the Middle East. Members previously held positions at AT&T Bell Labs, IBM Research, Nortel Networks, American Institutes for Research, CHEO Research Institute, Cognos, Cataphora, AIRBUS, and Amadeus.