Specific geographical context-sensitive search engines

The solutions of Eptisa TI (previously Sitesa) identify any text and place it on a map - Bitext has created a tool, NaturalFinder, for intelligent search engines.

El Pais, May 08 2008

In February, Yahoo! had 100 web addresses in its directory. Only 0.24% of the population of the world used Internet. Today, 1,300 million people are connected to the Net and can navigate around over 165 active web sites, according to Netcraft.

Internet stores all kinds of content: texts, documents, photographs, music, databases, maps, etc., though it is not all valid nor, especially easy to find. It is an exponentially growing problem that not even Google solves: you look for the needle and it returns the haystack.

Social searches, guided searches, semantic searches, faceted searches (looking for information from one or more perspectives) or based on natural languages, virtual assistants and artificial intelligence try to solve this problem not just in the macro-universe of Internet, but in environments which though more delimited are also full of content, such as the Administration, banks or large companies.

The Spanish company Eptisa TI is a specialist in geosearches: it facilitates access to information from the geographical point of view. The company has various solutions, such as those of the American company Metacarta, which allow "any non-structured information format to be identified and placed on a map." "Bear in mind that over 75% of searches have a geographic component," says Pedro Torres, Innovation and Marketing Manager.

Their geosearcher, which finds the information on a place, delimited in a given space of time, is used in strategic environments (defence, security, oil companies) where a great deal of information is analysed in a short period of time. It is a complex solution incorporating various technologies, such as "geographic information extraction, visual data mining, and natural language recognition."


NaturalFinder is the solution Eptisa TI incorporates in its geosearcher. It is developed by the Spanish company Bitext. "It is a complement for any search engine that uses spontaneous human language – natural language – and allows questions to be asked directly, without entering keywords," says Antonio Valderrábanos, founder of this company created in 2001 and specializing in organizing contents by their linguistic properties. NaturalFinder is used in the Instituto Nacional de Administraciones Públicas (INAP), which supplies information to civil servants, and the Ministry of Defence, among others.

Today, Isoco, dedicated to information management, and Bitext present the second version of Naveganza. According to Jesús Contreras, Isoco's Director of R+D+i de Isoco, will be capable of "understanding the Internet user and providing an answer to his question." Naveganza is a third-generation search engine. "While traditional search engines work on keywords," explains Contreras, "a semantic search engine relates documents by ideas and meanings. The former provides links containing key terms; the latter, answers."

Isoco, which creates applications to enable machines to understand humans, introduced its first semantic search engine in 2004, for the Real Instituto Elcano. The second version, named Naveganza, incorporated Bitext's natural language and was installed in the file search engine of Zaragoza City Council and for internal use by operators of the 11811 telephone guide.

The second version of Naveganza presented by the two companies today "will have even greater comprehension capacity because it includes all technologies of language and semantics. The goal is for them to become virtual assistants in the near future. For example, Internet users will be able to talk naturally to them to ask for what they want to find," says Contreras.

Antonio Valderrábanos, the philologist founder of Bitext, says that Naveganza will be better than the most popular search engine. "Google is unique in Internet, especially exploiting links to different pages, but is less useful in corporate environments."

Bitext was founded in 2001 and markets its products since 2005. "Our software analyses any format with natural language so that the application understands the content." Bitext supplies four modules for integration in different solutions, and available in Catalan, Spanish, Basque and English: Natural Finder (search engines), Natural Mail (e-mail), Natural GIS (geographic information systems) and Natural Assistant (customer attention). Its clients are the Instituto Nacional de Administraciones Públicas and the Provincial Government of Guipúzcoa "with a version in Basque, a language which is a challenge for any linguist because of its complexity," says Valderrábanos.

It also creates customized solutions, mixing its four products. For example, it and Eptisa TI are introducing an affiliate search engine in the 060 network which centralizes 300 web sites of the central administration and links with those of regional and local authorities. The affiliate search engine combines the functionalities of Google Search Appliance and those of its Natural Finder.

In practice, says Pedro Torres of Eptisa TI, "If the Internet user is searching on the website of the Institute of Employment, he will be able to use colloquial rather than bureaucratic language and the results will include those of other institutions of the central administration."

For the moment, these companies are involved in corporate environments, but as Torres says, "these technologies can be integrated in any search engine."

 Specific geographical context-sensitive search engines