We are in section:

Semantic search example

This page shows an example of the semantic search engine built by the Esperteyu Project. This search tool has indexed +20,000 legal documents published by Boletín Oficial del Principado de Asturias. The contents of these documents are of general interest, but they are written in legal and specialized jargon.

Syntactic search

To better show the advantages of the semantic search, we will first describe a syntactic search. As our documents are written in Spanish, both queries and results will be shown in Spanish with appropiate translations.

Let the user execute a query entering the word "viejo" (old). Of course, the search process returns some results, but they are limited and not very meaningful. In the common language, "viejo" is not just an adjective, but also a common word to refer to elderly people. Legal documents do not use this noun, as it is considered disrespectful. As a consequence, results do not match user's expectatives.

Results page is shown below:

Screenshot showing the results of a syntactic search

  1. The actual terms used in the matching. This is not the case in our example, but usually words can be singled out or even removed (stop words).
  2. The concepts recognized in the user's query. These are the concepts in the ontologies that best fit the underlying intentions of the query. In the image, there is one concept, "persona mayor" (the elders), which is the term actually used in the documents.
  3. Suggested terms: several words selected by their syntactical proximity with "viejo". This words may help the user to fix a miss-spelling.
  4. Results. As this is a syntactical search, all the documents in this list contain the word "viejo". In many cases, the documents have little relation with the intention of the query, as "viejo" is also a surname and a place name.
  5. Search mode. The selected mode (the first one) is the syntactic search mode. By using any of the two remaining modes, the user can change to the semantic search mode.

Semantic search

Semantic search allows the user to find revelant documents about a topic, even if the documents do not contain the query terms. Our semantic search tool takes one or several concepts and a context. In our example, the concept is "persona mayor" (the elders) and the context is "accesibilidad" (accessibility).

The page structure is very similar to the one previously described, with some minor differences:

Screenshot showing the results of the semantic search

  1. Concepts actually used in the query. This box enumerates several concepts from the ontologies that are semantically related with the original concept entered by the user, such as "ciudadano" (citizen) or "ayuda técnica de accesibilidad" (devices for people with dissabilities).
  2. Related terms. Some words semantically related with with original concept entered by the user, such as "anciano" (the elders).

In the main area of the page, results are shown. These results are semantically related with the query, even if they do not contain any of the terms typed by the user. In the example, most of the results are documents about "residencias para personas mayores" (old people's home).

Menu

Conforms with W3C XHTML 1.0 | CSS 2.0 | WAI 'Double A'