STILUS Fuzzy

Fuzzy search

The search function incorporated into the systems of information (databases, collections of documents, etc.) aims at finding the most relevant elements, similar to a certain search term (entered in the system by the users). For this purpose, the term is compared to an index constructed from different processes and sources of information, linking the terms considered as relevant and the associated documents.

If the correspondence is not perfect, the systems generally give an empty answer, indicating that they have not found any search term. This causes the users' dissatisfaction, as the need for information is not met efficiently.

However, a system of fuzzy search can process more profoundly the search terms and solve this lack of results. For this purpose, possible alternatives are built to correct common spelling mistakes, calculating the similarity between the search terms and the indexing terms and offering possibilities of selection to the user (such as "perhaps you wanted to say..."), with the possibility of using synonyms of the terms of the original search. In this way, the system manages to refine the search and anticipate the users' wishes.

STILUS Fuzzy is the system of fuzzy search developed by Daedalus that can be used by itself (as an independent executable programme) or only integrated into a system of production to complete its functionality.

To generate the search suggestions, STILUS Fuzzy is based on a dictionary of terms: a list of words in a certain language, people's names, organizations, places, etc. From the version 2.0, it incorporates multidictionary support that is it makes it possible to use simultaneously several dictionaries.

The STILUS Fuzzy terms can have one or several linked alias, used to include easily the different inflectional forms of a word, ("comer" -> alias "comiendo", "comido"...), synonyms or other linked terms ("policía" -> alias "comisaría", "cuartel", "prefectura", "guardia civil"...).

Besides, it is possible to incorporate lists of stop words (ignored in the processing) and frequent mistakes (usual suggestions of substitution) in order to improve the performance.

STILUS Fuzzy is available for Windows platforms such as Unix, in any of their variants, and in particular for Linux with their different deliveries, without establishing any restriction as for the version of the nucleus and the C standard libraries (libc).

In the Daedalus demo website a service of fuzzy search on song titles based on STILUS Fuzzy is included.