Die linguistic engine EXTRAKT ist in dem Suchportal LeXXiNet (früher ExtraktSearch) integriert und sorgt für optimale Suchergebnisse.
TRAPHO allows to carry out a phonetic ("fuzzy") search of words, especially of proper nouns.
By applying the approx. 100 language specific rules within TRAPHO to a given word, this word is translated into an abstract representation. This representation is then compared to the representations of the words which the user wants to search. Names/nouns, whose correct spelling is unknown, or variants of such names or variants produced by typing errors, can be found.
For instance "Deutschland" can be detected, even if the spelling in the document is "teutschland" or "duutschland". If someone searches for "Kournikova", but types in "Cornigowa", this hit can still be found.
In contrast to other phonetic systems, TRAPHO is based on the following idea: the spelling variants of a word depend on the language of the searcher. For instance, a German speaker would write a name which they heard in a different way than an English speaker. If a German speaker searches for "Achternbusch", they would normally key in the same characters, but an English speaker would probably type out "Ekternbush", because this is the way this name is pronounced in English.
Therefore, TRAPHO has different language versions (German, French, English, Latin, Italian, Dutch, Norwegian and Spanish), which can be combined in a cascading way for an even broader search.
Each language version consists of two different phonetic functions, which differ from each other in the strictness of the applied rules, thus making TRAPHO a powerful additional tool within EXTRAKT's whole spectrum of linguistic features.
To illustrate this, we tested the list of the German Soccer league players (Bundesliga). The results can be found here!