Le 24/02/04 15:59, ��Jose L. Rodriguez Illera�� <[EMAIL PROTECTED]> a �crit�:
> I have a pseudo-concordance tool that calculates the word index, frecuency > and lenght from a text. Then, you may select any word and it calculates all > its contexts with 5 words before and after. If you select one of them it > finds the page where it appears and select the word. May be need small > tunning. I do not know if it is as fast as FreeText, but it is enough for > most uses. If you're interested we have a text analysis code that brings strings to their dictionary entries (singular or infinitive form). It does so by first going through a grammar analysis of the phrase to tell if a string is a verb or noun or... (like "end" in English or "partie" in French, Spanish and German version are in the works). We use it to index documents on what they "talk about" purging text of redundant information necessary for human to human comprehension (I love to read Shannon & Weaver�;-). For example the preceding paragraph could be resumed to only nouns due to it's length and still keep its subject for future search: text analysis code string dictionary form grammar phrase verb noun end English French Spanish German version work document information human comprehension Shannon Weaver. The only meaningful strings missing when you limit to only nouns are verbs and adjectives which in the example would have been "singular, infinitive and index" but you also leave out "be, have, love, read, like, go..." that are rarely searched for, so it is the master's decision to know what he wants to index. Another thing concerning French nouns, a lot don�t have an "s" for a plural form but a totally different string. And bringing everything back to singular form assures that people looking for an "oeil" (eye) will also find documents about "yeux" (eyes). -- Beno�t Leraillez Souvenez-vous qu'on peut �tre herm�tique -- et ne rien renfermer. N'oubliez pas qu'herm�tique cela veut dire aussi ��bouch頻�! (S. Guitry) _______________________________________________ use-revolution mailing list [EMAIL PROTECTED] http://lists.runrev.com/mailman/listinfo/use-revolution
