Thanks for your answer.

Ok, so here is my problem : for french WB, searching for the "red" word in
"red/white" is ok. But, if there is a space after the '/' char (like in
"red/ white"), it won't match..... There are tons of such example so that
frankly, this is a sort of bug for me, but it's not the subject here :)
I could use neutral WB, but then there are a lot other complications and
this is not an option.

Currently, I have to use a program of mine that :
1) export a column from a table into a file
2) for that file do some transformations (cut accents etc...)
3) import this "new" column in table dedicated to full-text
I don't really wan't to spend my time with that anymore; I have to replicate
data because of some strange WB behaviour.

So my real problem is that our developpers are going crazy if there is not a
simple rule for word breaking.
My idea is then to build my own WB, so that I can decide __exactly__ when to
break or not.
By the way, is there "somewhere" the real specification/algorithm for the
different languages WB ??

From your answer, I understand I have to implement IWordBreaker interface.
Problem is I 've had built only one COM IFilter in the past, and from what I
remember, it did not work very well :)
So, I'm not familiar with COM programming but I understand it is the way to
go.

Is there any other simple sample someone can share please ? MS samples are
good, but do not focus enough on what is important IMO.

Thanks again,

Sylvain,
DBA.