Identify that 2 text deal with the same subject

How could we do that?

[ticatoc]

Hello,

I 'm looking for a script that could say if 2 texts are similars, if they are about the same subject.
I'm looking for that to try to create my own memtracker script (that I 'd like to be GPL :-) )...
Could you help me please?

[jatar_k]

for similar, as in chars, you could use levenshtein
[php.net...]

for whether they sound similarly you could use soundex
[php.net...]

for deriving the subject of a string you are looking at something significantly more complex, the search engines do this so maybe you could find a search engine script somewhere.