contains and tokenize

[David Carlisle]

you could use matches instead of contains and add  some word boundaries
to your regex. Schema/xpath regexp don't have built in word boundary
excpressions, but depending on your language rules something like

matches($text,concat('(^|[ ,\.;])',$keyword,($|[ ,\.;])'),'i')
for a case insensitive match against the keyword combined with either
the beginning or end of a string or a space , or ;.

If you want to use tokenize then you want = not contains, something like

upper-case($keyword) = tokenize(upper-case($text),'(\s|[\.,;])+'

which returns the sequence of words and then tests if any of them is
equal to the keyword. 

(There are some differences, a case insenitive match for example isn't
the same thing as upper casing and then comparing, see 
http://www.w3.org/TR/xpath-functions/#flags
)

David