I have a xml file that has the word "see" or "see also"(indefinite to any children of the root) which has to be looked. But only it should be considered if the following word to it is a tag, <hi>.
For example:
<p>...We believe in the power of the mind;; see <hi rend="smallcaps"> inspiration</hi>...</p>
or
<p>This situation may be a reflection of a debate... (see also <hi rend="smallcaps">debate</hi>)...</p>
I need to do this because I will manipulate the content of the <hi> element.
Any idea on how to do this? Thanks.

I have already found the solution:
tokenize(normalize-space(preceding-sibling::text()), ' ')[last()]='see'
or
tokenize(normalize-space(preceding-sibling::text()), ' ')[last()]='also'
and
tokenize(normalize-space(preceding-sibling::text()), ' ')[last()-1]='see'

preceding-sibling::text()[1] would be safer. In XSLT 2.0 you'll get an error if there's more than one.

I would do it with matches() and a regex myself, but your solution is OK.

Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference

You're definitely correct with doing the task with match and regex because I encountered a problem with the case when the word "see" is preceded by a character like "(see" or ";see" and the uppercase and lower case of the words. Can you give some code examples on how to do the match and regex. Thanks.

Don't bother anymore, I already found my solution:
<xsl:variable name="tex" as="xs:string*" select="preceding-sibling::text()"/>
<xsl:choose>
<xsl:when test="(tokenize(normalize-space(lower-case($tex[last()])), '\s[^s]?')[last()]='see' and not(ancestor::dummyarticle)) or (tokenize(normalize-space($tex[last()]), ' ')[last()]='also' and tokenize(normalize-space(lower-case($tex[last()])), '\s[^s]?')[last()-1]='see' and not(ancestor::dummyarticle))">
...
</xsl:when>
</xsl:choose>
Thanks anyway..

I think there is a flaw with my work,
<xsl:variable name="tex" as="xs:string*" select="preceding-sibling::text()[1]"/>
<xsl:choose>
<xsl:when test="(tokenize(normalize-space(lower-case($tex[last()])), '\s[^s]?')[last()]='see' and not(ancestor::dummyarticle)) or (tokenize(normalize-space($tex[last()]), ' ')[last()]='also' and tokenize(normalize-space(lower-case($tex[last()])), '\s[^s]?')[last()-1]='see' and not(ancestor::dummyarticle))">
...
</xsl:when>
</xsl:choose>

when the text is " (see", it does not go to the when block, how come? Thanks.