extract words of a certain language out of an xml file - xslt-3.0

Question

given the following xml (which of course consists of many records),
i would like to output unique values out of it, and also generate a report, that would have the records each word was found.
<collection>
<record>
<controlfield tag="001">1</controlfield>
<datafield tag="200" ind1="1" ind2=" ">
<subfield code="a">Metafore po</subfield>
<subfield code="e">Δοκίμια</subfield>
<subfield code="f">Περικλής αρχαία Ελλάδα</subfield>
</datafield>
<datafield tag="210" ind1="|" ind2="|">
<subfield code="a">Η Αθήνα</subfield>
<subfield code="c">Νοέμβριος</subfield>
<subfield code="d">1999</subfield>
</datafield>
<datafield tag="215" ind1=" " ind2=" ">
<subfield code="a">263 s.</subfield>
</datafield>
<datafield tag="606" ind1="|" ind2=" ">
<subfield code="3">250000087120140311174609</subfield>
<subfield code="a">Πλάτων ιστορία</subfield>
</datafield>
<datafield tag="700" ind1=" " ind2="1">
<subfield code="3">200000000120140228092156</subfield>
<subfield code="4">070</subfield>
<subfield code="a">Liper</subfield>
<subfield code="b">Berit von der</subfield>
</datafield>
</record>
<record>
<controlfield tag="001">here text may also exist</controlfield>
<datafield tag="200" ind1="1" ind2=" ">
<subfield code="a">Metafore po</subfield>
<subfield code="e">Δοκίμια</subfield>
<subfield code="f">Περικλής</subfield>
</datafield>
</collection>
desired output (xml format, or whatever is more easily achieved)
Δοκίμια: 1, here text may also exist
Περικλής: 1, here text may also exist
αρχαία: 1
Η: 1
etc...
regex i have tried with:
/[Α-Ωα-ω]{1,}/

Answer 1

It seems you can treat that like a grouping problem:
<xsl:template match="collection">
<xsl:where-populated>
<ul>
<xsl:for-each-group select="record" group-by="datafield/subfield!tokenize(., '\s')[matches(., '\p{IsGreek}')]">
<li>
{current-grouping-key()} : <xsl:value-of select="current-group()/controlfield" separator=", "/>
</li>
</xsl:for-each-group>
</ul>
</xsl:where-populated>
</xsl:template>
https://xsltfiddle.liberty-development.net/gWmuiKi/1 outputs
<ul>
<li>
Δοκίμια : 1, here text may also exist
</li>
<li>
Περικλής : 1, here text may also exist
</li>
<li>
αρχαία : 1
</li>
<li>
Ελλάδα : 1
</li>
<li>
Η : 1
</li>
<li>
Αθήνα : 1
</li>
<li>
Νοέμβριος : 1
</li>
<li>
Πλάτων : 1
</li>
<li>
ιστορία : 1
</li>
</ul>
that way.
Of course identifying a "word" by simply tokenizing on white space is going to fail in mosts texts and languages, due to punctuation characters and language specific rules. But XSLT/XPath/XQuery regular expressions don't have a word break metacharacter anyway so somehow one has to use tokenize or analyze-string.