[xsl] Re: Similarity metric in XSLT 2?

[Martin Holmes <mholmes@xxxxxxx>]

On 12-03-30 01:24 PM, Imsieke, Gerrit, le-tex wrote:
> I can only affirm that I'd be interested in such a library, too.
>
> The last time that I needed string similarity metrics (4 yrs ago), I
> used Perl with XML::LibXML and String::Similarity.
>
> If there were such a module / extension function for XPath / XSLT, I'd
> probably used it more often. If you find a Java library that is easy to
> interface with from Java-based XSLT processors, please let me know. I
> think that Levenshtein or more advanced algorithms will be too slow when
> implemented in XSLT, but may be readily available as an extension function.

I once implemented the Universal Similarity Metric (Normalized 
Compression Distance) in Pascal and Java:

<http://dh2010.cch.kcl.ac.uk/academic-programme/abstracts/papers/html/ab-693.html>

and found that it was surprisingly effective for short strings, as well 
as being very fast. I might look at figuring out how to call the Java 
library from Saxon. Implementing the metric was trivial.

Cheers,
Martin

> Gerrit
>
> On 2012-03-30 20:18, Martin Holmes wrote:
> > Hi all,
> >
> > I'm faced with a situation in which I have to match an input string
> > against a set of possible candidates, and I need to find the match which
> > is most similar to it (I'm trying to identify correspondences between
> > two sets of files which have similar, but not identical, content).
> >
> > Has anyone done anything like measuring string similarity in XSLT 2.0?
> > If so, how did you approach it?
> >
> > All help appreciated,
> > Martin