Wednesday, June 11, 2008

Removing Noise Words from a String with XQuery

MarkLogic doesn't offer a way to do stop words (a/k/a suppression lists a/k/a noise words) by default for various reasons -- and I didn't want to block them from being used in searches -- but I was asked to remove them from consideration when using hit highlighting. Here's the code I used to remove a fixed set of noise words from a user's search string.

define variable $NOISE_WORDS as xs:string*
(: \b is a word boundary. This catches beginning,
end, and middle of string matches on whole words. :)
('\bthe\b', '\bof\b', '\ban\b', '\bor\b',
'\bis\b', '\bon\b', '\bbut\b', '\ba\b')

define function remove-noise-words($string, $noise)
(: This is a recursive function. :)
if(not(empty($noise))) then
replace($string, $noise[1], '', 'i'),
(: This passes along the noise words after
the one just evaluated. :)
$noise[position() > 1]
else normalize-space($string)

let $source-string1 := "The Tragedy of King Lear"
let $source-string2 := "The Tragedy OF King Lear These an"
let $source-string3 :=
"The Tragedy of the an of King Lear These of"
let $source-string4 := "The of an of"
(: Need to handle empty result if all noise words,
as in #4 above. :)
let $final :=
remove-noise-words($source-string1, $NOISE_WORDS)
return $final

No comments: