KWIC (Keywords in Context) Output


Keywords In Context (KWIC) helps users to quickly scan through search results by listing hits surrounded by their context. eXist provides a KWIC module that is not bound to a specific index or query operation. It but can be applied to query results from all indexes that support match highlighting. This includes the Lucene-based index and the ngram index.

The documentation search function on eXist's home page is a good example. It queries documents written in DocBook format. However, the KWIC module has also been successfully used with different schemas (e.g. TEI) and languages (e.g. Chinese).

Using the Module

The KWIC module is entirely written in XQuery. To use the module, import its namespace into your query (you don't need to specify a location):

import module namespace kwic="";

The easiest way to get KWIC output is to call the kwic:summarize function on an element node returned from a full text or ngram query:

import module namespace kwic="";
for $hit in doc("/db/shakespeare/plays/hamlet.xml")//SPEECH[ft:query(., "'nature'")] 
order by ft:score($hit) descending
    kwic:summarize($hit, <config width="40"/>)

Every call to kwic:summarize will return an HTML paragraph containing 3 span elements with the text before and after each match, as well as the match text itself:

  <span class="previous">
    ... s effect, sir; after what flourish your
  <span class="hi">
  <span class="following">

The <config> element, passed to kwic:summarize (as second parameter) determines the appearance of the generated HTML. It recognizes 3 attributes:


The maximum number of characters to be printed before and after the match


By default kwic:summarize returns an HTML paragraph with spans.

If table="yes" it will return an HTML table row <tr> element. The text chunks will be enclosed in a table column <td> element.


If present, each match will be enclosed within a link, using the URI in the link attribute as target.

Using a callback function for more fine-grained control

If you look at the output of query above you may notice that a space is missing between words if the previous or following chunk extends to a different <LINE> element. And it would also be nicer to display text from <LINE> elements only and to ignore <SPEAKER> or <STAGEDIR> elements. This can be achieved with the help of a callback function:

import module namespace kwic="";
declare function local:filter($node as node(), $mode as xs:string) as xs:string? {
  if ($node/parent::SPEAKER or $node/parent::STAGEDIR) then 
  else if ($mode eq 'before') then 
      concat($node, ' ')
      concat(' ', $node)

for $hit in doc("/db/shakespeare/plays/hamlet.xml")//SPEECH[ft:query(., "'nature'")] 
order by ft:score($hit) descending
kwic:summarize($hit, <config width="40"/>, local:filter#2)

The third parameter to kwic:summarize here is a reference to a function accepting 2 arguments:

  1. A single text node which should be appended or prepended to the current text chunk

  2. A string indicating the current direction in which text is appended: before or after.

The function can return the empty sequence if the current node should be ignored (for instance if it belongs to a footnote which should not be displayed). Otherwise it must return a single string.

The local:filter function above first checks if the passed node has a SPEAKER or STAGEDIR parent. If so, it ignores that node by returning the empty sequence. If not, the function adds a single whitespace before or after the string, so adjacent lines will be properly separated.

Advanced Use

Using kwic:summarize, you will get one KWIC-formatted item for every match, even if the matches are in the same paragraph. Also, the context from which the text is taken is always the same: the element you queried. To get more control over the output, you can directly call kwic:get-summary, which is the module's core function.

kwic:get-summary expects 3 or 4 parameters.

  • The current context root

  • The match object to process

  • Parameters 3 and 4 are the same as for kwic:summarize

Before passing nodes to kwic:get-summary you have to expand them, which basically means to create an in-memory copy in which all matches are properly marked up with <exist:match> tags. The main part of the query should look as follows:

for $hit in doc("/db/shakespeare/plays/hamlet.xml")//SPEECH[ft:query(., "'nature'")]
let $expanded := kwic:expand($hit)
order by ft:score($hit) descending
    kwic:get-summary($expanded, ($expanded//exist:match)[1], <config width="40"/>,

In this example, we select the first <exist:match> only, thus ignoring all other matches within $expanded.

Sometimes you may also want to change the context to restrict the KWIC display to certain elements within the larger query context, for instance paragraphs within sections. The following example still queries <SPEECH> but displays a KWIC entry for each <LINE> with a match, grouped by speech:

for $hit in doc("/db/shakespeare/plays/hamlet.xml")//SPEECH[ft:query(., "nature")]
let $expanded := kwic:expand($hit)
order by ft:score($hit) descending
    <div class="speech">{
        for $line in $expanded//LINE[.//exist:match]
            kwic:get-summary($line, ($line/exist:match)[1], <config width="40"/>,

You might wonder why we don't query <LINE> directly to get a different context, as in:

//SPEECH[ft:query(LINE, "nature")]

This is because Lucene computes the relevance of each match with respect to the SPEECH context, not LINE. If we queried LINE, each single line would get a match score and the matches would end up in a completely different order.

Marking up Matches without using KWIC

Sometimes you don't want to use the KWIC module, but still would like an indication where matches were found in the text. eXist's XML serializer can automatically highlight matches when it writes out a piece of XML. All the matches will be surrounded by an <exist:match> tag.

You can achieve the same within an XQuery by calling the extension function util:expand:

let $expanded := util:expand($hit, "expand-xincludes=no")
return $expanded//exist:match

util:expand returns a copy of the XML fragment it received in its first parameter, which, unless configured otherwise, has all matches wrapped into <exist:match> tags.

Please note that util:expand will not expand matches in Lucene fields. Use ft:highlight-field-matches instead. For more information, see lucene.xml#expand-fields.