lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrzej Bialecki>
Subject Re: Indexing single words and marked phrases
Date Sat, 03 Mar 2007 09:24:09 GMT
Mike O'Leary wrote:
> I am working on a project with team that is developing a named entity
> recognizer. I need to configure Lucene indexing so that it indexes the
> individual words in the text that the named entity recognizer outputs as
> well as the phrases that it marks. For example, in a string like
> <BODY>"The population of <CITY>New York City</CITY> is not as large
as that
> of <CITY>Mexico City</CITY>."</BODY>
> I would want to index the words "New", "York" and "City" and "Mexico" and
> "City" as well as the phrases "New York City" and "Mexico City", along with
> the other non-stopwords in the string in a field labeled "BODY". Would it be
> better to write an Analyzer that can do this, or to adjust the XML parser
> that I am using so that it adds the text within tag pairs like <CITY> and
> </CITY> to the field that corresponds to the tag that is one level up. If it
> is better to write an Analyzer, could someone point me to information on how
> to do this? Thanks.

Please take a look at the analyzer implementation that Nutch uses 
(org.apache.nutch.analysis.NutchDocumentAnalyzer), which uses 
CommonGrams to detect which tokens need to be output as phrases. In your 
case you already know what are the phrases you want to output as tokens, 
but the idea is still the same - the original token stream is modified 
to output both individual terms and phrases.

Best regards,
Andrzej Bialecki     <><
 ___. ___ ___ ___ _ _   __________________________________
[__ || __|__/|__||\/|  Information Retrieval, Semantic Web
___|||__||  \|  ||  |  Embedded Unix, System Integration  Contact: info at sigram dot com

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message