lucenenet-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael Garski (JIRA)" <>
Subject [jira] Commented: (LUCENENET-337) TokenAttribute for Selectively Including Tokens in Length Norm
Date Wed, 10 Mar 2010 22:59:27 GMT


Michael Garski commented on LUCENENET-337:

There is an alternate way to avoid having synonyms add to the length norm of a field, which
is currently in the trunk.

Rather than using the default SImilarity as-is, if you call SetDiscountOverlaps(true) on DefaultSimilarity
overlapping tokens such as synonyms will not be considered for the length norm.

That being said, there are cases where the approach in the patch would work better, such as
when you wish to maintain positional information but specify the length norm to be something
other than the number of tokens such as using synonyms with acronyms.  This is a case I run
into quite a bit with names of musicians and bands.

> TokenAttribute for Selectively Including Tokens in Length Norm
> --------------------------------------------------------------
>                 Key: LUCENENET-337
>                 URL:
>             Project: Lucene.Net
>          Issue Type: Improvement
>            Reporter: Michael Garski
>            Priority: Minor
>         Attachments: LengthNorm.patch
> This patch adds functionality to Lucene.Net that allow a TokenFilter to mark a Token
as not to be included in the length norm calculation through the use of a new TokenAttribute
interface LengthNormAttribute and a corresponding implementation LengthNormAttributeImpl.
 This functionality is useful to prevent the increase of the length norm during synonym injection,
particularly in cases where there are a large number of synonyms in relation to the number
of original tokens.
> Following is an example of how to use the new attribute.
> Within your custom TokenFilter, define a field to persist a reference to the attribute
and set it's value in the constructor.  When a the stream advances to a new Token within the
call to IncrementToken() the value of the IncludeInLengthNorm property of the attribute is
set to false for Tokens which should not be included in the length norm calculation.  It defaults
to true and is reset to true after each Token is consumed within DocInverterPerField.ProcessFields.
> {code:title=CustomTokenFilter.cs|borderStyle=solid}
> public class CustomTokenFilter : TokenFilter
> {
> 	private LengthNormAttribute lnAttribute;
> 	public CustomTokenFilter(TokenStream input) : base(input)
> 	{
> 		this.lnAttribute = (LengthNormAttribute)AddAttribute(typeof(LengthNormAttribute));
> 	}
> 	public override bool IncrementToken()
> 	{
> 		if (input.IncrementToken())
> 		{
> 			// make determination that the token is not to be 
> 			// included in the length norm value
> 			// this example marks all tokens to not be 
> 			// included in the length norm value
> 			this.lnAttribute.IncludeInLengthNorm = false;
> 			return true;
> 		}
> 		else
> 		{
> 			return false;
> 		}
> 	}    
> }
> {code}

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message