Controlling metadata field weighting

Introduction

Funnelback’s ranking algorithm includes settings that control metadata weightings. These settings can be used to adjust the weightings applied to results where the query terms appear within specific metadata classes. This is achieved by setting the sco and wmeta ranking options.

Scoring mode 2

Setting the -sco=2 ranking option provides a way to define the metadata classes that will be considered as part of the ranking algorithm.

By default link text, clicked queries and titles are included (-sco=2[k,K,t]) when scoring mode 2 is enabled. The comma-delimited list of metadata classes to use with sco=2 is defined within square brackets when setting the value.

E.g. apply scoring to the default metadata classes as well as customField1 and customField2.

-sco=2[k,K,t,customField1,customField2]

Metadata weighting

Once scoring mode 2 is enabled separate weightings can be assigned to each defined metadata class using a corresponding wmeta ranking option.

A default weighting of 1.0 is applied to all listed metadata fields except for anchor text (k) and click information (K) which both receive a default weighting of 0.5.

Setting a weighting using the corresponding wmeta option will override the default value.

A larger value provides a bigger upweight.

Individual weights can be applied. For example, change the weighting of the t metadata class (from the default 1.0):

-wmeta.t=0.6

Example

Assume that the following metadata classes are configured for a data source that is included in the search:

  • description

  • author

  • section

  • datePublished

  • dateModified

  • articleText

  • articleTitle

  • articleKeywords

  • articleAbstract

The following ranking options (set as part of the query_processor_options within the results page configuration) could be used to upweight the text within the articleTitle, articleAbstract and articleText metadata classes.

-sco=2[articleText,articleTitle,articleAbstract] -wmeta.articleText=0.8 -wmeta.articleTitle=10

This tells Funnelback to apply metadata weightings to the articleText, articleTitle and articleAbstract metadata classes (the -sco=2 parameter) then apply non-default weightings to articleText and articleTitle.

© 2015- Squiz Pty Ltd