[
https://issues.apache.org/jira/browse/NUTCH-1037?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13063363#comment-13063363
]
Julien Nioche commented on NUTCH-1037:
--------------------------------------
Deduplicating the anchors would have an impact on the score on the search side
with all the terms having a doc freq of 1. As a result an anchor appearing once
for a doc would yield the same score as one appearing 100 times which is not
right.
A less drastic option would be to pass the number of occurrences for a given
anchor using the payloads format. This is partially supported in SOLR - see
http://digitalpebble.blogspot.com/2010/08/using-payloads-with-dismaxqparser-in.html
for a workaround. This way, we'd send less data to SOLR for indexing while
retaining the information about the frequencies and being able to take that
into account for the scoring
> Deduplicate anchors before indexing
> -----------------------------------
>
> Key: NUTCH-1037
> URL: https://issues.apache.org/jira/browse/NUTCH-1037
> Project: Nutch
> Issue Type: Improvement
> Reporter: Markus Jelsma
> Assignee: Markus Jelsma
> Fix For: 1.4, 2.0
>
> Attachments: NUTCH-1037-1.4-1.patch, NUTCH-1037-2.0-1.patch
>
>
> Anchors are not deduplicated before indexing. This can result in a very high
> number of similar and identical anchors being indexed. Before indexing,
> anchors must be deduplicated at least on case.
> Should this be implemented as a fix or as a new feature that needs to be
> configured?
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
