[
https://issues.apache.org/jira/browse/LUCENE-2587?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13145052#comment-13145052
]
Steven Rowe commented on LUCENE-2587:
-------------------------------------
Hi Roberto,
I've looked at your patch, and you've done a good job of finding the problem,
fixing it, and providing a test that fails before your fix and succeeds
afterward.
However, I see a problem. Your fix changes fragments' beginning offsets to
exclude preceding inter-fragment non-tokenized characters, but this causes
adjacent fragments to be non-contiguous (contiguity is defined as
frag1.end==frag2.start). As a result, your fix prevents adjacent fragments
from being merged. (See {{mergeContiguousFragments()}}, which is called from
{{getBestTextFragments()}}.)
Ideally, another test should be added that checks for the problem your current
fix causes (adjacency != contiguity), and then your fix should be changed to
enable contiguous fragment merging.
> Highlighter picks wrong offset for fragment boundaries
> ------------------------------------------------------
>
> Key: LUCENE-2587
> URL: https://issues.apache.org/jira/browse/LUCENE-2587
> Project: Lucene - Java
> Issue Type: Bug
> Components: modules/highlighter
> Affects Versions: 3.0.2
> Environment: Java 6 + Lucene 3.0.2
> Reporter: Terje Eggestad
> Priority: Trivial
> Labels: newdev
> Attachments: IMSentenceFragmenter.java, LUCENE-2587.patch,
> TestIMSentenceFragmenter.java
>
>
> I have written a new Fragmenter since we need fragments for hitlines to be
> on sentence boundaries and not cross paragraphs.
> When using it with org.apache.lucene.search.highlight.Highlighter, I get
> hitlines that starts with ". ", "? ", "! "...
> Consider the text "A b c d e. F g h i j! K l m n o. "
> which become the tokenstream : (A) (b) (c) (d) (e) (F) (g) (h) (i) (j) (K)
> (l) (m) (n) (o)
> If the fragmenter return isNewFragment() = true on F and K and Highlighter
> pick the middle fragment, lets say we search on "g" the hitline becomes:
> ". F <B>g</B> h i j"
> The reason, it seems, is that the offset to the fragment boundaries found by
> taking the endOffset of the last token in a fragment ,
> not the startOffset of the first.
> TJ
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
