Package de.julielab.genemapper

Class GeneMentionFilter

java.lang.Object

de.julielab.genemapper.GeneMentionFilter

public class GeneMentionFilter
extends Object

Constructor Summary

Constructors

Constructor	Description
GeneMentionFilter()

Method Summary

Modifier and Type	Method	Description
static void	intersectionFilter(de.julielab.geneexpbase.genemodel.GeneDocument document, de.julielab.geneexpbase.TermNormalizer normalizer, boolean onlySortOutEmptyCandidateLists)	Must be done after all genes have received their candidates.
static void	preFilter(de.julielab.geneexpbase.genemodel.GeneDocument document)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

GeneMentionFilter

public GeneMentionFilter()

Method Detail

preFilter

public static void preFilter(de.julielab.geneexpbase.genemodel.GeneDocument document)

intersectionFilter

public static void intersectionFilter(de.julielab.geneexpbase.genemodel.GeneDocument document,
                                      de.julielab.geneexpbase.TermNormalizer normalizer,
                                      boolean onlySortOutEmptyCandidateLists)
                               throws IOException

Must be done after all genes have received their candidates. Preliminary experiments: This does a bit good for precision, a bit more bad for recall. Overall F-score on IGN train is a bit less with this on.

A variant is to perform the filtering, but in the end, only reject those mentions whose candidate list is left completely empty and leave all other lists untouched. This helps a tiny bit on IGN train.

Parameters:

document - The gene document.

normalizer - The used term normalizer from the mapping core or disambiguation instance.

onlySortOutEmptyCandidateLists - Use the second strategy depicted above.

Throws:

IOException