edu.washington.cs.knowitall.extractor.conf.featureset
Class ExtractionFeature

java.lang.Object
  edu.washington.cs.knowitall.extractor.conf.featureset.ExtractionFeature

All Implemented Interfaces:

com.google.common.base.Predicate<ChunkedBinaryExtraction>

Direct Known Subclasses:

ChunkFeature, PosFeature, TokenFeature, VerbTokenFeature

public abstract class ExtractionFeature
extends Object
implements com.google.common.base.Predicate<ChunkedBinaryExtraction>

A parent class for any feature that picks a particular range and applies a test to all indices within that range. For example, the feature `return true if arg2 contains token "fish"` would be implemented by having rangeToExamine() return arg2.getRange() and testAtIndex() returning sentence.getToken(index).equalsIgnoreCase("fish");

Author:

Rob

Field Summary

protected BasicFieldNormalizer

stemmer

Constructor Summary

protected

ExtractionFeature()

Method Summary

boolean	apply(ChunkedBinaryExtraction cbe)
static Integer	indexOfHeadVerb(ChunkedExtraction relation, boolean exception) Implements the following naive algorithm for locating the head verb within a verb phrase: 1.
protected abstract edu.washington.cs.knowitall.commonlib.Range	rangeToExamine(ChunkedBinaryExtraction cbe)
protected abstract boolean	testAtIndex(Integer index, ChunkedSentence sentence)

Methods inherited from class java.lang.Object

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

Methods inherited from interface com.google.common.base.Predicate

equals

Field Detail

stemmer

protected BasicFieldNormalizer stemmer

Constructor Detail

ExtractionFeature

protected ExtractionFeature()

Method Detail

rangeToExamine

protected abstract edu.washington.cs.knowitall.commonlib.Range rangeToExamine(ChunkedBinaryExtraction cbe)

apply

public boolean apply(ChunkedBinaryExtraction cbe)

Specified by:

apply in interface com.google.common.base.Predicate<ChunkedBinaryExtraction>

testAtIndex

protected abstract boolean testAtIndex(Integer index,
                                       ChunkedSentence sentence)

indexOfHeadVerb

public static Integer indexOfHeadVerb(ChunkedExtraction relation,
                                      boolean exception)

Implements the following naive algorithm for locating the head verb within a verb phrase:

1. Start at end of phrase.
2. Work backward until you encounter a posTag.startsWith("V");
3. Return the corresponding token.
If exception, throw an IllegalArgumentException if no verb in the relation. If !exception, return null to the client if no verb in the relation.

Parameters:

relation -