edu.washington.cs.knowitall.normalization

Class RelationString

java.lang.Object

edu.washington.cs.knowitall.normalization.RelationString

All Implemented Interfaces:

Comparable<RelationString>

public class RelationString
extends Object
implements Comparable<RelationString>

Represents a Reverb predicate that contains (1) the original predicate from the text (2) normalized version (3) POS sequence (modified by Michael Schmitz)

Author:

Jonathan Berant

Constructor Summary

Constructors

Constructor and Description
RelationString(String m_pred, String m_normPred, String m_posPred)

Method Summary

Methods

Modifier and Type	Method and Description
int	compareTo(RelationString o)
void	correctNormalization() Normalizes the predicate - takes predicate normalized by TextRunner and applies rule based normalization: if predicate doesn't start with a letter or a abbreviated aux - return empty predicate transform to lower case omit adverbs (RB pos-tag) omit quasi-modals: "used to VB", "ought to VB", "be able to VB", "have to VB" omit modals: (MD pos-tag) handle "have": change "have VBN/VB/VBD" to "VBD" handle "be": reverse TextRunner normalization of "be VBN" to unnormalized form, omit "be" in "be VBG" unless is "be going to VB", which is treated like a quasi-modal handle "do": omit "do" in "do VB" expressions handle "have" again to solve some extreme cases where it needs to be applied twice ('d have had to have been doing") if what's left is "be" or "do" or "be be" or "do do" - return empty predicate since this predicate is not informative if what's left doesn't start with a letter or a abbreviated aux - return empty predicate since the extraction was bad if what's contains "be be " or "do do " or "have have " - return empty predicate since the extraction was bad
boolean	equals(Object obj)
String	getNormPred()
String	getPosPred()
String	getPred()
int	hashCode()
static void	main(String[] args)
void	setNormPred(String m_normPred)
void	setPosPred(String m_posPred)
void	setPred(String m_pred)
void	setRelationString(RelationString other)
String	toString()

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Constructor Detail

RelationString

public RelationString(String m_pred,
              String m_normPred,
              String m_posPred)

Method Detail

setRelationString

public void setRelationString(RelationString other)

getPred

public String getPred()

setPred

public void setPred(String m_pred)

getNormPred

public String getNormPred()

setNormPred

public void setNormPred(String m_normPred)

getPosPred

public String getPosPred()

setPosPred

public void setPosPred(String m_posPred)

toString

public String toString()

Overrides:

toString in class Object

hashCode

public int hashCode()

Overrides:

hashCode in class Object

equals

public boolean equals(Object obj)

Overrides:

equals in class Object

compareTo

public int compareTo(RelationString o)

Specified by:

compareTo in interface Comparable<RelationString>

correctNormalization

public void correctNormalization()

Normalizes the predicate - takes predicate normalized by TextRunner and applies rule based normalization:

1. if predicate doesn't start with a letter or a abbreviated aux - return empty predicate
2. transform to lower case
3. omit adverbs (RB pos-tag)
4. omit quasi-modals: "used to VB", "ought to VB", "be able to VB", "have to VB"
5. omit modals: (MD pos-tag)
6. handle "have": change "have VBN/VB/VBD" to "VBD"
7. handle "be": reverse TextRunner normalization of "be VBN" to unnormalized form, omit "be" in "be VBG" unless is "be going to VB", which is treated like a quasi-modal
8. handle "do": omit "do" in "do VB" expressions
9. handle "have" again to solve some extreme cases where it needs to be applied twice ('d have had to have been doing")
10. if what's left is "be" or "do" or "be be" or "do do" - return empty predicate since this predicate is not informative
11. if what's left doesn't start with a letter or a abbreviated aux - return empty predicate since the extraction was bad
12. if what's contains "be be " or "do do " or "have have " - return empty predicate since the extraction was bad

Throws:

IllegalArgumentException

main

public static void main(String[] args)