edu.stanford.nlp.trees

Class CoordinationTransformer

java.lang.Object

edu.stanford.nlp.trees.CoordinationTransformer

All Implemented Interfaces:

TreeTransformer, java.util.function.Function<Tree,Tree>

public class CoordinationTransformer
extends java.lang.Object
implements TreeTransformer

Coordination transformer transforms a PennTreebank tree containing a coordination in a flat structure in order to get the dependencies right.
The transformer goes through several steps:

• Removes empty nodes and simplifies many tags (DependencyTreeTransformer)
• Relabels UCP phrases to either ADVP or NP depending on their content
• Turn flat CC structures into structures with an intervening node
• Add extra structure to QP phrases - combine "well over", unflattened structures with CC (QPTreeTransformer)
• Flatten SQ structures to get the verb as the head
• Rearrange structures that appear to be dates
• Flatten X over only X structures
• Turn some fixed conjunction phrases into CONJP, such as "and yet", etc
• Attach RB such as "not" to the next phrase to get the RB headed by the phrase it modifies
• Turn SBAR to PP if parsed as SBAR in phrases such as "The day after the airline was planning ..."
• Rearrange "now that" into an SBAR phrase if it was misparsed as ADVP
• (Only for universal dependencies) Extracts multi-word expressions and attaches all nodes to a new MWE constituent

Author:

Marie-Catherine de Marneffe, John Bauer, Sebastian Schuster

Constructor Summary

Constructors

Constructor and Description
CoordinationTransformer(HeadFinder hf)
CoordinationTransformer(HeadFinder hf, boolean performMWETransformation) Constructor

Method Summary

Modifier and Type	Method and Description
static Tree	CCtransform(Tree t) Transforms t if it contains a coordination in a flat structure
static Tree	gappingTransform(Tree t)
static void	main(java.lang.String[] args)
static Tree	MWETransform(Tree t) Puts all multi-word expressions below a single constituent labeled "MWE".
static Tree	prepCCTransform(Tree t)
static Tree	removeXOverX(Tree t)
Tree	SQflatten(Tree t) Removes the SQ structure under a WHNP question, such as "Who am I to judge?".
Tree	transformTree(Tree t) Transforms t if it contains a coordination in a flat structure (CCtransform) and transforms UCP (UCPtransform).
static Tree	UCPtransform(Tree t) Transforms t if it contains an UCP, it will change the UCP tag into the phrasal tag of the first word of the UCP (UCP (JJ electronic) (, ,) (NN computer) (CC and) (NN building)) will become (ADJP (JJ electronic) (, ,) (NN computer) (CC and) (NN building))

Methods inherited from class java.lang.Object

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

Methods inherited from interface edu.stanford.nlp.trees.TreeTransformer

apply

Methods inherited from interface java.util.function.Function

andThen, compose, identity

Constructor Detail

CoordinationTransformer

public CoordinationTransformer(HeadFinder hf)

CoordinationTransformer

public CoordinationTransformer(HeadFinder hf,
                               boolean performMWETransformation)

Constructor

Parameters:

hf - the headfinder

performMWETransformation - Parameter for backwards compatibility. If set to false, multi-word expressions won't be attached to a new "MWE" node

Method Detail

transformTree

public Tree transformTree(Tree t)

Transforms t if it contains a coordination in a flat structure (CCtransform) and transforms UCP (UCPtransform).

Specified by:

transformTree in interface TreeTransformer

Parameters:

t - a tree to be transformed

Returns:

t transformed

SQflatten

public Tree SQflatten(Tree t)

Removes the SQ structure under a WHNP question, such as "Who am I to judge?". We do this so that it is easier to pick out the head and then easier to connect that head to all of the other words in the question in this situation. In the specific case of making the copula head, we don't do this so that the existing headfinder code can easily find the "am" or other copula verb.

removeXOverX

public static Tree removeXOverX(Tree t)

UCPtransform

public static Tree UCPtransform(Tree t)

Transforms t if it contains an UCP, it will change the UCP tag into the phrasal tag of the first word of the UCP (UCP (JJ electronic) (, ,) (NN computer) (CC and) (NN building)) will become (ADJP (JJ electronic) (, ,) (NN computer) (CC and) (NN building))

Parameters:

t - a tree to be transformed

Returns:

t transformed

CCtransform

public static Tree CCtransform(Tree t)

Transforms t if it contains a coordination in a flat structure

Parameters:

t - a tree to be transformed

Returns:

t transformed (give t not null, return will not be null)

MWETransform

public static Tree MWETransform(Tree t)

Puts all multi-word expressions below a single constituent labeled "MWE". Patterns for multi-word expressions are defined in MWE_PATTERNS.

prepCCTransform

public static Tree prepCCTransform(Tree t)

gappingTransform

public static Tree gappingTransform(Tree t)

main

public static void main(java.lang.String[] args)