it.uniroma2.sag.kelp.kernel.tree

Class PartialTreeKernel

java.lang.Object

it.uniroma2.sag.kelp.kernel.Kernel

it.uniroma2.sag.kelp.kernel.DirectKernel<TreeRepresentation>

it.uniroma2.sag.kelp.kernel.tree.PartialTreeKernel

public class PartialTreeKernel
extends DirectKernel<TreeRepresentation>

Partial Tree Kernel implementation. A Partial Tree Kernel is a convolution kernel that evaluates the tree fragments shared between two trees. The considered fragments are are partial trees, i.e. a node and its partial descendancy (the descendancy can be incomplete, i.e. a partial production is allowed). The kernel function is defined as:
\(K(T_1,T_2) = \sum_{n_1 \in N_{T_1}} \sum_{n_2 \in N_{T_2}} \Delta(n_1,n_2)\)
where \(\Delta(n_1,n_2)=\sum^{|F|}_i=1 I_i(n_1) I_i(n_2)\), that is the number of common fragments rooted at the n1 and n2. It can be computed as:
- if the node labels of \(n_1\) and \(n_2\) are different then \(\Delta(n_1,n_2)=0\)
- else \(\Delta(n_1,n_2)= \mu(\lambda^2 + \sum_{J_1, J_2, l(J_1)=l(J_2)} \lambda^{d(J_1)+d(J_2)} \prod_{i=1}^{l(J_1)} \Delta(c_{n_1}[J_{1i}], c_{n_2}[J_{2i}])\)

Fore details see [Moschitti, EACL2006] Alessandro Moschitti. Efficient convolution kernels for dependency and constituent syntactic trees. In ECML 2006, Berlin, Germany.

Author:

Danilo Croce, Giuseppe Castellucci

Field Summary

Fields inherited from class it.uniroma2.sag.kelp.kernel.DirectKernel

representation

Constructor Summary

Constructors

Constructor and Description
PartialTreeKernel() Default constructor.
PartialTreeKernel(float LAMBDA, float MU, float terminalFactor, String representationIdentifier) A Constructor for the Partial Tree Kernel in which parameters can be set manually.
PartialTreeKernel(String representationIdentifier) This constructor by default uses lambda=0.4, mu=0.4, terminalFactor=1

Method Summary

Methods

Modifier and Type	Method and Description
float	evaluateKernelNotNormalize(TreeRepresentation a, TreeRepresentation b) Evaluate the Partial Tree Kernel
DeltaMatrix	getDeltaMatrix()
float	getLambda() Get the Vertical Decay factor
int	getMaxSubseqLeng()
float	getMu() Get the Horizontal Decay factor
float	getTerminalFactor() Get the Terminal Factor
float	kernelComputation(TreeRepresentation repA, TreeRepresentation repB) Computes the kernel similarity between two specific representations
void	setDeltaMatrix(DeltaMatrix deltaMatrix) Deprecated.
void	setLambda(float lambda)
void	setMaxSubseqLeng(int maxSubseqLeng)
void	setMu(float mu)
void	setTerminalFactor(float terminalFactor)

Methods inherited from class it.uniroma2.sag.kelp.kernel.DirectKernel

getRepresentation, kernelComputation, setRepresentation

Methods inherited from class it.uniroma2.sag.kelp.kernel.Kernel

disableCache, getKernelCache, getKernelComputations, getNumberOfHits, getNumberOfMisses, getSquaredNormCache, innerProduct, load, reset, save, setKernelCache, setSquaredNormCache, squaredNorm, squaredNormOfTheDifference

Methods inherited from class java.lang.Object

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

Constructor Detail

PartialTreeKernel

public PartialTreeKernel()

Default constructor. It should be used only for json serialization/deserialization purposes This constructor by default uses lambda=0.4, mu=0.4, terminalFactor=1 and it forces to operate to the represenatation whose identifier is "0". Please use the PartialTreeKernel(String) or PartialTreeKernel(float,float,float,String) to use a Partial Tree Kernel in your application.

PartialTreeKernel

public PartialTreeKernel(float LAMBDA,
                 float MU,
                 float terminalFactor,
                 String representationIdentifier)

A Constructor for the Partial Tree Kernel in which parameters can be set manually.

Parameters:

LAMBDA - lambda value in the PTK formula

MU - mu value of the PTK formula

terminalFactor - terminal factor

representationIdentifier - the representation on which operate

PartialTreeKernel

public PartialTreeKernel(String representationIdentifier)

This constructor by default uses lambda=0.4, mu=0.4, terminalFactor=1

Method Detail

evaluateKernelNotNormalize

public float evaluateKernelNotNormalize(TreeRepresentation a,
                               TreeRepresentation b)

Evaluate the Partial Tree Kernel

Parameters:

a - First tree

b - Second Tree

Returns:

Kernel value

getDeltaMatrix

public DeltaMatrix getDeltaMatrix()

getLambda

public float getLambda()

Get the Vertical Decay factor

Returns:

Vertical Decay factor

getMu

public float getMu()

Get the Horizontal Decay factor

Returns:

Horizontal Decay factor

getTerminalFactor

public float getTerminalFactor()

Get the Terminal Factor

Returns:

kernelComputation

public float kernelComputation(TreeRepresentation repA,
                      TreeRepresentation repB)

Description copied from class: DirectKernel

Computes the kernel similarity between two specific representations

Specified by:

kernelComputation in class DirectKernel<TreeRepresentation>

Parameters:

repA - the first representation in the kernel similarity

repB - the second representation in the kernel similarity

Returns:

the kernel similarity

setDeltaMatrix

@Deprecated
public void setDeltaMatrix(DeltaMatrix deltaMatrix)

Deprecated.

Sets the delta matrix. This method should not be used, as the new KeLP versions are optimized to automatically set the proper delta matrix

Parameters:

deltaMatrix -

setLambda

public void setLambda(float lambda)

setMu

public void setMu(float mu)

setTerminalFactor

public void setTerminalFactor(float terminalFactor)

getMaxSubseqLeng

public int getMaxSubseqLeng()

Returns:

The maximum length of common subsequences considered in the recursion. It reflects the maximum branching factor allowed to the tree fragments.

setMaxSubseqLeng

public void setMaxSubseqLeng(int maxSubseqLeng)

Parameters:

maxSubseqLeng - The maximum length of common subsequences considered in the recursion. It reflects the maximum branching factor allowed to the tree fragments.