de.tudarmstadt.ukp.dkpro.core.api.frequency.util

Class ConditionalFrequencyDistribution<C,V>

java.lang.Object

de.tudarmstadt.ukp.dkpro.core.api.frequency.util.ConditionalFrequencyDistribution<C,V>

Type Parameters:

C - the type of the conditions

V - the type of the samples

public class ConditionalFrequencyDistribution<C,V>
extends Object

Inspired by nltk.probability.ConditionalFreqDist Maps a condition to a FrequencyDistribution.

This class could be used to learn how frequently a word collocates with another word. Suppose we want to know how frequently the word "of" appears after the word "because", then ConditionalFrequencyDistribution can be used as follows:

 ConditionalFrequencyDistribution<String, String> cfd =
     new ConditionalFrequencyDistribution<String, String>();

 cfd.addSample("because", "in");
 cfd.addSample("because", "of");
 cfd.addSample("despite", "in");
 cfd.addSample("because", "of");

 System.out.println(cfd.getCount("despite", "of"));
 

The last call to getCount(C, V) will yield 2, because given the condition that the first word in a two-word sequence is "because", the word "of" has appeared twice.

This class was inspired by NLTK's FreqDist.

See Also:

FrequencyDistribution

Constructor Summary

Constructors

Constructor and Description
ConditionalFrequencyDistribution() Creates a new empty ConditionalFrequencyDistribution.
ConditionalFrequencyDistribution(Map<C,Iterable<V>> samples) Creates a new ConditionalFrequencyDistribution and fills it with samples from a map.

Method Summary

Modifier and Type	Method and Description
void	addSample(C condition, V sample, long frequency) Adds a sample with a certain frequency under a given condition.
Set<C>	getConditions() Returns all conditions for which samples have been recorded.
long	getCount(C condition, V sample) Returns the total number of samples which equal a given sample under a given condition.
FrequencyDistribution<V>	getFrequencyDistribution(C condition) Returns the FrequencyDistribution under a given condition, or null if this distribution contains no such FrequencyDistribution for this condition.
long	getN()
boolean	hasCondition(C condition) Indicates whether samples have been recorded under a given condition.
void	inc(C condition, V sample) Increases a sample under a given condition.
void	incAll(C condition, Iterable<V> samples) Increases all provided samples under a given condition.
void	removeCondition(C condition) Remove a previously registered condition
void	setFrequencyDistribution(C condition, FrequencyDistribution<V> fd) Directly set the frequency distribution for a given condition.
String	toString()

Methods inherited from class java.lang.Object

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

Constructor Detail

ConditionalFrequencyDistribution

public ConditionalFrequencyDistribution()

Creates a new empty ConditionalFrequencyDistribution.

ConditionalFrequencyDistribution

public ConditionalFrequencyDistribution(Map<C,Iterable<V>> samples)

Creates a new ConditionalFrequencyDistribution and fills it with samples from a map.

Parameters:

samples - the Iterable to fill from

Method Detail

getN

public long getN()

Returns:

The total number of sample outcomes that have been recorded by this ConditionalFreqDist.

getCount

public long getCount(C condition,
                     V sample)

Returns the total number of samples which equal a given sample under a given condition.

If there are no samples for the condition in question on record, 0 will be returned.

Parameters:

condition - the condition

sample - the sample under a given condition

Returns:

the number of all samples which equal sample

getFrequencyDistribution

public FrequencyDistribution<V> getFrequencyDistribution(C condition)

Returns the FrequencyDistribution under a given condition, or null if this distribution contains no such FrequencyDistribution for this condition.

Parameters:

condition - the condition

Returns:

the distribution the condition maps to

setFrequencyDistribution

public void setFrequencyDistribution(C condition,
                                     FrequencyDistribution<V> fd)

Directly set the frequency distribution for a given condition.

Parameters:

condition - the condition.

fd - the distribution.

getConditions

public Set<C> getConditions()

Returns all conditions for which samples have been recorded.

Returns:

a Set of all recorded conditions

hasCondition

public boolean hasCondition(C condition)

Indicates whether samples have been recorded under a given condition.

Parameters:

condition - the condition in question

Returns:

true if samples for condition exist

inc

public void inc(C condition,
                V sample)

Increases a sample under a given condition.

Parameters:

condition - the condition for this sample

sample - the sample to add

incAll

public void incAll(C condition,
                   Iterable<V> samples)

Increases all provided samples under a given condition.

If there is no FrequencyDistribution present for the given condition, a new empty one will be created and populated from the given samples.

Parameters:

condition - the condition for the samples

samples - the samples to add

addSample

public void addSample(C condition,
                      V sample,
                      long frequency)

Adds a sample with a certain frequency under a given condition.

Parameters:

condition - the condition for this sample

sample - the sample to add

frequency - the frequenc of the sample

removeCondition

public void removeCondition(C condition)

Remove a previously registered condition

Parameters:

condition - the condition to be removed

toString

public String toString()

Overrides:

toString in class Object