edu.ucla.sspace.wordsi

Class EvaluationWordsi

java.lang.Object

edu.ucla.sspace.wordsi.BaseWordsi

edu.ucla.sspace.wordsi.EvaluationWordsi

All Implemented Interfaces:

SemanticSpace, Wordsi

public class EvaluationWordsi
extends BaseWordsi

An Wordsi implementation to be used for evaluations. The word senses are not updated during processing, instead, each generated context vector is compared to the existing word senses and context vector is labeled with the id for the most similar sense. The sense labeling is passed on directly to the AssignmentReporter for each context vector generated.

Word sense must be provided by a SemanticSpace. For any polysemous words, the first sense must be keyed by the raw word and all other sense must be keyed by the raw word plus "-senseNumber" where senseNumber is an integer starting at 1, for the second sense, and goes up to N-1, for the last sense.

Author:

Keith Stevens

Constructor Summary

Constructors

Constructor and Description
EvaluationWordsi(Set<String> acceptedWords, ContextExtractor extractor, SemanticSpace sspace, AssignmentReporter reporter) Creates a new EvaluationWordsi.

Method Summary

Methods

Modifier and Type	Method and Description
Vector	getVector(String term) Returns the semantic vector for the provided word.
Set<String>	getWords() Returns the set of words that are represented in this semantic space.
void	handleContextVector(String focusKey, String secondaryKey, SparseDoubleVector context) Performs some operation with contextVector, which can be indexed by either primaryKey, secondaryKey, or both.
void	processSpace(Properties props) Once all the documents have been processed, performs any post-processing steps on the data.

Methods inherited from class edu.ucla.sspace.wordsi.BaseWordsi

acceptWord, getSpaceName, getVectorLength, processDocument

Methods inherited from class java.lang.Object

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

Constructor Detail

EvaluationWordsi

public EvaluationWordsi(Set<String> acceptedWords,
                ContextExtractor extractor,
                SemanticSpace sspace,
                AssignmentReporter reporter)

Creates a new EvaluationWordsi.

Parameters:

acceptedWords - The set of accepted words. Only these words will have context vectors generated.

extractor - The ContextExtractor responsible for generating context vectors.

sspace - The SemanticSpace responsible for provided existing word senses.

reporter - The AssignmentReporter reponsible for reporting sense labelings.

Method Detail

getWords

public Set<String> getWords()

Returns the set of words that are represented in this semantic space.

Returns:

the set of words that are represented in this semantic space.

getVector

public Vector getVector(String term)

Returns the semantic vector for the provided word.

Parameters:

term - a word that may be in the semantic space

Returns:

The Vector for the provided word or null if the word was not in the space.

handleContextVector

public void handleContextVector(String focusKey,
                       String secondaryKey,
                       SparseDoubleVector context)

Performs some operation with contextVector, which can be indexed by either primaryKey, secondaryKey, or both. This operation will likely assign the contextVector to some cluster immediately or store the contextVector so that it may be clustered with all other other context vecetors generated for primaryKey.

The secondaryKey does not need to be used, but some experiments may require it, such as the SenseEval/SemEval evaluation or pseudo-word disambiguation. For SenseEval/SemEval evaluations, a SenseEvalContextExtractor should be used, which will provide the context id as the secondaryKey; reporting should be done with a SenseEvalReporter. For pseudo-word disambiguation/discrimination, a PseudoWordContextExtractor should be used, which will create pseudo-words for some set of tokens. This extractor will use the pseudo-word for the primaryKey and the original token as the secondaryKey.

Parameters:

focusKey - The primary key for contextVector

context - a SparseDoubleVector that represents a single context for a word

processSpace

public void processSpace(Properties props)

Once all the documents have been processed, performs any post-processing steps on the data. An algorithm should treat this as a no-op if no post-processing is required. Callers may specify the values for any exposed parameters using the properties argument.

By general contract, once this method has been called, processDocument will not be called again.

Parameters:

props - a set of properties and values that may be used to configure any exposed parameters of the algorithm.