edu.ucla.sspace.wordsi.EvaluationWordsi Maven / Gradle / Ivy

Go to download

Show more of this group Show more artifacts with this name
Show all versions of sspace-wordsi Show documentation

The S-Space Package is a collection of algorithms for building Semantic Spaces as well as a highly-scalable library for designing new distributional semantics algorithms. Distributional algorithms process text corpora and represent the semantic for words as high dimensional feature vectors. This package also includes matrices, vectors, and numerous clustering algorithms. These approaches are known by many names, such as word spaces, semantic spaces, or distributed semantics and rest upon the Distributional Hypothesis: words that appear in similar contexts have similar meanings.

The newest version!

/*
 * Copyright 2010 Keith Stevens 
 *
 * This file is part of the S-Space package and is covered under the terms and
 * conditions therein.
 *
 * The S-Space package is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as published
 * by the Free Software Foundation and distributed hereunder to you.
 *
 * THIS SOFTWARE IS PROVIDED "AS IS" AND NO REPRESENTATIONS OR WARRANTIES,
 * EXPRESS OR IMPLIED ARE MADE.  BY WAY OF EXAMPLE, BUT NOT LIMITATION, WE MAKE
 * NO REPRESENTATIONS OR WARRANTIES OF MERCHANT- ABILITY OR FITNESS FOR ANY
 * PARTICULAR PURPOSE OR THAT THE USE OF THE LICENSED SOFTWARE OR DOCUMENTATION
 * WILL NOT INFRINGE ANY THIRD PARTY PATENTS, COPYRIGHTS, TRADEMARKS OR OTHER
 * RIGHTS.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program. If not, see .
 */

package edu.ucla.sspace.wordsi;

import edu.ucla.sspace.common.SemanticSpace;
import edu.ucla.sspace.common.Similarity;

import edu.ucla.sspace.vector.SparseDoubleVector;
import edu.ucla.sspace.vector.Vector;

import java.util.HashSet;
import java.util.Properties;
import java.util.Set;


/**
 * An {@link 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 {@link AssignmentReporter} for each context vector generated.
 *
 * 
 *
 * Word sense must be provided by a {@link 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
 */
public class EvaluationWordsi extends BaseWordsi {

    /**
     * The underlying {@link SemanticSpace} that provides word senses.
     */
    private final SemanticSpace wordSpace;

    /**
     * The {@link AssignmentReporter} used to report sense labels for generated
     * contexts.
     */
    private final AssignmentReporter reporter;

    /**
     * Creates a new {@link EvaluationWordsi}.
     *
     * @param acceptedWords The set of accepted words.  Only these words will
     *        have context vectors generated.
     * @param extractor The {@link ContextExtractor} responsible for generating
     *        context vectors.
     * @param sspace The {@link SemanticSpace} responsible for provided existing
     *        word senses.
     * @param reporter The {@link AssignmentReporter} reponsible for reporting
     *        sense labelings.
     */
    public EvaluationWordsi(Set acceptedWords,
                            ContextExtractor extractor,
                            SemanticSpace sspace,
                            AssignmentReporter reporter) {
        super(acceptedWords, extractor);
        this.wordSpace = sspace;
        this.reporter = reporter;
    }

    /**
     * {@inheritDoc}
     */
    public Set getWords() {
        return new HashSet();
    }

    /**
     * {@inheritDoc}
     */
    public Vector getVector(String term) {
        return wordSpace.getVector(term);
    }

    /**
     * {@inheritDoc}
     */
    public void handleContextVector(String focusKey,
                                    String secondaryKey,
                                    SparseDoubleVector context) {
        // Find the most similar existing word sense.
        int senseNumber = 0;
        int bestSense = 0;
        double bestSimilarity = -1;

        while (true) {
            // Create the word sense key based on the sense number.
            Vector wordSense = getVector((senseNumber == 0)
                    ? focusKey
                    : focusKey + "-" + senseNumber);

            // If no word sense exists then we have examined all known word senses.
            if (wordSense == null)
                break;

            // Compute the similarity of this context vector to the current word
            // sense.
            double similarity = Similarity.cosineSimilarity(wordSense, context);
            if (similarity > bestSimilarity) {
                bestSimilarity = similarity;
                bestSense = senseNumber;
            }

            senseNumber++;
        }

        // If a reporter is provided, report the sense labeling.
        if (reporter != null)
            reporter.updateAssignment(focusKey, secondaryKey, bestSense);
    }


    /**
     * {@inheritDoc}
     */
    public void processSpace(Properties props) {
        if (reporter != null)
            reporter.finalizeReport();
    }
}