All Downloads are FREE. Search and download functionalities are using the official Maven repository.

edu.ucla.sspace.hal.GeometricWeighting Maven / Gradle / Ivy

Go to download

The S-Space Package is a Natural Language Processing library for distributional semantics representations. Distributional semantics representations model the meaning of words, phrases, and sentences as high dimensional vectors or probability distributions. The library includes common algorithms such as Latent Semantic Analysis, Random Indexing, and Latent Dirichlet Allocation. The S-Space package also includes software libraries for matrices, vectors, graphs, and numerous clustering algorithms.

The newest version!
/*
 * Copyright 2009 David Jurgens
 *
 * 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.hal;

/**
 * A geometically-decreasing weighting scheme for specifying how a {@link
 * HyperspaceAnalogueToLanguage} instance should weigh co-occurrences based on
 * the word distance.
 */
public class GeometricWeighting implements WeightingFunction {

    /**
     * Returns the weighed value where the closest words receive a weight equal
     * to the window size and the most distance words receive a weight of {@code
     * 1}, using a geometric (1 / 2n) decrease for in-between values.
     *
     * @param positionOffset {@inheritDoc}
     * @param windowSize {@inheritDoc}
     *
     * @return {@inheritDoc}
     */
    public double weight(int positionOffset, int windowSize) {
	return ((1 << (windowSize - (Math.abs(positionOffset) - 1))) / 
		(double)(1 << windowSize)) * windowSize;
    }

}




© 2015 - 2024 Weber Informatics LLC | Privacy Policy