jube-fork/Accord.MachineLearning/Representations/BagOfWords.cs

// Accord Machine Learning Library
// The Accord.NET Framework
// http://accord-framework.net
//
// Copyright © César Souza, 2009-2017
// cesarsouza at gmail.com
//
//    This library is free software; you can redistribute it and/or
//    modify it under the terms of the GNU Lesser General Public
//    License as published by the Free Software Foundation; either
//    version 2.1 of the License, or (at your option) any later version.
//
//    This library is distributed in the hope that it will be useful,
//    but WITHOUT ANY WARRANTY; without even the implied warranty of
//    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
//    Lesser General Public License for more details.
//
//    You should have received a copy of the GNU Lesser General Public
//    License along with this library; if not, write to the Free Software
//    Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
//

namespace Accord.MachineLearning
{
    using Accord.Math;
    using System;
    using System.Collections.Generic;
    using System.Runtime.Serialization;
    using Accord.MachineLearning.VectorMachines;
    using System.Threading.Tasks;
    using System.Collections.ObjectModel;
    using Accord.Compat;

    /// <summary>
    ///   Bag of words.
    /// </summary>
    /// 
    /// <remarks>
    ///   The bag-of-words (BoW) model can be used to extract finite
    ///   length features from otherwise varying length representations.
    /// </remarks>
    /// 
    /// <example>
    ///   <code source="Unit Tests\Accord.Tests.MachineLearning\BagOfWordsTest.cs" region="doc_learn"/>
    ///   
    /// <para>
    ///   The following example shows how to use Bag-of-Words to convert other kinds of sequences
    ///   into fixed-length representations. In particular, we apply Bag-of-Words to convert data
    ///   from the PENDIGITS handwritten digit recognition dataset and afterwards convert their
    ///   representations using a <see cref="SupportVectorMachine"/>.</para>
    ///   
    ///   <code source="Unit Tests\Accord.Tests.MachineLearning\BagOfWordsTest.cs" region="doc_learn_pendigits"/>
    /// </example>
    /// 
    [Serializable]
    public class BagOfWords : ParallelLearningBase, IBagOfWords<string[]>,
        ITransform<string[], Sparse<double>>,
        IUnsupervisedLearning<BagOfWords, string[], int[]>
    {
        // TODO: Inherit from BaseBagOfWords

        // TODO: Replace by TwoWayDictionary
        private Dictionary<string, int> stringToCode;
        private Dictionary<int, string> codeToString;


        [NonSerialized]
        private IDictionary<string, int> readOnlyStringToCode;

        [NonSerialized]
        private IDictionary<int, string> readOnlyCodeToString;



        /// <summary>
        ///   Gets the number of words in this codebook.
        /// </summary>
        /// 
        public int NumberOfWords
        {
            get { return stringToCode.Count; }
        }


        /// <summary>
        /// Gets the number of outputs generated by the model.
        /// </summary>
        /// <value>The number of outputs.</value>
        public int NumberOfOutputs
        {
            get { return this.NumberOfWords; }
            set { throw new InvalidOperationException("This property is read-only."); }
        }

        /// <summary>
        /// Gets the number of inputs accepted by the model.
        /// </summary>
        /// <value>The number of inputs.</value>
        public int NumberOfInputs
        {
            get { return -1; }
            set { throw new InvalidOperationException("This property is read-only."); }
        }

        /// <summary>
        ///   Gets the forward dictionary which translates
        ///   string tokens to integer labels.
        /// </summary>
        /// 
        public IDictionary<string, int> StringToCode
        {
            get { return readOnlyStringToCode; }
        }

        /// <summary>
        ///   Gets the reverse dictionary which translates
        ///   integer labels into string tokens.
        /// </summary>
        /// 
        public IDictionary<int, string> CodeToString
        {
            get { return readOnlyCodeToString; }
        }

        /// <summary>
        ///   Gets or sets the maximum number of occurrences of a word which
        ///   should be registered in the feature vector. Default is 1 (if a
        ///   word occurs, corresponding feature is set to 1).
        /// </summary>
        /// 
        public int MaximumOccurance { get; set; }

        /// <summary>
        ///   Constructs a new <see cref="BagOfWords"/>.
        /// </summary>
        /// 
        /// <param name="texts">The texts to build the bag of words model from.</param>
        /// 
        [Obsolete("Please use the default constructor and pass the texts to the Learn() method.")]
        public BagOfWords(params string[][] texts)
        {
            if (texts == null)
                throw new ArgumentNullException("texts");

            initialize(texts);
        }

        /// <summary>
        ///   Constructs a new <see cref="BagOfWords"/>.
        /// </summary>
        /// 
        /// <param name="texts">The texts to build the bag of words model from.</param>
        /// 
        [Obsolete("Please use the default constructor and pass the texts to the Learn() method.")]
        public BagOfWords(params string[] texts)
        {
            if (texts == null)
                throw new ArgumentNullException("texts");

            initialize(new[] { texts });
        }

        /// <summary>
        ///   Constructs a new <see cref="BagOfWords"/>.
        /// </summary>
        /// 
        public BagOfWords()
        {
            initialize(null);
        }

        private void initialize(string[][] texts)
        {
            stringToCode = new Dictionary<string, int>();
            codeToString = new Dictionary<int, string>();

            readOnlyStringToCode = new ReadOnlyDictionary<string, int>(stringToCode);
            readOnlyCodeToString = new ReadOnlyDictionary<int, string>(codeToString);

            MaximumOccurance = 1;

            if (texts != null)
                Learn(texts);
        }


        /// <summary>
        ///   Computes the Bag of Words model.
        /// </summary>
        /// 
        [Obsolete("Please use the Learn() method instead.")]
        public void Compute(string[][] texts)
        {
            Learn(texts);
        }

        private static void checkArgs(string[][] texts, double[] weights)
        {
            if (weights != null)
                throw new ArgumentException("Not Supported Weights", "weights");

            if (texts == null)
                throw new ArgumentNullException("texts");

            for (int i = 0; i < texts.Length; i++)
            {
                if (texts[i] == null)
                    throw new ArgumentNullException("texts",
                        "Text " + i + "cannot be null.");

                for (int j = 0; j < texts[i].Length; j++)
                    if (texts[i][j] == null)
                        throw new ArgumentNullException("texts",
                            "Token at text " + i + ", position " + i + " cannot be null.");
            }
        }

        /// <summary>
        ///   Gets the codeword representation of a given text.
        /// </summary>
        /// 
        /// <param name="text">The text to be processed.</param>
        /// 
        /// <returns>An integer vector with the same length as words
        /// in the code book.</returns>
        /// 
        [Obsolete("Please use Transform() instead.")]
        public int[] GetFeatureVector(params string[] text)
        {
            return ((ITransform<string[], int[]>)this).Transform(text);
        }


        double[] IBagOfWords<string[]>.GetFeatureVector(string[] value)
        {
            return ((ITransform<string[], double[]>)this).Transform(value);
        }


        [OnDeserialized]
        private void OnDeserialized(StreamingContext context)
        {
            readOnlyStringToCode = new ReadOnlyDictionary<string, int>(stringToCode);
            readOnlyCodeToString = new ReadOnlyDictionary<int, string>(codeToString);
        }

        /// <summary>
        /// Applies the transformation to an input, producing an associated output.
        /// </summary>
        /// <param name="input">The input data to which
        /// the transformation should be applied.</param>
        /// <param name="result">The location to where to store the
        /// result of this transformation.</param>
        /// <returns>The output generated by applying this
        /// transformation to the given input.</returns>
        public double[] Transform(string[] input, double[] result)
        {
            // Detect all feature words
            foreach (string word in input)
            {
                int j;
                if (!stringToCode.TryGetValue(word, out j))
                    continue;

                if (result[j] < MaximumOccurance)
                    result[j]++;
            }

            return result;
        }

        /// <summary>
        /// Applies the transformation to an input, producing an associated output.
        /// </summary>
        /// <param name="input">The input data to which
        /// the transformation should be applied.</param>
        /// <param name="result">The location to where to store the
        /// result of this transformation.</param>
        /// <returns>The output generated by applying this
        /// transformation to the given input.</returns>
        public int[] Transform(string[] input, int[] result)
        {
            // Detect all feature words
            foreach (string word in input)
            {
                int j;
                if (!stringToCode.TryGetValue(word, out j))
                    continue;

                if (result[j] < MaximumOccurance)
                    result[j]++;
            }

            return result;
        }

        int[] ICovariantTransform<string[], int[]>.Transform(string[] input)
        {
            return Transform(input, new int[NumberOfWords]);
        }

        /// <summary>
        /// Applies the transformation to an input, producing an associated output.
        /// </summary>
        /// <param name="input">The input data to which the transformation should be applied.</param>
        /// <returns>The output generated by applying this transformation to the given input.</returns>
        public double[] Transform(string[] input)
        {
            return Transform(input, new double[NumberOfWords]);
        }

        /// <summary>
        /// Applies the transformation to a set of input vectors,
        /// producing an associated set of output vectors.
        /// </summary>
        /// <param name="input">The input data to which
        /// the transformation should be applied.</param>
        /// <param name="result">The location to where to store the
        /// result of this transformation.</param>
        /// <returns>The output generated by applying this
        /// transformation to the given input.</returns>
        public int[][] Transform(string[][] input, int[][] result)
        {
            Parallel.For(0, input.Length, ParallelOptions, i =>
                Transform(input[i], result[i]));
            return result;
        }

        /// <summary>
        /// Applies the transformation to a set of input vectors,
        /// producing an associated set of output vectors.
        /// </summary>
        /// <param name="input">The input data to which
        /// the transformation should be applied.</param>
        /// <param name="result">The location to where to store the
        /// result of this transformation.</param>
        /// <returns>The output generated by applying this
        /// transformation to the given input.</returns>
        public double[][] Transform(string[][] input, double[][] result)
        {
            Parallel.For(0, input.Length, ParallelOptions, i =>
                Transform(input[i], result[i]));
            return result;
        }

        /// <summary>
        /// Applies the transformation to a set of input vectors,
        /// producing an associated set of output vectors.
        /// </summary>
        /// <param name="input">The input data to which
        /// the transformation should be applied.</param>
        /// <returns>The output generated by applying this
        /// transformation to the given input.</returns>
        public double[][] Transform(string[][] input)
        {
            return Transform(input, Jagged.Zeros<double>(input.Length, NumberOfWords));
        }

        int[][] ICovariantTransform<string[], int[]>.Transform(string[][] input)
        {
            return Transform(input, Jagged.Zeros<int>(input.Length, NumberOfWords));
        }

        /// <summary>
        /// Learns a model that can map the given inputs to the desired outputs.
        /// </summary>
        /// <param name="x">The model inputs.</param>
        /// <param name="weights">The weight of importance for each input sample.</param>
        /// <returns>A model that has learned how to produce suitable outputs
        /// given the input data <paramref name="x" />.</returns>
        public BagOfWords Learn(string[] x, double[] weights = null)
        {
            return Learn(x.Tokenize(), weights);
        }

        /// <summary>
        /// Learns a model that can map the given inputs to the desired outputs.
        /// </summary>
        /// <param name="x">The model inputs.</param>
        /// <param name="weights">The weight of importance for each input sample.</param>
        /// <returns>A model that has learned how to produce suitable outputs
        /// given the input data <paramref name="x" />.</returns>
        public BagOfWords Learn(string[][] x, double[] weights = null)
        {
            checkArgs(x, weights);

            // TODO: Use Parallel.For
            int symbol = 0;
            foreach (string[] text in x)
            {
                foreach (string word in text)
                {
                    if (!stringToCode.ContainsKey(word))
                    {
                        stringToCode[word] = symbol;
                        codeToString[symbol] = word;
                        symbol++;
                    }
                }
            }

            return this;
        }





        /// <summary>
        /// Applies the transformation to an input, producing an associated output.
        /// </summary>
        /// <param name="input">The input data to which
        /// the transformation should be applied.</param>
        /// <param name="result">The location to where to store the
        /// result of this transformation.</param>
        /// <returns>The output generated by applying this
        /// transformation to the given input.</returns>
        public Sparse<double> Transform(string[] input, Sparse<double> result)
        {
            // Detect all feature words
            foreach (string word in input)
            {
                int j;
                if (!stringToCode.TryGetValue(word, out j))
                    continue;

                if (result[j] < MaximumOccurance)
                    result[j]++;
            }

            return result;
        }

        Sparse<double> ICovariantTransform<string[], Sparse<double>>.Transform(string[] input)
        {
            return Sparse.FromDense(Transform(input));
        }

        Sparse<double>[] ICovariantTransform<string[], Sparse<double>>.Transform(string[][] input)
        {
            return Transform(input, new Sparse<double>[input.Length]);
        }

        /// <summary>
        /// Applies the transformation to a set of input vectors,
        /// producing an associated set of output vectors.
        /// </summary>
        /// <param name="input">The input data to which
        /// the transformation should be applied.</param>
        /// <param name="result">The location to where to store the
        /// result of this transformation.</param>
        /// <returns>The output generated by applying this
        /// transformation to the given input.</returns>
        public Sparse<double>[] Transform(string[][] input, Sparse<double>[] result)
        {
            var t = ((ITransform<string[], Sparse<double>>)this);
            Parallel.For(0, input.Length, ParallelOptions, i =>
            {
                result[i] = t.Transform(input[i]);
            });
            return result;
        }



        /// <summary>
        ///   Creates the default clustering algorithm for Bag-of-Words models (<see cref="KMeans"/>).
        /// </summary>
        /// 
        /// <param name="numberOfWords">The number of clusters for k-means.</param>
        /// 
        public static KMeans GetDefaultClusteringAlgorithm(int numberOfWords)
        {
            return new KMeans(numberOfWords)
            {
                ComputeCovariances = false,
                UseSeeding = Seeding.KMeansPlusPlus,
            };
        }
    }
}