diff options
author | Liquan Pei <lpei@gopivotal.com> | 2014-08-03 23:55:58 -0700 |
---|---|---|
committer | Xiangrui Meng <meng@databricks.com> | 2014-08-03 23:55:58 -0700 |
commit | e053c55819363fab7068bb9165e3379f0c2f570c (patch) | |
tree | 4c59ae31a557ded8507e036d10d9e8f2300c83d9 /mllib/src/main | |
parent | ae58aea2d1435b5bb011e68127e1bcddc2edf5b2 (diff) | |
download | spark-e053c55819363fab7068bb9165e3379f0c2f570c.tar.gz spark-e053c55819363fab7068bb9165e3379f0c2f570c.tar.bz2 spark-e053c55819363fab7068bb9165e3379f0c2f570c.zip |
[MLlib] [SPARK-2510]Word2Vec: Distributed Representation of Words
This is a pull request regarding SPARK-2510 at https://issues.apache.org/jira/browse/SPARK-2510. Word2Vec creates vector representation of words in a text corpus. The algorithm first constructs a vocabulary from the corpus and then learns vector representation of words in the vocabulary. The vector representation can be used as features in natural language processing and machine learning algorithms.
To make our implementation more scalable, we train each partition separately and merge the model of each partition after each iteration. To make the model more accurate, multiple iterations may be needed.
To investigate the vector representations is to find the closest words for a query word. For example, the top 20 closest words to "china" are for 1 partition and 1 iteration :
taiwan 0.8077646146334014
korea 0.740913304563621
japan 0.7240667798885471
republic 0.7107151279078352
thailand 0.6953217332072862
tibet 0.6916782118129544
mongolia 0.6800858715972612
macau 0.6794925677480378
singapore 0.6594048695593799
manchuria 0.658989931844148
laos 0.6512978726001666
nepal 0.6380792327845325
mainland 0.6365469459587788
myanmar 0.6358614338840394
macedonia 0.6322366180313249
xinjiang 0.6285291551708028
russia 0.6279951236068411
india 0.6272874944023487
shanghai 0.6234544135576999
macao 0.6220588462925876
The result with 10 partitions and 5 iterations is:
taiwan 0.8310495079388313
india 0.7737171315919039
japan 0.756777901233668
korea 0.7429767187102452
indonesia 0.7407557427278356
pakistan 0.712883426985585
mainland 0.7053379963140822
thailand 0.696298191073948
mongolia 0.693690656871415
laos 0.6913069680735292
macau 0.6903427690029617
republic 0.6766381604813666
malaysia 0.676460699141784
singapore 0.6728790997360923
malaya 0.672345232966194
manchuria 0.6703732292753156
macedonia 0.6637955686322028
myanmar 0.6589462882439646
kazakhstan 0.657017801081494
cambodia 0.6542383836451932
Author: Liquan Pei <lpei@gopivotal.com>
Author: Xiangrui Meng <meng@databricks.com>
Author: Liquan Pei <liquanpei@gmail.com>
Closes #1719 from Ishiihara/master and squashes the following commits:
2ba9483 [Liquan Pei] minor fix for Word2Vec test
e248441 [Liquan Pei] minor style change
26a948d [Liquan Pei] Merge pull request #1 from mengxr/Ishiihara-master
c14da41 [Xiangrui Meng] fix styles
384c771 [Xiangrui Meng] remove minCount and window from constructor change model to use float instead of double
e93e726 [Liquan Pei] use treeAggregate instead of aggregate
1a8fb41 [Liquan Pei] use weighted sum in combOp
7efbb6f [Liquan Pei] use broadcast version of vocab in aggregate
6bcc8be [Liquan Pei] add multiple iteration support
720b5a3 [Liquan Pei] Add test for Word2Vec algorithm, minor fixes
2e92b59 [Liquan Pei] modify according to feedback
57dc50d [Liquan Pei] code formatting
e4a04d3 [Liquan Pei] minor fix
0aafb1b [Liquan Pei] Add comments, minor fixes
8d6befe [Liquan Pei] initial commit
Diffstat (limited to 'mllib/src/main')
-rw-r--r-- | mllib/src/main/scala/org/apache/spark/mllib/feature/Word2Vec.scala | 424 |
1 files changed, 424 insertions, 0 deletions
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/feature/Word2Vec.scala b/mllib/src/main/scala/org/apache/spark/mllib/feature/Word2Vec.scala new file mode 100644 index 0000000000..87c81e7b0b --- /dev/null +++ b/mllib/src/main/scala/org/apache/spark/mllib/feature/Word2Vec.scala @@ -0,0 +1,424 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.mllib.feature + +import scala.collection.mutable +import scala.collection.mutable.ArrayBuffer +import scala.util.Random + +import com.github.fommil.netlib.BLAS.{getInstance => blas} +import org.apache.spark.{HashPartitioner, Logging} +import org.apache.spark.SparkContext._ +import org.apache.spark.annotation.Experimental +import org.apache.spark.mllib.linalg.{Vector, Vectors} +import org.apache.spark.mllib.rdd.RDDFunctions._ +import org.apache.spark.rdd._ +import org.apache.spark.storage.StorageLevel + +/** + * Entry in vocabulary + */ +private case class VocabWord( + var word: String, + var cn: Int, + var point: Array[Int], + var code: Array[Int], + var codeLen:Int +) + +/** + * :: Experimental :: + * Word2Vec creates vector representation of words in a text corpus. + * The algorithm first constructs a vocabulary from the corpus + * and then learns vector representation of words in the vocabulary. + * The vector representation can be used as features in + * natural language processing and machine learning algorithms. + * + * We used skip-gram model in our implementation and hierarchical softmax + * method to train the model. The variable names in the implementation + * matches the original C implementation. + * + * For original C implementation, see https://code.google.com/p/word2vec/ + * For research papers, see + * Efficient Estimation of Word Representations in Vector Space + * and + * Distributed Representations of Words and Phrases and their Compositionality. + * @param size vector dimension + * @param startingAlpha initial learning rate + * @param parallelism number of partitions to run Word2Vec (using a small number for accuracy) + * @param numIterations number of iterations to run, should be smaller than or equal to parallelism + */ +@Experimental +class Word2Vec( + val size: Int, + val startingAlpha: Double, + val parallelism: Int, + val numIterations: Int) extends Serializable with Logging { + + /** + * Word2Vec with a single thread. + */ + def this(size: Int, startingAlpha: Int) = this(size, startingAlpha, 1, 1) + + private val EXP_TABLE_SIZE = 1000 + private val MAX_EXP = 6 + private val MAX_CODE_LENGTH = 40 + private val MAX_SENTENCE_LENGTH = 1000 + private val layer1Size = size + private val modelPartitionNum = 100 + + /** context words from [-window, window] */ + private val window = 5 + + /** minimum frequency to consider a vocabulary word */ + private val minCount = 5 + + private var trainWordsCount = 0 + private var vocabSize = 0 + private var vocab: Array[VocabWord] = null + private var vocabHash = mutable.HashMap.empty[String, Int] + private var alpha = startingAlpha + + private def learnVocab(words:RDD[String]): Unit = { + vocab = words.map(w => (w, 1)) + .reduceByKey(_ + _) + .map(x => VocabWord( + x._1, + x._2, + new Array[Int](MAX_CODE_LENGTH), + new Array[Int](MAX_CODE_LENGTH), + 0)) + .filter(_.cn >= minCount) + .collect() + .sortWith((a, b) => a.cn > b.cn) + + vocabSize = vocab.length + var a = 0 + while (a < vocabSize) { + vocabHash += vocab(a).word -> a + trainWordsCount += vocab(a).cn + a += 1 + } + logInfo("trainWordsCount = " + trainWordsCount) + } + + private def createExpTable(): Array[Float] = { + val expTable = new Array[Float](EXP_TABLE_SIZE) + var i = 0 + while (i < EXP_TABLE_SIZE) { + val tmp = math.exp((2.0 * i / EXP_TABLE_SIZE - 1.0) * MAX_EXP) + expTable(i) = (tmp / (tmp + 1.0)).toFloat + i += 1 + } + expTable + } + + private def createBinaryTree(): Unit = { + val count = new Array[Long](vocabSize * 2 + 1) + val binary = new Array[Int](vocabSize * 2 + 1) + val parentNode = new Array[Int](vocabSize * 2 + 1) + val code = new Array[Int](MAX_CODE_LENGTH) + val point = new Array[Int](MAX_CODE_LENGTH) + var a = 0 + while (a < vocabSize) { + count(a) = vocab(a).cn + a += 1 + } + while (a < 2 * vocabSize) { + count(a) = 1e9.toInt + a += 1 + } + var pos1 = vocabSize - 1 + var pos2 = vocabSize + + var min1i = 0 + var min2i = 0 + + a = 0 + while (a < vocabSize - 1) { + if (pos1 >= 0) { + if (count(pos1) < count(pos2)) { + min1i = pos1 + pos1 -= 1 + } else { + min1i = pos2 + pos2 += 1 + } + } else { + min1i = pos2 + pos2 += 1 + } + if (pos1 >= 0) { + if (count(pos1) < count(pos2)) { + min2i = pos1 + pos1 -= 1 + } else { + min2i = pos2 + pos2 += 1 + } + } else { + min2i = pos2 + pos2 += 1 + } + count(vocabSize + a) = count(min1i) + count(min2i) + parentNode(min1i) = vocabSize + a + parentNode(min2i) = vocabSize + a + binary(min2i) = 1 + a += 1 + } + // Now assign binary code to each vocabulary word + var i = 0 + a = 0 + while (a < vocabSize) { + var b = a + i = 0 + while (b != vocabSize * 2 - 2) { + code(i) = binary(b) + point(i) = b + i += 1 + b = parentNode(b) + } + vocab(a).codeLen = i + vocab(a).point(0) = vocabSize - 2 + b = 0 + while (b < i) { + vocab(a).code(i - b - 1) = code(b) + vocab(a).point(i - b) = point(b) - vocabSize + b += 1 + } + a += 1 + } + } + + /** + * Computes the vector representation of each word in vocabulary. + * @param dataset an RDD of words + * @return a Word2VecModel + */ + def fit[S <: Iterable[String]](dataset: RDD[S]): Word2VecModel = { + + val words = dataset.flatMap(x => x) + + learnVocab(words) + + createBinaryTree() + + val sc = dataset.context + + val expTable = sc.broadcast(createExpTable()) + val bcVocab = sc.broadcast(vocab) + val bcVocabHash = sc.broadcast(vocabHash) + + val sentences: RDD[Array[Int]] = words.mapPartitions { iter => + new Iterator[Array[Int]] { + def hasNext: Boolean = iter.hasNext + + def next(): Array[Int] = { + var sentence = new ArrayBuffer[Int] + var sentenceLength = 0 + while (iter.hasNext && sentenceLength < MAX_SENTENCE_LENGTH) { + val word = bcVocabHash.value.get(iter.next()) + word match { + case Some(w) => + sentence += w + sentenceLength += 1 + case None => + } + } + sentence.toArray + } + } + } + + val newSentences = sentences.repartition(parallelism).cache() + var syn0Global = + Array.fill[Float](vocabSize * layer1Size)((Random.nextFloat() - 0.5f) / layer1Size) + var syn1Global = new Array[Float](vocabSize * layer1Size) + + for(iter <- 1 to numIterations) { + val (aggSyn0, aggSyn1, _, _) = + // TODO: broadcast temp instead of serializing it directly + // or initialize the model in each executor + newSentences.treeAggregate((syn0Global, syn1Global, 0, 0))( + seqOp = (c, v) => (c, v) match { + case ((syn0, syn1, lastWordCount, wordCount), sentence) => + var lwc = lastWordCount + var wc = wordCount + if (wordCount - lastWordCount > 10000) { + lwc = wordCount + alpha = startingAlpha * (1 - parallelism * wordCount.toDouble / (trainWordsCount + 1)) + if (alpha < startingAlpha * 0.0001) alpha = startingAlpha * 0.0001 + logInfo("wordCount = " + wordCount + ", alpha = " + alpha) + } + wc += sentence.size + var pos = 0 + while (pos < sentence.size) { + val word = sentence(pos) + // TODO: fix random seed + val b = Random.nextInt(window) + // Train Skip-gram + var a = b + while (a < window * 2 + 1 - b) { + if (a != window) { + val c = pos - window + a + if (c >= 0 && c < sentence.size) { + val lastWord = sentence(c) + val l1 = lastWord * layer1Size + val neu1e = new Array[Float](layer1Size) + // Hierarchical softmax + var d = 0 + while (d < bcVocab.value(word).codeLen) { + val l2 = bcVocab.value(word).point(d) * layer1Size + // Propagate hidden -> output + var f = blas.sdot(layer1Size, syn0, l1, 1, syn1, l2, 1) + if (f > -MAX_EXP && f < MAX_EXP) { + val ind = ((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2.0)).toInt + f = expTable.value(ind) + val g = ((1 - bcVocab.value(word).code(d) - f) * alpha).toFloat + blas.saxpy(layer1Size, g, syn1, l2, 1, neu1e, 0, 1) + blas.saxpy(layer1Size, g, syn0, l1, 1, syn1, l2, 1) + } + d += 1 + } + blas.saxpy(layer1Size, 1.0f, neu1e, 0, 1, syn0, l1, 1) + } + } + a += 1 + } + pos += 1 + } + (syn0, syn1, lwc, wc) + }, + combOp = (c1, c2) => (c1, c2) match { + case ((syn0_1, syn1_1, lwc_1, wc_1), (syn0_2, syn1_2, lwc_2, wc_2)) => + val n = syn0_1.length + val weight1 = 1.0f * wc_1 / (wc_1 + wc_2) + val weight2 = 1.0f * wc_2 / (wc_1 + wc_2) + blas.sscal(n, weight1, syn0_1, 1) + blas.sscal(n, weight1, syn1_1, 1) + blas.saxpy(n, weight2, syn0_2, 1, syn0_1, 1) + blas.saxpy(n, weight2, syn1_2, 1, syn1_1, 1) + (syn0_1, syn1_1, lwc_1 + lwc_2, wc_1 + wc_2) + }) + syn0Global = aggSyn0 + syn1Global = aggSyn1 + } + newSentences.unpersist() + + val wordMap = new Array[(String, Array[Float])](vocabSize) + var i = 0 + while (i < vocabSize) { + val word = bcVocab.value(i).word + val vector = new Array[Float](layer1Size) + Array.copy(syn0Global, i * layer1Size, vector, 0, layer1Size) + wordMap(i) = (word, vector) + i += 1 + } + val modelRDD = sc.parallelize(wordMap, modelPartitionNum) + .partitionBy(new HashPartitioner(modelPartitionNum)) + .persist(StorageLevel.MEMORY_AND_DISK) + + new Word2VecModel(modelRDD) + } +} + +/** +* Word2Vec model +*/ +class Word2VecModel(private val model: RDD[(String, Array[Float])]) extends Serializable { + + private def cosineSimilarity(v1: Array[Float], v2: Array[Float]): Double = { + require(v1.length == v2.length, "Vectors should have the same length") + val n = v1.length + val norm1 = blas.snrm2(n, v1, 1) + val norm2 = blas.snrm2(n, v2, 1) + if (norm1 == 0 || norm2 == 0) return 0.0 + blas.sdot(n, v1, 1, v2,1) / norm1 / norm2 + } + + /** + * Transforms a word to its vector representation + * @param word a word + * @return vector representation of word + */ + def transform(word: String): Vector = { + val result = model.lookup(word) + if (result.isEmpty) { + throw new IllegalStateException(s"$word not in vocabulary") + } + else Vectors.dense(result(0).map(_.toDouble)) + } + + /** + * Transforms an RDD to its vector representation + * @param dataset a an RDD of words + * @return RDD of vector representation + */ + def transform(dataset: RDD[String]): RDD[Vector] = { + dataset.map(word => transform(word)) + } + + /** + * Find synonyms of a word + * @param word a word + * @param num number of synonyms to find + * @return array of (word, similarity) + */ + def findSynonyms(word: String, num: Int): Array[(String, Double)] = { + val vector = transform(word) + findSynonyms(vector,num) + } + + /** + * Find synonyms of the vector representation of a word + * @param vector vector representation of a word + * @param num number of synonyms to find + * @return array of (word, cosineSimilarity) + */ + def findSynonyms(vector: Vector, num: Int): Array[(String, Double)] = { + require(num > 0, "Number of similar words should > 0") + val topK = model.map { case(w, vec) => + (cosineSimilarity(vector.toArray.map(_.toFloat), vec), w) } + .sortByKey(ascending = false) + .take(num + 1) + .map(_.swap) + .tail + + topK + } +} + +object Word2Vec{ + /** + * Train Word2Vec model + * @param input RDD of words + * @param size vector dimension + * @param startingAlpha initial learning rate + * @param parallelism number of partitions to run Word2Vec (using a small number for accuracy) + * @param numIterations number of iterations, should be smaller than or equal to parallelism + * @return Word2Vec model + */ + def train[S <: Iterable[String]]( + input: RDD[S], + size: Int, + startingAlpha: Double, + parallelism: Int = 1, + numIterations:Int = 1): Word2VecModel = { + new Word2Vec(size,startingAlpha, parallelism, numIterations).fit[S](input) + } +} |