[SPARK-12685][MLLIB][BACKPORT TO 1.4] word2vec trainWordsCount gets overflow

jira: https://issues.apache.org/jira/browse/SPARK-12685

master PR: #10627

the log of word2vec reports
trainWordsCount = -785727483
during computation over a large dataset.

Update the priority as it will affect the computation process.
alpha = learningRate * (1 - numPartitions * wordCount.toDouble / (trainWordsCount + 1))

Author: Yuhao Yang <hhbyyh@gmail.com>

Closes #10721 from hhbyyh/branch-1.4.

Author: hhbyyh

mllib/src/main/scala/org/apache/spark/mllib/feature/Word2Vec.scala

@@ -139,21 +139,21 @@ class Word2Vec extends Serializable with Logging {
   /** context words from [-window, window] */
   private val window = 5
 
-  private var trainWordsCount = 0
+  private var trainWordsCount = 0L
   private var vocabSize = 0
   @transient private var vocab: Array[VocabWord] = null
   @transient private var vocabHash = mutable.HashMap.empty[String, Int]
 
   private def learnVocab(words: RDD[String]): Unit = {
     vocab = words.map(w => (w, 1))
       .reduceByKey(_ + _)
+      .filter(_._2 >= minCount)
       .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)
 
@@ -164,7 +164,7 @@ class Word2Vec extends Serializable with Logging {
       trainWordsCount += vocab(a).cn
       a += 1
     }
-    logInfo("trainWordsCount = " + trainWordsCount)
+    logInfo(s"vocabSize = $vocabSize, trainWordsCount = $trainWordsCount")
   }
 
   private def createExpTable(): Array[Float] = {
@@ -313,7 +313,7 @@ class Word2Vec extends Serializable with Logging {
         val random = new XORShiftRandom(seed ^ ((idx + 1) << 16) ^ ((-k - 1) << 8))
         val syn0Modify = new Array[Int](vocabSize)
         val syn1Modify = new Array[Int](vocabSize)
-        val model = iter.foldLeft((syn0Global, syn1Global, 0, 0)) {
+        val model = iter.foldLeft((syn0Global, syn1Global, 0L, 0L)) {
           case ((syn0, syn1, lastWordCount, wordCount), sentence) =>
             var lwc = lastWordCount
             var wc = wordCount