gistfile1.scala

import io.Source

object SpellChecker {
  val NWORDS = train(words(Source.fromFile("/tmp/big.txt", "ISO-8859-1").mkString))
  val alphabet = ('a' to 'z').toList

  def words(text: String):Iterator[String] = """[a-z]+""".r.findAllIn(text.toLowerCase)

  def train(features: Iterator[String]):Map[String,Int] = {
    val empty = Map[String,Int]().withDefaultValue(1)
    features.foldLeft(empty) { case (model, f) => model(f) += 1 }
  }

  def edits1(word: String):Set[String] = {
    val s = (0 to word.length).toList map { i => (word.take(i), word.drop(i)) }
    val deletes = for ((a,b) <- s if b.length > 0) yield a + b.drop(1)
    val transposes = for ((a,b) <- s if b.length > 1) yield a + b(1) + b(0) + b.drop(2)
    val replaces = for ((a,b) <- s if b.length > 0; c <- alphabet) yield a + c + b.drop(1)
    val inserts = for ((a,b) <- s; c <- alphabet) yield a + c + b
    
    Set.empty ++ deletes ++ transposes ++ replaces ++ inserts
  }

  def known(words: Set[String]):Set[String] = words.filter(NWORDS.contains)

  def correct(word: String):String = {
    lazy val wordKnown = known(Set(word))
    lazy val wordEdits1 = known(edits1(word))
    lazy val wordEdits2 = known(edits1(word).flatMap(edits1))

    val candidates = (if (!wordKnown.isEmpty) wordKnown else
    if (!wordEdits1.isEmpty) wordEdits1 else
    if (!wordEdits2.isEmpty) wordEdits2 else Set(word))
      
    candidates.reduceLeft { (a, b) => (if (NWORDS(a) > NWORDS(b)) a else b) }
  } 
}