axle.nlp.UnweightedDocumentVectorSpace.scala Maven / Gradle / Ivy

Go to download

Show more of this group Show more artifacts with this name
Show all versions of axle-core_2.11 Show documentation

axle-core

The newest version!

package axle.nlp

import spire.algebra.Field
import spire.algebra.InnerProductSpace
import spire.algebra.MetricSpace
import spire.implicits.DoubleAlgebra

object UnweightedDocumentVectorSpace {

  /**
   *
   * distance(v1: TermVector, v2: TermVector) = 1 - dot(v1, v2) / (norm(v1) * norm(v2))
   *
   */

  def apply(_stopwords: Set[String],
    corpusIterable: Iterable[String]): DocumentVectorSpace =
    new DocumentVectorSpace {

      val _vectors = corpusIterable.iterator.map(doc2vector).toIndexedSeq

      def stopwords: Set[String] = _stopwords

      val innerProductSpace = new InnerProductSpace[TermVector, Double] {

        def negate(x: TermVector): TermVector = x.map(kv => (kv._1, -1 * kv._2)) // Not sure this makes much sense

        def zero: TermVector = Map()

        def plus(x: TermVector, y: TermVector): TermVector =
          (x.keySet union y.keySet).toIterable.map(k => (k, x.get(k).getOrElse(0) + y.get(k).getOrElse(0))).toMap

        def timesl(r: Double, v: TermVector): TermVector = v.map(kv => (kv._1, (kv._2 * r).toInt))

        def scalar: Field[Double] = DoubleAlgebra

        def dot(v1: TermVector, v2: TermVector): Double =
          (v1.keySet intersect v2.keySet).toList.map(w => v1(w) * v2(w)).sum

      }

      def space: MetricSpace[TermVector, Double] = innerProductSpace.normed

    }

}