You could do the following:
trait Converter[T] {
  def convert(attrType: String, attrValue: String): T
}
object ConverterTest {
  implicit object IntConverter extends Converter[Int] {
    def convert(attrType: String, attrValue: String): Int = {
      attrType match {
        case "N" => attrValue.toInt
        case _ => -1
      }
    }
  }
  implicit object StringConverter extends Converter[String] {
    def convert(attrType: String, attrValue: String): String = {
      attrType match {
        case "S" => attrValue
        case _ => ""
      }
    }
  }
  def to[T: Converter](attrType: String, attrValue: String): T = {
    implicitly[Converter[T]].convert(attrType, attrValue)
  }
  def main(args: Array[String]) {
    println(to[String]("S", "B"))
    println(to[String]("N", "B"))
    println(to[Int]("S", "23"))
    println(to[Int]("N", "23"))
  }
}
Its more code, and I couldn't get type inferencing to work, so it is probably of limited use.
But it is a single method plus a bunch of converters that can get controlled at the call site, so you get some extra flexibility.
Is it worth the effort? Depends on the actual use case.