Speech Understanding in a Multiple Recognizer with an Anaphora Resolution Process