Parameter-Free Attentive Scoring for Speaker Verification