No Audiogram: Leveraging Existing Scores for Personalized Speech Intelligibility Prediction