The music landscape today is becoming increasingly complex, with a growing diversity of structures and styles. This poses a challenge for many listeners, particularly those without formal training, in expressing their intuitive or abstract musical impressions and preferences in precise and descriptive language. Although current music captioning and tagging models can provide track-level descriptions, they offer limited cross-track insights and cannot fully support listeners to explore the underlying reasons for their music preferences. These limitations motivate exploring whether an interactive interface can strengthen listeners’ ability to articulate the specific musical elements that underpin their taste. We propose Euphrasis, an interactive system that integrates music caption outputs from multiple tracks into a structured representation with a user-friendly interface. Euphrasis assists listeners in verbalizing emotional responses and abstract impressions by iteratively extracting and refining descriptors, specific and isolatable musical keywords, across multiple tracks. By comparing these descriptors, listeners recognize recurring patterns that clarify their music preferences. In preliminary user evaluations, participants produced more precise descriptors and reported higher self‑awareness of taste, revealing the system’s potential to enhance personal engagement with music.