Abstract
This paper investigates how segment-based audio retrieval can support professional music curation through Talaria, a human-centered system that allows curators to query with perceptually salient excerpts (e.g., choruses, vocal lines) and refine results via acoustic and editorial filters to prioritize control, flexibility, and interpret ability over automated recommendations. Developed through a Research through Design (RtD) process and evaluated with fourteen professionals from a major Asian streaming platform, the study combined semi-structured interviews, a scenario-based questionnaire, and usability testing to examine three research questions: (i) how segment-based retrieval supports creativity and control; (ii) where it is most effective or inadequate; and (iii) which design factors shape trust in AI-assisted tools. Findings show that Talaria enhances early-stage exploration and mid-stage stylistic refinement, excelling in perceptual tasks such as playlist flow optimization and context-specific selection, while exhibiting limits in concept-driven work that depends on symbolic reasoning or lyrical interpretation (e.g., artist branding, narrative playlists). Trust hinged on perceived consistency, transparency, and editorial autonomy, and was strengthened by lightweight, interaction-level explanations that clarify similarity logic without constraining judgment. We distill design implications for human-centered curation systems: treat interaction-level interpret ability as a first-class requirement, preserve user agency through non-prescriptive controls, adopt genre-aware segmentation and preview strategies, and support fluid, iterative workflows. The work offers design evidence that segment-level retrieval can augment professional curation when interpret ability and agency are built into the interaction.
Keywords
Segment-based retrieval; Human-centered AI; Music curation; Interpretability
DOI
https://doi.org/10.21606/iasdr.2025.87
Citation
Chen, Y., Lou, J.,and Guan, S.(2025) Talaria: Designing Segment-Based Audio Retrieval to Support Human-Centered Music Curation, in Chang, C.-Y., and Hsu, Y. (eds.), IASDR 2025: Design Next, 02-05 December, Taiwan. https://doi.org/10.21606/iasdr.2025.87
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
Conference Track
Track 4 - Human-Centered AI
Talaria: Designing Segment-Based Audio Retrieval to Support Human-Centered Music Curation
This paper investigates how segment-based audio retrieval can support professional music curation through Talaria, a human-centered system that allows curators to query with perceptually salient excerpts (e.g., choruses, vocal lines) and refine results via acoustic and editorial filters to prioritize control, flexibility, and interpret ability over automated recommendations. Developed through a Research through Design (RtD) process and evaluated with fourteen professionals from a major Asian streaming platform, the study combined semi-structured interviews, a scenario-based questionnaire, and usability testing to examine three research questions: (i) how segment-based retrieval supports creativity and control; (ii) where it is most effective or inadequate; and (iii) which design factors shape trust in AI-assisted tools. Findings show that Talaria enhances early-stage exploration and mid-stage stylistic refinement, excelling in perceptual tasks such as playlist flow optimization and context-specific selection, while exhibiting limits in concept-driven work that depends on symbolic reasoning or lyrical interpretation (e.g., artist branding, narrative playlists). Trust hinged on perceived consistency, transparency, and editorial autonomy, and was strengthened by lightweight, interaction-level explanations that clarify similarity logic without constraining judgment. We distill design implications for human-centered curation systems: treat interaction-level interpret ability as a first-class requirement, preserve user agency through non-prescriptive controls, adopt genre-aware segmentation and preview strategies, and support fluid, iterative workflows. The work offers design evidence that segment-level retrieval can augment professional curation when interpret ability and agency are built into the interaction.