Designing and Evaluating the Usability of a Machine Learning API for Rapid Prototyping Music Technology

Area of Science:

• Human-Computer Interaction
• Machine Learning Engineering
• Music Technology Software Development

Background:

• Usability and developer experience are crucial for machine learning (ML) tools in creative software.
• Existing research on application programming interface (API) design and evaluation is relevant to ML for music technology.
• Understanding developer needs empowers them as ML users and innovators.

Purpose of the Study:

• To evaluate the usability and developer experience of ML tools for music technology.
• To present the design rationale and evaluation of the RAPID-MIX API for rapid prototyping.
• To provide design recommendations for ML APIs in music technology.

Main Methods:

• Literature review of API design and evaluation in ML for music technology.
• Design and implementation of the RAPID-MIX API for interactive ML.
• Usability evaluation using a cognitive dimensions questionnaire with 12 music technology software developers.

Main Results:

• Participants found the RAPID-MIX API easy to learn and use.
• The API was perceived as fun and effective for rapid prototyping with interactive ML.
• Analysis using the cognitive dimensions framework identified design trade-offs and usability issues.

Conclusions:

• The RAPID-MIX API is a usable and effective tool for rapid prototyping in music technology.
• The cognitive dimensions framework can be applied to evaluate ML APIs, offering valuable insights.
• Design recommendations are provided for future ML APIs aimed at music technology rapid prototyping.