WAR-CRAFT: Refining Web API Recommendations through Requirement Auditing
Summary
- Researchers introduced WAR-CRAFT, a post-recommendation refinement framework for web API recommendations that checks outputs against provider documentation.
- The method uses an LLM to remove unsupported or redundant APIs, audits requirements against a checklist, and employs a catalog-constrained Resolver for uncovered needs.
- Tested on ProgrammableWeb with seven upstream recommenders, WAR-CRAFT raised mean SetF1 from 0.558 to 0.644 and halved functional redundancy.
- The approach outperformed five refinement baselines and showed performance gains ranging from +2.2 to +18.2 points depending on the upstream recommender type.
Why it matters
WAR-CRAFT improves the accuracy and reliability of automated web API composition by filtering out unsupported recommendations and reducing redundancy, addressing key limitations in API recommender systems.