Service recommendation method based on text view and interaction view.

Area of Science:

• Software Engineering
• Artificial Intelligence
• Data Mining

Background:

• Web service reuse is prevalent in software development for cost-effectiveness and functionality.
• Developers face challenges selecting appropriate web services from numerous options on sharing platforms.
• Existing recommendation systems struggle with arbitrary user input and sparse interaction data.

Purpose of the Study:

• To propose a novel service recommendation method, SRTI, addressing limitations in current web service discovery.
• To enhance the accuracy of web service recommendations by integrating multiple data perspectives.

Main Methods:

• Utilized graph neural networks to extract features and preferences from historical application and service interaction data.
• Employed Transformer models for analyzing development requirements and fully connected neural networks for matching candidate services.
• Integrated text and interaction views to generate a final ranked list of recommended services.

Main Results:

• The proposed SRTI method demonstrated superior performance compared to state-of-the-art methods in web service recommendation.
• Extensive experiments on real-world datasets validated the effectiveness of SRTI in improving service discovery.

Conclusions:

• SRTI effectively addresses the challenges of arbitrary user input and sparse interaction data in service recommendation.
• The integrated approach of SRTI offers a more robust and accurate solution for software developers seeking web services.