A Folksonomy-Based Social Recommendation System for Scientific Workflow Reuse

In the past decade, scientific workflow systems have significantly improved scientists´ ability to structure scientific processes, use computational resources, and analyze their data more efficiently. Such productivity can be further enhanced by sharing, reusing, and repurposing existing tasks and workflows across different users and institutes. However, existing scientific workflow systems are mainly single-user oriented with limited sharing and reusing functionalities. To overcome such limitations, we propose a folksonomy-based social workflow recommendation system to improve workflow design productivity. Our contributions are: i) We developed a web-based workflow design environment (called Web bench) to allow users to create workflows and collaboratively annotate and categorize them using social tags. The resulted folksonomy improves workflow search ability and share ability. ii) We proposed several workflow recommendation strategies to automatically or semi-automatically augment an in-progress workflow, leveraging both structural and semantic similarities between workflows and guiding information extracted from previously created workflows in the database. iii) We implemented the proposed environment and strategies in a prototype based on the DATAVIEW scientific workflow management system and validated our approach with numerous use cases.