Story-driven approach to software evolution

Juergen Rilling, Wen Jun Meng, René Witte, Philippe CharlandIET Softw 2008 · DOI

TL;DR

This paper introduces a novel approach to software evolution by using an ontology to formally represent processes and knowledge resources, enabling automated reasoning. They propose a "story metaphor" to guide maintainers through complex evolution tasks, making the process more intuitive and interactive.

What Problem Does This Paper Address?

Software evolution, particularly maintenance, is a knowledge-intensive process. Maintainers need to integrate information from various sources (code, documentation, historical data) and follow complex procedures. Existing process models often lack detail, fail to integrate diverse knowledge resources effectively, and don't provide sufficient guidance. This makes it challenging for maintainers to understand and evolve software efficiently, leading to potential errors and reduced quality. The authors aim to bridge this gap by providing a structured, knowledge-driven, and user-friendly approach to software evolution.

Methods

The core of the proposed solution is a formal ontological representation using Description Logics (DL), specifically OWL-DL. This ontology models both software evolution processes (activities, tasks) and relevant knowledge resources (tools, techniques, user expertise). The use of DL allows for automated reasoning services to infer implicit relationships between these elements, enriching the knowledge base.

To make this complex system more accessible, the authors introduce a "story metaphor." This metaphor maps the software evolution process and user interactions onto a narrative structure, making it familiar and intuitive for users. The metaphor is designed to be:

  • Familiar: Leveraging common storytelling elements.
  • Relevant: Closely mapping to software evolution activities.
  • Abstractable: Expressible at different levels of detail.

The system is implemented as a client-server architecture using Eclipse plug-ins, with the server handling ontology management, storage, and reasoning (using Jena and Racer).

Datasets

Not discussed in the paper. The paper focuses on a conceptual framework and its implementation, with case studies described as a method for evaluation and ontology population, rather than using pre-existing datasets for training or benchmarking.

Key Insights and Ideas

  • Ontologies as a Unifying Framework: Formal ontologies provide a powerful way to integrate disparate knowledge sources in software engineering, moving beyond simple data storage to enable reasoning and inference.
  • The Power of Metaphor: A well-chosen metaphor, like storytelling, can significantly improve user adoption and understanding of complex technical systems by providing an intuitive interaction model.
  • Bridging the Gap Between Process and Knowledge: The authors demonstrate how to formally link process steps with relevant knowledge resources, enabling context-aware guidance for maintainers.
  • "Open World" for Evolution: The ontological approach supports an "open world" assumption, allowing for the continuous integration of newly gained knowledge, which is crucial for evolving systems.
  • Reasoning for Implicit Knowledge: Automated reasoning can uncover hidden connections and provide deeper insights than explicitly stated information, which is invaluable in complex maintenance tasks.

Key Results

The paper presents a conceptual model and its implementation. The evaluation is based on case studies involving graduate students performing component substitution tasks on the "Debrief" software.

  • Proof of Concept: The case studies demonstrated that the ontological representation could capture relevant information and model context levels for software maintenance tasks.
  • Ontology Population: After two case studies, the ontology contained 295 concepts, 129 relationships, and 765 instances.
  • Positive Feedback: Maintainers found the integration of process, task, and resource information beneficial.
  • Iterative Refinement: Feedback from case studies was used to extend and refine the ontological model, improving its ability to support task-specific queries.

Limitations and Future Work

  • Ontology Population Challenge: The authors acknowledge that populating the knowledge base (ontology) is a significant challenge, common to all knowledge-based systems. They are exploring strategies like internet community portals and industrial adoption.
  • Limited Scope of Case Studies: The initial case studies were performed in a controlled environment with a restricted task type (component substitution) and resources.
  • Visual Format Under Development: The visual format for the task story model is still under development.

Future Work: * Conduct further case studies to enrich ontologies and optimize the tool. * Address the challenge of ontology population. * Extend the environment by including additional tasks and resources.

Summary updated

BibTeX
@article{rilling2008story,
  author = {Juergen Rilling and Wen Jun Meng and René Witte and Philippe Charland},
  title = {{Story-driven approach to software evolution}},
  journal = {IET Softw.},
  year = {2008},
  doi = {10.1049/iet-sen:20070095},
}
Figure