TL;DR
This paper presents a systematic roadmap and literature review of approaches that map, transform, or integrate goal models with other software development artifacts. It reveals that while significant work exists, the field is fragmented, with many approaches focusing on specific stages and lacking extensive evaluation, suggesting the field is still in its early stages of maturity.
What Problem Does This Paper Address?
Goal models are valuable for capturing requirements in the early stages of software development. However, their utility can be significantly enhanced if they are integrated into downstream phases of the software lifecycle, such as architectural design, coding, testing, and system evolution. The challenge lies in transforming the qualitative and social nature of goal models (e.g., softgoals, roles) into concrete, functional system elements. This paper aims to understand the current landscape of approaches that bridge this gap, asking: what methods exist to map, integrate, or transform goal models into later-stage software artifacts?
Methods
The authors employ a systematic literature mapping and review approach, inspired by Evidence-Based Software Engineering. Their methodology involves:
- Systematic Search: A comprehensive search of research databases using keywords related to requirements engineering, software engineering, goal models, and transformation/mapping/integration.
- Snowballing: Starting with a core set of relevant papers and expanding the search by examining their references.
- Roadmap Creation: Summarizing 243 identified publications based on inclusion/exclusion criteria, categorizing them by transformation type, source/target artifacts, motivations, and venues.
- In-depth Literature Review: A detailed analysis of the top 50 most cited papers (based on citations per year) to gain deeper insights into the research landscape.
- Taxonomy Development: Creating taxonomies for "how" (source/target of transformation) and "why" (motivations) tags using a grounded theory approach to classify publications.
Datasets
Not discussed in the paper. The study focuses on analyzing existing research publications rather than using specific datasets for evaluation.
Key Insights and Ideas
- Fragmentation and Isolation: The field is characterized by many small, isolated groups of researchers, hindering the consolidation of knowledge and leading to the proposal of similar approaches with minor variations.
- Dominance of "Solution" Papers: The vast majority of publications are classified as "solution" papers, indicating a lack of extensive application or evaluation, suggesting the field's immaturity.
- The "Softgoal" Challenge: Transforming the qualitative and "fuzzy" aspects of goal models, particularly softgoals, remains a significant challenge, with many approaches either ignoring them or handling them informally.
- Downstream Focus: While goal models are traditionally for early RE, a surprising number of techniques transform them to downstream artifacts, indicating their utility beyond the initial requirements phase.
- "Schools" of Thought: Research tends to cluster around specific goal modeling frameworks (e.g., KAOS, Tropos, i*, GRL), with papers within these "schools" being more likely to cite each other.
Key Results
- Roadmap Summary: A roadmap of 243 publications was created, identifying various transformation, mapping, and integration approaches.
- Transformation Types: Mappings are the most frequent technique, particularly in the horizontal direction. Vertical transformations are also significant, especially for moving downstream. Exogenous transformations (from goal models to other artifacts) are dominant.
- Goal Model Sources: "Unidentified Goal Model" is the most frequent source, followed by i* and KAOS.
- "How" Taxonomy: The "how" taxonomy reveals a wide variety of source and target artifacts, with significant activity in transforming goal models to software artifacts, business processes, and modeling languages (especially non-UML).
- "Why" Taxonomy: Enhanced RE is the most frequent motivation, followed by Software Quality, Paradigm, and Change Management.
- Trends: Interest in goal model integration has seen peaks and dips, with a recent rise. Transformations from goal models to non-UML languages are increasing, while transformations to UML seem to be declining.
- Citation Analysis: The top 50 cited papers are heavily weighted towards a few highly cited works, with a significant portion of citations attributed to the original Tropos paper. Cross-citation analysis reveals a relatively low level of interconnectedness among papers, suggesting isolation.
Limitations and Future Work
- Study Completeness: The study acknowledges that the snowballing process was not fully completed, and the systematic search might not have captured all relevant literature.
- Subjectivity: The inclusion/exclusion criteria, classification of transformation types, and tagging of papers involved some degree of subjectivity, though mitigated by multiple reviewers and group discussions.
- DBLP Limitations: The use of the DBLP database for citation analysis has limitations in coverage, particularly for papers outside of core computer science.
- Future Work: The authors suggest applying their process to a broader set of research papers, examining the "how" of utilization (not just reference) in citation networks, and integrating their approach with other methods for analyzing literature. They also highlight the need for more research on transforming softgoals and for more extensive evaluation and application of proposed techniques.
Summary updated