- The paper presents a semantic extension for Jira that automates expert finding by leveraging multi-stage ticket similarity and expert score aggregation.
- It utilizes tf-idf, ontology-based similarity measures, and wiki integration to achieve 82.7% recall and 88.6% precision in expert recommendations.
- The approach enhances organizational knowledge management by reducing duplicate work and expediting issue triage in software projects.
Semantic Expert Recommendation in Bug Tracking: An Analysis of "Semantic Jira"
Introduction
"Semantic Jira - Semantic Expert Finder in the Bug Tracking Tool Jira" (1312.5150) presents a semantic extension to the Jira issue tracking system aimed at automating expert finding and collaborative knowledge reuse in software engineering project environments. The proposed approach leverages semantic search and ontology-based information extraction to identify similar tickets, recommend relevant experts, and link organizational knowledge bases, thus reducing redundant work and enhancing collaborative maintenance and development processes.
Related Work and Contextual Positioning
The methodology distinguishes itself from prior research by forgoing reliance on source code analysis and instead focusing on data natively present within bug tracking systems. Approaches such as Matter et al. (2011), McDonald & Ackerman (2000), and Reichelt & Wulf (2009) primarily depend on prolonged developer interaction with code repositories or proprietary knowledge bases, impeding applicability to non-IT users or requiring extensive ramp-up. By integrating semantic analysis directly within Jira, this system achieves tighter coupling with operational workflows and addresses limitations related to usability and accessibility for diverse project roles. Furthermore, the workflow-centric design enhances discoverability of tacit and explicit expertise, addressing gaps in previous expert recommender solutions that lack deep integration with task management systems.
Methodology: Statistical and Ontological Expert Inference
The expert finder introduced in Semantic Jira operates via a multi-stage retrieval and inference model:
- Semantic Ticket Similarity: Ticket similarity is first measured using tf-idf weighted keyword extraction from ticket descriptions, leveraging vector space similarity metrics to locate related historical tickets. Supplementing lexical analysis, a domain ontology taxonomy is applied to capture deeper semantic relatedness between ticket pairs through taxonomic distance functions and hierarchical concept similarity measures.
- Expert Identification: For each set of similar tickets, "experts" are inferred by aggregating roles (reporter, assignee, follower). An expert score is accumulated based on the volume and relevance of their contributions to topic-proximal tickets, comprising both developer and organizational dimensions.
- Wiki-Based Expertise Augmentation: Integration with organizational wikis enhances expert profiling. Contributions to wiki sections mapped to ontology concepts are weighted both by specificity (hierarchy level in the ontology) and by semantic relatedness to the ticket domain. The final expertise score for an individual is thus consolidated across both direct and related knowledge areas, using ontology-based relevance propagation.
- Implementation: The system is realized through a Jira plugin interfacing with Lucene/Solr for indexing and retrieval, and the CSW Semantic Similarity Matchmaking Framework for ontology-based processing. RESTful interfacing ensures compatibility and extensibility with Jira’s core components and MediaWiki systems.
Empirical Evaluation
A user study conducted in a midsized enterprise IT context assessed system precision and recall against human expert assignment. On a sample of 32 tickets and 98 ground-truth expert annotations:
- 82.7% recall demonstrates the system’s high coverage in proposing relevant experts.
- 88.6% precision within the top-10 system recommendations confirms effective ranking, with 45.7% of correct experts appearing at the top position.
- The average number of experts suggested per ticket was 10.03, indicating balanced selectivity.
The evaluation also establishes the influence of configurable parameters such as the number of relevant keywords (k) on performance, reflecting the system's flexibility for further tuning.
Implications and Future Directions
The results support the practical viability of ontology-driven, semantic expert recommendation systems integrated within bug tracking tools. For organizational knowledge management, this approach enhances the discoverability of both explicit (documented) and implicit (activity-inferred) expertise. Practically, semantic recommender systems can reduce duplication in bug resolution, expedite issue triage, and foster knowledge transfer across technical and non-technical roles.
The study suggests several promising extensions, such as multi-lingual support and integration with external knowledge sources (e.g., DBpedia, WordNet) to further enhance the semantic reach and robustness of expert inference. The results also motivate the incorporation of more nuanced user interaction and historical activity modeling—e.g., assignment change histories, ticket status dynamics—to optimize expertise scoring.
Conclusion
"Semantic Jira - Semantic Expert Finder in the Bug Tracking Tool Jira" (1312.5150) delivers a comprehensive framework for semantic, ontology-based expert recommendation in enterprise issue tracking. By leveraging both statistical and taxonomic similarity, and consolidating expertise across ticket and wiki activities, the system demonstrates robust empirical accuracy and meaningful workflow integration. The approach informs future research avenues in organizational knowledge discovery, expert recommendation, and semantic enterprise search, particularly as software engineering processes grow in complexity and collaboration scope.