Enhancing project management with AI - An NLP based Jira plugin for detecting similar tickets

This thesis presents the design and implementation of a Natural Language Processing (NLP) based Jira plugin that detects similar tickets to improve efficiency and decision making in software project management. Large projects often suffer from scattered tickets that are semantically related, leading to loss of historical project knowledge, redundant effort, and delayed or inconsistent resolution. Traditional Jira Query Language (JQL) keyword-based search mechanisms are effective only when tickets share overlapping terms and are limited in capturing semantic meaning, making accurate identification of similar tickets difficult. To address this challenge, the proposed solution applies NLP techniques that combines TF-IDF with cosine similarity and Sentence-BERT based semantic embeddings. The TF-IDF component captures lexical similarity or tickets with overlapping words using ticket summaries and descriptions, while the semantic model identifies conceptual relationships that go beyond exact keyword matching. Text normalization techniques, including tokenization, phrase-level mapping, and domain-specific vocabulary standardization, are applied to enhance consistency in the lexical similarity model. In addition, metadata-based heuristic boosting using Jira fields further refines the similarity scores between tickets. The plugin is implemented using Atlassian Forge and follows a modular architecture, with semantic similarity computation handled by a lightweight FastAPI service to support scalability and efficient deployment. Evaluation on curated Jira datasets using precision, recall, F1-score, and response time demonstrates that the proposed approach outperforms traditional JQL-based keyword search. The results show that the plugin effectively identifies similar tickets, improves issue traceability, promotes better use of historical project knowledge, and supports efficient project planning and management workflows.