Welcome to the Software Maintenance and Analytics Research Team (SMART lab), established at the Faculty of Computer Science of the Dalhousie University. The lab is lead by Tushar Sharma. We research, develop, and disseminate methods, tools, techniques associated with Software engineering (specifically software design and architecture, code quality, refactoring, technical debt, and program comprehension) and machine learning techniques for software engineering applications.

News

2025

• Grant - Digital Research Alliance of Canada (DRA) granted computing resources worth more than $49K via Resource Allocation Competition, Co-PI with Dr. Masud Rahman, Mar 2025.
• Publication - "Reinforcement~Learning vs Supervised~Learning: A tug of war to generate refactored code accurately". Accepted in EASE (Research track) 2025. Mar 2025.
• Publication - "TS-Detector : Detecting Feature Toggle Usage Patterns". Accepted in FSE (Tools track) 2025. Mar 2025.
• Media coverage - "The artificial-intelligence boom is straining environmental and infrastructural limits", Feb 2025.
• Publication - "An adaptive language-agnostic pruning method for greener language models for code". Accepted in FSE (Research track) 2025. Jan 2025. Preprint
• Publication - "MaRV: A Manually Validated Refactoring Dataset". Accepted in FORGE 2025 (Benchmarking track). Jan 2025.
• Publication - "DPy: Code Smells Detection Tool for Python". Accepted in MSR (tools track) 2025. Jan 2025. Preprint
• Publication - "It Works (only) on My Machine: A Study on Reproducibility Smells in Ansible Scripts". Accepted in MSR (research track) 2025. Jan 2025. Preprint

2024

• Publication - "COMET: Generating Commit Messages using Delta Graph Context Representation". Accepted in JSS. Dec 2024. Preprint
• Publication - "On Inter-dataset Code Duplication and Data Leakage in Large Language Models". Accepted in TSE. Nov 2024. Preprint
• Media coverage - "The hidden cost of the AI revolution: carbon emissions", Oct 2024.
• Media coverage - "This Dal researcher wants to ensure AI doesn’t ruin the environment", Sept 2024.
• Grant - Mitacs Accelerate: "Fine-tuning an LLM for patent drafting" ($30K); industry partner: XLSCOUT, Aug 2024.
• Publication - "Enhancing Identifier Naming Through Multi-Mask Fine-tuning of Language Models of Code". Accepted in IEEE SCAM 2024. Preprint
• Publication - "Enhancing Energy-Awareness in Deep Learning through Fine-Grained Energy Measurement". Accepted in TOSEM. Preprint
• Publication - "Broken Windows: Exploring the Applicability of a Controversial Theory on Code Quality". Accepted in ICSME 2024. Preprint
• Publication - "Exploring Influence of Feature Toggles on Code Complexity". Accepted in EASE 2024. Preprint
• Grant - Digital Research Alliance of Canada (DRA) granted computing resources worth more than $51K via Resource Allocation Competition, Co-PI with Dr. Masud Rahman, Mar 2024.
• Grant - Mitacs Accelerate: "Exploring the connection between internal and external software quality attributes" ($225K); industry partner: Spokesfan, Mar 2024.
• Technical briefing - "LLMs for code: the potential, prospects, and problems", Accepted in International Conference on Software Architecture (ICSA) 2024.
• Publication - "Benchmarking Emerging Deep Learning Quantization Methods for Energy Efficiency", Accepted in GREENS 2024. Preprint
• Grant - Mitacs Accelerate: "Subsystem Optimization of Intelligent Automated Distribution Kiosk" ($15K), Mar 2024.
• Publication - "Greenlight: Highlighting TensorFlow APIs Energy Footprint", Accepted in MSR (Data/Tools track) 2024. Preprint
• Publication - "Multi-faceted Code Smell Detection at Scale using DesigniteJava 2.0". Accepted in MSR (Data/Tools track) 2024. Preprint

2023

• Publication - "A Survey on Machine Learning Techniques Applied to Source Code", Accepted in Journal of Systems and Software, Dec 2023. Preprint
• Publication - "Naturalness of Attention: Revisiting Attention in Code Language Models", Accepted in ICSE (NIER), Nov 2023. Preprint
• Award - Best artifact award for SCAM 2023 - "Calibrating Deep Learning-based Code Smell Detection using Human Feedback"
• Grant - Lab2Market grant for CoderGate, Sept 2023.
• Publication - "Calibrating Deep Learning-based Code Smell Detection using Human Feedback", Accepted in IEEE SCAM, Aug 2023. Preprint
• Publication - "Mining and Fusing Productivity Metrics with Code Quality Information at Scale", Accepted in IEEE ICSME (Tools track), Aug 2023. Preprint
• Publication - "Automatic Refactoring Candidate Identification Leveraging Effective Code Representation", Accepted in IEEE ICSME (NIER track), Aug 2023. Preprint
• Publication - "Investigating Developers' Perception on Software Testability and its Effects". Accepted in Empirical Software Engineering Journal, Jul 2023. Preprint
• Grant - Canada First Research Excellence Fund (CFREF) accepted to fund "Transforming Climate Action: Addressing the Missing Ocean" for $154 million for the next seven years. Glad to be one of the 170 researchers' team from Dalhousie to participate in this program.
• Grant, Discovery grant - "DevQOps: A Continuous Software Quality Assurance Framework", Accepted to be funded for five years by NSERC, Apr 2023
• Publication - "DACOS-A Manually Annotated Dataset of Code Smells", Mar 2023, Accepted at MSR 2023 (dataset and tools track).
• Grant - Lab2Market grant for QConnect, Jan 2023.

2022 and older

• Publication - "Green AI: Do Deep Learning Frameworks Have Different Costs?", Dec 2021, Accepted at ICSE 2022 (Research track). Preprint
• Publication - "Lessons from Research to Practice on Writing Better Quality Puppet Scripts". in SANER, 2022. Preprint