Selected work

Representative Papers

Multilanguage Detection of Design Pattern Instances

Feb 01, 2025Journal of Software: Evolution and Process

Existing design pattern detection tools are predominantly language-specific, hindering consistent identification of pattern instances across multi-language codebases. To address this, we propose DP-LARA—a cross-language design pattern detection approach built upon the LARA framework and a novel virtual Abstract Syntax Tree (vAST). DP-LARA establishes the first unified, language-agnostic pattern matching paradigm by mapping Java and C/C++ source code to semantically equivalent vAST representations, then applying static analysis augmented with an extensible rule engine for precise and consistent pattern instance identification. Experimental evaluation demonstrates that DP-LARA achieves detection accuracy on par with state-of-the-art single-language tools for both Java and C/C++, while significantly improving cross-language result consistency and reducing language adaptation effort by over 60%. This work introduces a scalable, maintainable paradigm for multi-language software architecture analysis.

20 citations1 influentialRead paper

CitiLink-Minutes: A Multilayer Annotated Dataset of Municipal Meeting Minutes

Feb 12, 2026

This study addresses the scarcity of high-quality annotated data in municipal council minutes, which has hindered the application of natural language processing (NLP) to research on local governance transparency. To bridge this gap, the authors present a multilayer-annotated dataset comprising 120 Portuguese-language municipal meeting records, structured across three dimensions: metadata, agenda items, and voting outcomes. This resource is the first to offer structured interlinking and dual manual annotation, rigorously validated by linguists and anonymized to ensure privacy. Adhering strictly to FAIR principles, the dataset encompasses over one million tokens and more than 38,000 annotations. Accompanied by baseline model performance across multiple NLP tasks, it fills a critical void in local governance text resources and supports the development of downstream applications.

2 citationsRead paper

CitiLink: Enhancing Municipal Transparency and Citizen Engagement through Searchable Meeting Minutes

Jan 26, 2026

This study addresses the challenge posed by lengthy and unstructured municipal meeting minutes, which hinder efficient public access to critical information. To overcome this, the authors propose a novel framework that integrates large language models—such as Gemini—with traditional retrieval techniques. The approach leverages the language model to automatically extract metadata and agenda items, which are then combined with BM25-based full-text search and faceted filtering to create an interactive, searchable, and structured system. Evaluated on 120 meeting minutes from six Portuguese municipalities, user studies demonstrate that the system significantly enhances usability and information retrieval efficiency, with Gemini excelling in the information extraction task. This work offers a scalable technical pathway to improve government transparency and civic engagement.

1 citationsRead paper

The Role of Deep Learning in Financial Asset Management: A Systematic Review

Mar 03, 2025

This study addresses the lack of systematic analysis on the evolution of deep learning (DL) in financial asset management. We conduct a bibliometric and systematic literature review of 612 high-quality empirical studies (2018–2023) indexed in Scopus, focusing on three emerging trends: the integration of eXplainable AI (XAI) with Deep Reinforcement Learning (DRL), Transformer-based hybrid architectures, and fusion of alternative data—including ESG metrics and sentiment signals. Our method identifies critical gaps and synthesizes methodological advances across domains. Key contributions include: (i) proposing the first XAI-DRL joint framework, enhancing both transparency and dynamic adaptability of investment decisions; (ii) empirically validating that alternative-data-driven models achieve simultaneous improvements in interpretability and robustness. Results show DL adoption improves portfolio Sharpe ratios by an average of 22% and reduces mean absolute error in price forecasting by 17–34%.

1 citationsRead paper

Improving Debugging in Verification-Aware Languages Through Automated Fault Localization: A Case Study in Dafny

Aug 05, 2026

This work addresses the limited diagnostic feedback provided by verification-aware languages like Dafny upon verification failure, which forces developers to manually inspect execution traces to identify root causes. To alleviate this burden, the paper proposes two automated fault localization strategies—state-based and counterexample-based—that leverage verifier-generated counterexamples through structured single-trace ranking and multi-trace aggregation. Evaluated on MutDafny mutants and the DafnyBench benchmark using EXAM scores, the results demonstrate that counterexample-based methods substantially outperform traditional state-based approaches. Notably, structured single-trace ranking yields the most significant improvement, while multi-trace aggregation further enhances robustness and practical utility for debugging.

0 citationsRead paper

Recent publications

Latest Papers

Improving Debugging in Verification-Aware Languages Through Automated Fault Localization: A Case Study in Dafny

Aug 05, 2026

This work addresses the limited diagnostic feedback provided by verification-aware languages like Dafny upon verification failure, which forces developers to manually inspect execution traces to identify root causes. To alleviate this burden, the paper proposes two automated fault localization strategies—state-based and counterexample-based—that leverage verifier-generated counterexamples through structured single-trace ranking and multi-trace aggregation. Evaluated on MutDafny mutants and the DafnyBench benchmark using EXAM scores, the results demonstrate that counterexample-based methods substantially outperform traditional state-based approaches. Notably, structured single-trace ranking yields the most significant improvement, while multi-trace aggregation further enhances robustness and practical utility for debugging.

0 citationsRead paper

WiFuse: An Attention Mechanism for Human Activity Recognition using Fused CSI Amplitude and Delay-Doppler Channel Features

Aug 01, 2026

Wi-Fi signals are highly susceptible to multipath effects and noise interference, which degrade the performance of channel state information (CSI)-based human activity recognition. To address this, this work proposes WiFuse, a novel framework that, for the first time, integrates denoised CSI amplitude variations and phase-derived Delay-Doppler two-dimensional motion representations within a dual-stream architecture. The framework combines ResNet with a temporal convolutional network (TCN) and incorporates both channel-wise and spatiotemporal attention mechanisms to enhance feature learning. Furthermore, a two-stage decoupled transfer learning strategy is introduced to mitigate domain shift and class overlap challenges. Evaluated on the XRF55 and Wi-MIR datasets, WiFuse achieves recognition accuracies of 95.28% and 98.20%, respectively, significantly outperforming existing state-of-the-art methods.

0 citationsRead paper

Linearising Explicit Substitutions using Intersection Types

Jul 22, 2026

This study addresses the challenge of establishing a precise correspondence between explicit substitution λ-calculi and resource-aware linear calculi under constraints of limited parameter availability. To this end, the work introduces a novel term expansion technique that integrates intersection types with a multiplicity mechanism, mapping explicit substitution λ-terms into weakly linear terms in Boudol’s resource-sensitive λ-calculus. This is the first extension of term expansion methods to calculi with explicit substitution, achieving semantic alignment between strongly normalizing terms and weakly linear terms while guaranteeing they share identical normal forms. The result provides a foundational theoretical framework for integrating substructural type systems with explicit substitution formalisms.

0 citationsRead paper

Verified LLM-Driven Synthesis for Concept Design

Jul 17, 2026

This work addresses the challenge in conceptual design where relying solely on safety invariants often fails to uniquely determine reaction rules aligned with user intent, leading to inconsistent synthesis outcomes. To overcome this, the authors propose a novel synthesis approach that integrates formal semantics with an LLM-driven Counterexample-Guided Inductive Synthesis (CEGIS) framework. The method leverages either positive/negative example scenarios or natural language prompts to guide the generation of rules satisfying given safety invariants, and introduces, for the first time, an LLM-assisted scenario-based elicitation mechanism to support early-stage design exploration. As the first effort to combine formal verification with LLM-based synthesis in conceptual design, experiments demonstrate that scenario-based guidance more reliably reproduces intended designs than natural language alone; with sufficient scenarios, LLM-augmented elicitation effectively recovers expected behaviors for most variants, though behavior omission and non-determinism remain key obstacles to achieving full coverage.

0 citationsRead paper

From Mobile Data to Business Insights: An End-to-End Analytics Framework for Large-Scale Urban Mobility Analysis and Decision Support

Jul 03, 2026

This study addresses the growing demand for mobility insights across urban planning, transportation, and retail by proposing an end-to-end urban mobility analytics framework. Built upon a reusable modular architecture, the framework integrates high spatiotemporal resolution mobility modeling with multi-scenario business applications. It establishes a closed-loop pipeline—from raw geolocation data to strategic insights—through anonymization, ETL workflows, BigQuery-based data management, Vertex AI–driven model training, and Power BI visualization. The system effectively supports diverse analytical tasks, including traveler profiling, trajectory mining, catchment area analysis, traffic anomaly detection, and origin–destination pattern recognition, thereby delivering scalable and efficient decision support for both urban governance and commercial strategy.

0 citationsRead paper