Institution profile

University of Pisa

Academic institutioneurope · it
Official website
Research library437linked papers
Opportunities0open roles
Selected work

Representative Papers

Geo-semantic-parsing: AI-powered geoparsing by traversing semantic knowledge graphs

Jul 07, 2020Decision Support Systems

Geographic information in online social media text is often unstructured, ambiguous, or abbreviated, hindering real-time geolocation. To address this, we propose an end-to-end geoparsing method integrating semantic understanding and knowledge graph reasoning. First, BERT encodes contextual semantics; second, a semantic knowledge graph (GeoNames/Wikidata) traversal mechanism—implemented via a graph neural network (GNN)—models hierarchical geographic relations and resolves toponym ambiguity; third, a constraint-aware coordinate optimization algorithm outputs precise latitude-longitude coordinates. Our approach is the first to systematically incorporate knowledge graph path reasoning into the geoparsing pipeline, overcoming limitations of conventional dictionary-based matching and statistical models. Evaluated on the GeoCorpora dataset, it achieves an F1-score of 89.3%, outperforming the state-of-the-art by 6.2 percentage points, and demonstrates significantly improved robustness to fuzzy, abbreviated, and historical toponyms.

31 citations1 influentialRead paper

Declarative Application Management in the Fog

Oct 27, 2021Journal of Grid Computing

To address the lack of decentralized, adaptive application management mechanisms for massive heterogeneous devices in fog computing environments, this paper proposes the first declarative application management framework tailored for fog computing. The framework employs high-level intent specifications (in YAML) to decouple user requirements from underlying heterogeneous execution details, enabling automated cross-edge-and-cloud deployment, elastic scaling, and self-healing capabilities. Its key innovation lies in systematically introducing the declarative paradigm into fog computing lifecycle management, realized through a closed-loop control architecture comprising a Kubernetes-based extensible orchestration layer, lightweight runtime agents, and an event-driven coordinator. Evaluated on a real-world multi-domain fog testbed, the framework achieves a 42% reduction in deployment latency, sub-1.2-second policy update response time, and a 28% improvement in resource utilization.

6 citationsRead paper

ranx: A Blazing-Fast Python Library for Ranking Evaluation and Comparison

Mar 04, 2025European Conference on Information Retrieval

To address the dual challenges of accuracy and fairness in expert user identification within community question-answering platforms, this paper introduces ranx—a high-performance ranking evaluation library. Implemented entirely in NumPy and Cython, ranx features a novel sub-millisecond parallel multi-metric computation architecture, enabling pairwise and multi-group statistical significance testing (e.g., Wilcoxon signed-rank test, t-test) across arbitrary numbers of rankers. Through memory mapping and vectorized logic, it achieves extreme computational efficiency. On TREC and MSMARCO benchmarks, ranx outperforms ir-measures and prior versions of ranx by 10–100×, supporting real-time evaluation over million-scale queries. Its core innovation lies in the tight integration of ranking evaluation, statistical inference, and systems-level engineering optimization. As an open-source infrastructure, ranx provides both flexibility and scalability, advancing reproducible and verifiable research in information retrieval and recommender systems.

4 citationsRead paper

A Frank System for Co-Evolutionary Hybrid Decision-Making

Mar 08, 2025International Symposium on Intelligent Data Analysis

Accurately and fairly labeling unlabeled datasets remains challenging due to the inherent limitations of both purely human and purely AI-driven approaches. Method: This paper proposes FRANK, a human–AI co-evolutionary hybrid decision-making framework that jointly models human cognitive patterns and machine learning models through bidirectional co-evolution. FRANK integrates Bayesian optimization, interpretable neuro-symbolic reasoning, and real-time human-in-the-loop reinforcement learning to dynamically optimize both model architecture and decision policies. Contribution/Results: Evaluated on multi-agent collaborative tasks, FRANK improves decision quality by 37% and reduces human annotation effort by 52%, while achieving significantly stronger generalization than either fully automated or fully manual baselines. Its core innovation lies in establishing an interpretable, tunable collaborative decision paradigm that supports bidirectional evolution between human cognition and algorithmic processes.

2 citationsRead paper

A taxonomy of categories for relations

Feb 14, 2025arXiv.org

This paper addresses the lack of a unified classification framework for the structural properties of relational abstractions. Methodologically, it introduces the first hierarchical and systematic taxonomy of relational categories, grounded in the Kleisli category of the symmetric monoidal monad as a unifying generative mechanism. This framework subsumes diverse relational structures—including relational database schemas, program semantics models, and relational representations in AI—along with their enriched variants, within a single categorical setting. The key contribution is the identification of a common origin: all major relational categories in the literature arise as instances of this monadic Kleisli construction. By exposing this deep structural unity, the taxonomy enhances theoretical coherence and conceptual clarity. It provides a rigorous, general mathematical foundation applicable across program semantics, database theory, and AI-based relational modeling.

2 citationsRead paper
Recent publications

Latest Papers

Control-Theoretic Content Moderation

Sep 16, 2026

本文通过引入一种基于反馈控制的全局、自适应和序列决策框架来解决平台级内容管理策略问题，该方法在平衡多个竞争目标方面优于局部策略。

0 citationsRead paper