- Semantic Web and Ontologies
- Recommender Systems and Techniques
- Service-Oriented Architecture and Web Services
- Topic Modeling
- Advanced Database Systems and Queries
- Advanced Graph Neural Networks
- Data Management and Algorithms
- Advanced Bandit Algorithms Research
- Logic, Reasoning, and Knowledge
- Multi-Agent Systems and Negotiation
- Privacy-Preserving Technologies in Data
- Image Retrieval and Classification Techniques
- Adversarial Robustness in Machine Learning
- Business Process Modeling and Analysis
- Explainable Artificial Intelligence (XAI)
- Machine Learning in Healthcare
- Generative Adversarial Networks and Image Synthesis
- Context-Aware Activity Recognition Systems
- RFID technology advancements
- Mobile Agent-Based Network Management
- EEG and Brain-Computer Interfaces
- Renal Diseases and Glomerulopathies
- Constraint Satisfaction and Optimization
- Web Data Mining and Analysis
- Stochastic Gradient Optimization Techniques
Polytechnic University of Bari
2016-2025
Engineering (Italy)
2011-2023
Instituto Politécnico Nacional
2009-2021
RELX Group (Netherlands)
2021
RELX Group (United States)
2021
University of New Brunswick
2020
University of Cambridge
2020
Shenzhen Institutes of Advanced Technology
2020
University of Bristol
2020
Anderson University - Indiana
2020
The World Wide Web is moving from a of hyper-linked Documents to linked Data. Thanks the Semantic spread and more recent Linked Open Data (LOD) initiative, vast amount RDF data have been published in freely accessible datasets. These datasets are connected with each other form so called cloud. As today, there tons available Data, but only few applications really exploit their potential power. In this paper we show how these can successfully be used develop recommender system (RS) that relies...
Linked Open Data has been recognized as a valuable source for background information in many data mining and retrieval tasks. However, most of the existing tools require features propositional form, i.e., vector nominal or numerical associated with an instance, while Li nked sources are graphs by nature. In this paper, we present RDF2Vec, approach that uses language modeling approaches unsupervised feature extraction from sequences words, adapts them to RDF graphs. We generate leveraging...
This work takes a critical stance on previous studies concerning fairness evaluation in Large Language Model (LLM)-based recommender systems, which have primarily assessed consumer by comparing recommendation lists generated with and without sensitive user attributes. Such approaches implicitly treat discrepancies recommended items as biases, overlooking whether these changes might stem from genuine personalization aligned true preferences of users. Moreover, earlier typically address single...
The advent of the Linked Open Data (LOD) initiative gave birth to a variety open knowledge bases freely accessible on Web. They provide valuable source information that can improve conventional recommender systems, if properly exploited. In this paper we present SPrank, novel hybrid recommendation algorithm able compute top-N item recommendations from implicit feedback exploiting available in so called Web Data. We leverage DBpedia, well-known base LOD compass, extract semantic path-based...
The Web has moved, slowly but steadily, from a collection of documents towards structured data. Knowledge graphs have then emerged as way representing the knowledge encoded in such data well tool to reason on them order extract new and implicit information. are currently used, for example, explain search results, explore spaces, semantically enrich textual documents, or feed knowledge-intensive applications recommender systems. In this work, we describe how create exploit graph supply hybrid...
In most real-world scenarios, the ultimate goal of recommender system applications is to suggest a short ranked list items, namely top- N recommendations, that will appeal end user. Often, problem computing recommendations mainly tackled with two-step approach. The focuses first on predicting unknown ratings, which are eventually used generate recommendation list. Actually, task can be directly seen as ranking where main not accurately predict ratings but find best-ranked items recommend....
Recommender Systems have shown to be an effective way alleviate the over-choice problem and provide accurate tailored recommendations. However, impressive number of proposed recommendation algorithms, splitting strategies, evaluation protocols, metrics, tasks, has made rigorous experimental particularly challenging. Puzzled frustrated by continuous recreation appropriate benchmarks, pipelines, hyperparameter optimization, procedures, we developed exhaustive framework address such needs....
research-article Free Access Share on Recommender systems under European AI regulations Authors: Tommaso Di Noia Politecnico di Bari, Italy ItalyView Profile , Nava Tintarev TU Delft, The Netherlands NetherlandsView Panagiota Fatourou University of Crete, Greece GreeceView Markus Schedl Linz Institute Technology Lab LabView Authors Info & Claims Communications the ACMVolume 65Issue 4April 2022 pp 69–73https://doi.org/10.1145/3512728Online:19 March 2022Publication History...
Matchmaking arises when supply and demand meet in an electronic marketplace, or agents search for a web service to perform some task, even recruiting agencies match curricula job profiles. In such open environments, the objective of matchmaking process is discover best available offers given request. We address problem from knowledge representation perspective, with formalization based on Description Logics. devise Concept Abduction Contraction as non-monotonic inferences Logics suitable...
Providing very accurate recommendations to end users has been nowadays recognized be just one of the main tasks a recommender systems must able perform. While predicting relevant suggestions, attention needs paid their diversification in order avoid monotony recommendation. In this paper we focus on modeling users' inclination toward selecting diverse items, where diversity is computed by means content-based item attributes. We then exploit such present novel approach re-rank list Top-N...
The impact of data characteristics on the performance classical recommender systems has been recently investigated and produced fruitful results about relationship they have with recommendation accuracy. This work provides a systematic study broadly chosen (DCs) systems. is applied to accuracy fairness several variations CF models. We focus suite DCs that capture properties structure user–item interaction matrix, rating frequency, item properties, or distribution values. Experimental...
Point-of-interest (POI) recommendation is an essential service to location-based social networks (LBSNs), benefiting both users providing them the chance explore new locations and businesses by discovering potential customers. These systems learn preferences of their mobility patterns generate relevant POI recommendations. Previous studies have shown that incorporating contextual information such as geographical, temporal, social, categorical substantially improves quality However, fewer...
Research on recommender systems algorithms, like other areas of applied machine learning, is largely dominated by efforts to improve the state-of-the-art, typically in terms accuracy measures. Several recent research works however indicate that reported improvements over years sometimes "don't add up", and methods were published several ago often outperform latest models when evaluated independently. Different factors contribute this phenomenon, including some researchers probably only...
The availability of a huge amount interconnected data in the so called Web Data (WoD) paves way to new generation applications able exploit information encoded it. In this paper we present model-based recommender system leveraging datasets publicly available Linked Open (LOD) cloud as DBpedia and LinkedMDB. proposed approach adapts support vector machine (SVM) deal with RDF triples. We tested our showed its effectiveness by comparison different systems techniques -- both content-based...