arXiv subjects

Francesco Poggi

Publications and source records attributed to Francesco Poggi.

13 recordsLinked to original sources

arXiv2026

RAGEAR: Retrieval-Augmented Graph-Enhanced Academic Recommender

We present RAGEAR (Retrieval-Augmented Graph-Enhanced Academic Recommender), a neurosymbolic recommender system for academic course recommendation. RAGEAR combines dense retrieval over full lecture transcripts with a symbolic Knowledge Graph modelling courses, lessons, transcript chunks, credits, study plans, and curricular information. The Knowledge Graph supports symbolic filtering and contextualisation based on structured constraints, such as credits, academic disciplines, study plans, and prerequisites. Unlike metadata-based approaches, it exploits fine-grained instructional content by retrieving transcript chunks semantically aligned with a student's query. The main contribution is a graph-aware aggregation function that propagates chunk-level evidence to course-level recommendations. The score combines three factors: the share of retrieved similarity associated with a course, the rank-based strength of its relevant chunks, and the distribution of evidence across lessons. We evaluate RAGEAR on 152 student-like queries through a human evaluation sample and a large-scale LLM-based relevance assessment. Results show that lecture transcripts improve over metadata-only retrieval, and that RAGEAR further improves ranking quality over a transcript-based normalized SumP baseline, especially for top-ranked recommendations.

cs.IR↗

arXiv2026

Tacit Knowledge Extraction via Logic Augmented Generation and Active Inference

Tacit knowledge plays a central role in human expertise, yet it remains difficult to capture, formalize, and reuse in machine-interpretable form. This challenge is especially relevant in procedural domains, where successful execution depends not only on explicit instructions, but also on implicit assumptions, contextual constraints, embodied skills, and experience-based judgments rarely documented. As a result, current knowledge engineering pipelines struggle to transform tacit and process-centric knowledge into formally specified, machine-interpretable representations that can be queried, validated, reasoned over, and reused. In this paper, we introduce a neuro-symbolic framework that combines Logic-Augmented Generation and an Active-Inference-inspired approach for ontology-grounded Knowledge Graph construction. We evaluate the approach in a knowledge transfer case study in manufacturing, using assembly-like repair procedures from instructional videos as a reproducible proxy domain. Results show that the proposed solution improves completeness and semantic quality, advancing neuro-symbolic knowledge engineering for industrial domains.

cs.AI↗

arXiv2025

Enhancing Retrieval-Augmented Generation with Entity Linking for Educational Platforms

In the era of Large Language Models (LLMs), Retrieval-Augmented Generation (RAG) architectures are gaining significant attention for their ability to ground language generation in reliable knowledge sources. Despite their effectiveness, RAG systems based solely on semantic similarity often fail to ensure factual accuracy in specialized domains, where terminological ambiguity can affect retrieval relevance. This study proposes ELERAG, an enhanced RAG architecture that integrates a factual signal derived from Entity Linking to improve the accuracy of educational question-answering systems in Italian. The system includes a Wikidata-based Entity Linking module and implements a hybrid re-ranking strategy based on Reciprocal Rank Fusion (RRF). To validate our approach, we compared it against standard baselines and state-of-the-art methods, including a Weighted-Score Re-ranking, a standalone Cross-Encoder and a combined RRF+Cross-Encoder pipeline. Experiments were conducted on two benchmarks: a custom academic dataset and the standard SQuAD-it dataset. Results show that, in domain-specific contexts, ELERAG significantly outperforms both the baseline and the Cross-Encoder configurations. Conversely, the Cross-Encoder approaches achieve the best results on the general-domain dataset. These findings provide strong experimental evidence of the domain mismatch effect, highlighting the importance of domain-adapted hybrid strategies to enhance factual precision in educational RAG systems without relying on computationally expensive models trained on disparate data distributions. They also demonstrate the potential of entity-aware RAG systems in educational environments, fostering adaptive and reliable AI-based tutoring tools.

cs.IR↗

arXiv2025

The Belief-Desire-Intention Ontology for modelling mental reality and agency

The Belief-Desire-Intention (BDI) model is a cornerstone for representing rational agency in artificial intelligence and cognitive sciences. Yet, its integration into structured, semantically interoperable knowledge representations remains limited. This paper presents a formal BDI Ontology, conceived as a modular Ontology Design Pattern (ODP) that captures the cognitive architecture of agents through beliefs, desires, intentions, and their dynamic interrelations. The ontology ensures semantic precision and reusability by aligning with foundational ontologies and best practices in modular design. Two complementary lines of experimentation demonstrate its applicability: (i) coupling the ontology with Large Language Models (LLMs) via Logic Augmented Generation (LAG) to assess the contribution of ontological grounding to inferential coherence and consistency; and (ii) integrating the ontology within the Semas reasoning platform, which implements the Triples-to-Beliefs-to-Triples (T2B2T) paradigm, enabling a bidirectional flow between RDF triples and agent mental states. Together, these experiments illustrate how the BDI Ontology acts as both a conceptual and operational bridge between declarative and procedural intelligence, paving the way for cognitively grounded, explainable, and semantically interoperable multi-agent and neuro-symbolic systems operating within the Web of Data.

cs.AI↗

arXiv2025

LEAD: LLM-enhanced Engine for Author Disambiguation

Author Name Disambiguation (AND) is a long-standing challenge in bibliometrics and scientometrics, as name ambiguity undermines the accuracy of bibliographic databases and the reliability of research evaluation. This study addresses the problem of cross-source disambiguation by linking academic career records from CercaUniversit\`a, the official registry of Italian academics, with author profiles in Scopus. We introduce LEAD (LLM-enhanced Engine for Author Disambiguation), a novel hybrid framework that combines semantic features extracted through Large Language Models (LLMs) with structural evidence derived from co-authorship and citation networks. Using a gold standard of 606 ambiguous cases, we compare five methods: (i) Label Spreading on co-authorship networks; (ii) Bibliographic Coupling on citation networks; (iii) a standalone LLM-based approach; (iv) an LLM-enriched configuration; and (v) the proposed hybrid pipeline. LEAD achieves the best performance (F1 = 96.7%, accuracy = 95.7%) with lower computational cost than full LLM models. Bibliographic Coupling emerges as the fastest and strongest single-source method. These findings demonstrate that integrating semantic and structural signals within a selective hybrid strategy offers a robust and scalable solution to cross-database author identification. Beyond the Italian case, this work highlights the potential of hybrid LLM-based methods to improve data quality and reliability in scientometric analyses.

cs.DL↗

arXiv2022

Do open citations give insights on the qualitative peer-review evaluation in research assessments? An analysis of the Italian National Scientific Qualification

In the past, several works have investigated ways for combining quantitative and qualitative methods in research assessment exercises. Indeed, the Italian National Scientific Qualification (NSQ), i.e. the national assessment exercise which aims at deciding whether a scholar can apply to professorial academic positions as Associate Professor and Full Professor, adopts a quantitative and qualitative evaluation process: it makes use of bibliometrics followed by a peer-review process of candidates' CVs. The NSQ divides academic disciplines into two categories, i.e. citation-based disciplines (CDs) and non-citation-based disciplines (NDs), a division that affects the metrics used for assessing the candidates of that discipline in the first part of the process, which is based on bibliometrics. In this work, we aim at exploring whether citation-based metrics, calculated only considering open bibliographic and citation data, can support the human peer-review of NDs and yield insights on how it is conducted. To understand if and what citation-based (and, possibly, other) metrics provide relevant information, we created a series of machine learning models to replicate the decisions of the NSQ committees. As one of the main outcomes of our study, we noticed that the strength of the citational relationship between the candidate and the commission in charge of assessing his/her CV seems to play a role in the peer-review phase of the NSQ of NDs.

cs.DL↗

arXiv2022

Open bibliographic data and the Italian National Scientific Qualification: measuring coverage of academic fields

The importance of open bibliographic repositories is widely accepted by the scientific community. For evaluation processes, however, there is still some skepticism: even if large repositories of open access articles and free publication indexes exist and are continuously growing, assessment procedures still rely on proprietary databases, mainly due to the richness of the data available in these proprietary databases and the services provided by the companies they are offered by. This paper investigates the status of open bibliographic data of three of the most used open resources, namely Microsoft Academic Graph, Crossref and OpenAIRE, evaluating their potentialities as substitutes of proprietary databases for academic evaluation processes. We focused on the Italian National Scientific Qualification (NSQ), the Italian process for University Professor qualification, which uses data from commercial indexes, and investigated similarities and differences between research areas, disciplines and application roles. The main conclusion is that open datasets are ready to be used for some disciplines, among which mathematics, natural sciences, economics and statistics, even if there is still room for improvement; but there is still a large gap to fill in others - like history, philosophy, pedagogy and psychology - and a stronger effort is required from researchers and institutions.

cs.DL↗

arXiv2021

Academics evaluating academics: a methodology to inform the review process on top of open citations

In the past, several works have investigated ways for combining quantitative and qualitative methods in research assessment exercises. In this work, we aim at introducing a methodology to explore whether citation-based metrics, calculated only considering open bibliographic and citation data, can yield insights on how human peer-review of research assessment exercises is conducted. To understand if and what metrics provide relevant information, we propose to use a series of machine learning models to replicate the decisions of the committees of the research assessment exercises.

cs.DL↗

arXiv2021

Does the Venue of Scientific Conferences Leverage their Impact? A Large Scale study on Computer Science Conferences

Background: Conferences bring scientists together and provide one of the most timely means for disseminating new ideas and cutting-edge works.The importance of conferences in scientific areas is testified by quantitative indicators. In Computer Science, for instance, almost two out of three papers published on Scopus are conference papers. Objective/Purpose: The main goal of this paper is to investigate a novel research question: is there any correlation between the impact of a scientific conference and the venue where it took place? Approach: In order to measure the impact of conferences we conducted a large scale analysis on the bibliographic data extracted from 3,838 Computer Science conference series and over 2.5 million papers spanning more than 30 years of research. To quantify the "touristicity" of a venue we exploited some indicators such as the size of the Wikipedia page for the city hosting the venue and other indexes from reports of the World Economic Forum. Results/Findings: We found out that the two aspects are related, and the correlation with conference impact is stronger when considering country-wide touristic indicators, such as the Travel&Tourism Competitiveness Index. More-over the almost linear correlation with the Tourist Service Infrastructure index attests the specific importance of tourist/accommodation facilities in a given country. Conclusions: This is the first attempt to focus on the relationship of venue characteristics to conference papers. The results open up new possibilities, such as allowing conference organizers and authors to estimate in advance the impact of conferences, thus supporting them in their decisions.

cs.DL↗

arXiv2021

Can we assess research using open scientific knowledge graphs? A case study within the Italian National Scientific Qualification

The need for open scientific knowledge graphs is ever increasing. While there are large repositories of open access articles and free publication indexes, there are still few free knowledge graphs exposing citation networks, and often their coverage is partial. Consequently, most evaluation processes based on citation counts rely on commercial citation databases. Things are changing thanks to the Initiative for Open Citations (I4OC, https://i4oc.org) and the Initiative for Open Abstracts (I4OA, https://i4oa.org), whose goal is to campaign for scholarly publishers to open the reference lists and the other metadata of their articles. This paper investigates the growth of the open bibliographic metadata and open citations in two scientific knowledge graphs, OpenCitations' COCI and Crossref, with an experiment on the Italian National Scientific Qualification (NSQ), the National process for University Professor qualification which uses data from commercial indexes. We simulated the procedure by only using such open data and explored similarities and differences with the official results. The outcomes of the experiment show that the amount of open bibliographic metadata and open citation data currently available in the two scientific knowledge graphs adopted is not yet enough for obtaining results similar to those provided using commercial databases.

cs.DL↗

arXiv2020

The practice of self-citations: a longitudinal study

In this article, we discuss the outcomes of an experiment where we analysed whether and to what extent the introduction, in 2012, of the new research assessment exercise in Italy (a.k.a. Italian Scientific Habilitation) affected self-citation behaviours in the Italian research community. The Italian Scientific Habilitation attests to the scientific maturity of researchers and in Italy, as in many other countries, is a requirement for accessing to a professorship. To this end, we obtained from ScienceDirect 35,673 articles published from 1957 and 2016 by the participants to the 2012 Italian Scientific Habilitation, that resulted in the extraction of 1,379,050 citations retrieved through Semantic Publishing technologies. Our analysis showed an overall increment in author self-citations (i.e. where the citing article and the cited article share at least one author) in several of the 24 academic disciplines considered. However, we depicted a stronger causal relation between such increment and the rules introduced by the 2012 Italian Scientific Habilitation in 10 out of 24 disciplines analysed.

cs.DL↗

arXiv2019

Open data to evaluate academic researchers: an experiment with the Italian Scientific Habilitation

The need for scholarly open data is ever increasing. While there are large repositories of open access articles and free publication indexes, there are still a few examples of free citation networks and their coverage is partial. One of the results is that most of the evaluation processes based on citation counts rely on commercial citation databases. Things are changing under the pressure of the Initiative for Open Citations (I4OC), whose goal is to campaign for scholarly publishers to make their citations as totally open. This paper investigates the growth of open citations with an experiment on the Italian Scientific Habilitation, the National process for University Professor qualification which instead uses data from commercial indexes. We simulated the procedure by only using open data and explored similarities and differences with the official results. The outcomes of the experiment show that the amount of open citation data currently available is not yet enough for obtaining similar results.

cs.DL↗

arXiv2018

Do altmetrics work for assessing research quality?

Alternative metrics (aka altmetrics) are gaining increasing interest in the scientometrics community as they can capture both the volume and quality of attention that a research work receives online. Nevertheless, there is limited knowledge about their effectiveness as a mean for measuring the impact of research if compared to traditional citation-based indicators. This work aims at rigorously investigating if any correlation exists among indicators, either traditional (i.e. citation count and h-index) or alternative (i.e. altmetrics) and which of them may be effective for evaluating scholars. The study is based on the analysis of real data coming from the National Scientific Qualification procedure held in Italy by committees of peers on behalf of the Italian Ministry of Education, Universities and Research.

cs.DL↗