Research

My research develops interpretable statistical learning methods for two main classes of data: natural language text and preference rankings. The thread connecting them is a preference for methods whose decisions can be explained and whose results can be audited.

Sentiment analysis from raw text

My doctoral and post-doctoral work has produced General Sentiment Decomposition — a framework for mining opinions directly from unprocessed natural language, without depending on heavy supervised pipelines. The line continues with two recent classifiers I co-designed: the Threshold-based Naïve Bayes classifier (Tb-NB) and its iterative refinement (iTb-NB), which combine competitive accuracy with the interpretability that classical Bayes models afford. Applications have ranged from tourism reviews on Booking.com to a sentiment study of Dante's Divina Commedia.

key references: Romano, Conversano (2025, MLwA); Romano, Zammarchi, Conversano (2023, SMA); Romano, Contu, Mola, Conversano (2023, ADAC); Romano (PhD thesis, 2021).

Preference learning and rank aggregation

A second strand of work concerns consensus ranking when many objects must be ordered from heterogeneous judgments. With co-authors I have proposed a heuristic algorithm scalable to large object sets, a Particle Swarm Optimization approach to preference rankings, and most recently a distance-based aggregation method for the more general case of preference-approvals.

key references: Albano, Romano (2026, ADAC); Romano, Conversano, Siciliano, D'Ambrosio (2025, ADAC); Romano, Siciliano (CLADAG 2023).

Semi-supervised methods on text

With Marco Ortu, Andrea Carta, Luca Frigau and others, I work on semi-supervised clustering that exploits both topical structure and sentiment to extract meaning from large collections of reviews and social media texts. Recent work also investigates green computing trade-offs in multiclass text classification — accuracy against computational and environmental cost.

key references: Ortu, Romano, Carta (2024, BDR); Frigau, Romano, Ortu, Contu (2023, SMA); Priola, Romano (2025, EJASA).

Funded projects (selected)

• e.INS — Ecosystem of Innovation for Next Generation Sardinia, Spoke 6 (PNRR, NextGenerationEU; CUP F53C22000430001). Member; statistical methods for digital transformation and sustainable mobility.
• Digital Education Hub Higher Education (PNRR M4C1, Investment 3.4; CUP D43C23004530005). Member.
• General Sentiment Decomposition (MGR — Mobilità Giovani Ricercatori, 2025). Scientific Coordinator; Tokyo Metropolitan University, Japan.
• Context AdapteR Tree-based Framework (MGR, 2023). Scientific Coordinator; Constructor University Bremen, Germany.
• ISIDE — Innovation for Sea Safety (Interreg Italy–France Maritime 2014–2020). Member; data mining and predictive modeling on sea accidents.

Collaborators

Claudio Conversano, Francesco Mola, Giulia Contu, Giulia Zammarchi, Luca Frigau, Marco Ortu (Cagliari); Antonio D'Ambrosio, Roberta Siciliano (Naples Federico II); Antonio Albano (Palermo); Valeria Vitelli (Oslo); Adalbert F. X. Wilhelm (Bremen); Atsuho Nakayama (Tokyo Metropolitan); Tomàs Aluja Banet (UPC Barcelona); José Luis García-Lapresta (Valladolid).