Short CV
  • Academic positions
  • Education



Public engagement

Francesca A. Lisi's photo

Prof. Francesca A. Lisi

Professore Associato (ssd INF/01) / Associate Professor
Dipartimento di Informatica
Università degli Studi di Bari "Aldo Moro"

Campus Universitario "E. Quagliariello"
Via E. Orabona, 4 • I-70125 Bari • Italy

Phone: +39 080 5442296
Fax: +39 080 5442031
E-mail: francesca.lisi(a)
Skype: francesca.a.lisi

News & Highlights:

Short CV

Abilitazione a funzioni di Professore Associato (sc 01/B1: ASN 2018-2020, sc 09/H1: ASN 2013, ASN 2016-2018)

Academic positions

Other affiliations


[back to top]


AreasInterestsPublicationsInvited TalksAwardsGrantsProjects



I am active in the research field of Inductive Logic Programming (ILP) for 15 years. My current research is focused on:

Onto-Relational Learning

Since 2003, I have been working on frameworks for inductive inference which extend the methodological apparatus of ILP, originally conceived for learning in first-order Clausal Logics (CLs), to those hybrid Knowledge Representation (KR) systems that integrate CLs and Description Logics (DLs). In particular, I have considered two such KR systems: AL-log [Donini et al., IJIS 1998] and DL+log [Rosati, KR 2006]. The former integrates safely the DL ALC and the function-free CL fragment Datalog, whereas the latter is a weakly-safe integration of any DL and Disjunctive Datalog.

In the ILP framework for learning in AL-log [Lisi, TPLP 2008], hypotheses are represented as AL-log rules. The framework is general, meaning that it is valid whatever the scope of induction (prediction/description) is. Therefore the literal in the head of hypotheses represents a concept to be either discriminated from others or characterized. The generality relation for one such hypothesis language is an adaptation of the well-known ILP technique of generalized subsumption to the AL-log KR framework. It gives raise to a quasi-order and can be checked with a decidable procedure based on constrained SLD-resolution. Coverage relations for both ILP settings of learning from interpretations and learning from entailment have been defined on the basis of query answering in AL-log. The framework has been partially implemented in the ILP system SPADA that supports a variant of a very popular data mining task - Frequent Pattern Discovery - where rich prior conceptual knowledge is taken into account during the discovery process in order to find patterns at multiple levels of description granularity [Lisi & Malerba, MLJ 2004]. An evolution towards a variant of AL-log with a more expressive DL component (OWL DL) is implemented in the AL-QuIn system [Lisi, IJSWIS 2011].

The ILP framework for learning in DL+log [Lisi, TPLP 2010] represents hypotheses as DL+log rules and organizes them according to a generality ordering inspired by relative subsumption. Analogously to [Lisi, TPLP 2008], this framework encompasses both scopes of induction but, differently from (Lisi, 2008), it assumes the ILP setting of learning from entailment only. Both the coverage relation and the generality relation have been reformulated as satisfiability problems in DL+log. Compared to [Lisi, TPLP 2008], this framework shows an added value which can be summarized as follows. First, it can deal with incomplete information because disjunctive rules and default negation are allowed in DL+log . Second, it can induce rule-based definitions for new DL concepts because DL literals are allowed in DL+log rule heads.

Besides studying the theoretical aspects of these frameworks, I have been also investigating their applicability in Semantic Web Mining, Ontology Evolution, and Spatial Data Mining.

Learning in Fuzzy Description Logics

Since 2010, I have started a collaboration with Umberto Straccia on the extension of ILP to fuzzy DLs in order to deal with incomplete and vague knowledge. In particular, I have considered two KR formalisms: SoftFacts (Straccia, 2010) and fuzzy ALC(D) (Straccia, 2005). The former implements an ontology-based data access system where the ontology is represented in a fuzzy variant of DL-Lite, being therefore particularly suitable for data-intensive applications. The latter extends ALC with fuzzy concrete domains.

In [Lisi & Straccia, FI 2013], we describe a FOIL-like method, named SoftFoil, for learning fuzzy General Concept Inclusion (GCI) DL axioms from fuzzy examples. It builds upon SoftFacts and an information-theoretic heuristic adapted to account for multiple fuzzy instantiations of fuzzy predicates occurring in the induced GCIs.

Another FOIL-like method, called Foil-DL, is presented in [Lisi & Straccia, ILP 2013]. However, it adopts fuzzy ALC(D) as a KR framework and learns fuzzy GCI axioms from crisp examples. Also, it differs from SoftFoil as for the re finement operator and the heuristic. Finally, as opposite to SoftFoil, it is implemented and tested on toy problems (such as the Michalski trains) and real-world problems (such as hotel finding in the tourism application domain).

Declarative Modeling of Concept Learning in DLs

Since 2012, I have developed an interest in declarative modeling languages for Machine Learning and Data Mining. In particular, I have considered the case of Concept Learning in DLs [Lisi, ILP 2012]. Here, a fragment of Second-Order DLs is used as the starting point for the definition of a declarative language for modeling several variants of the Concept Learning problem (Concept Induction, Concept Refinement, and Concept Formation).

Invited Talks







[back to top]


Journal ArticlesBook ChaptersConference PapersEdited WorksAll

Institutional profile: IRIS
Other non-institutional profiles:  SCOPUS, Google Scholar, dblp

Selected from the production of the last 15 years.

Journal Articles

Book Chapters

Conference Papers

Edited Works

[back to top]

Professional service





PC member






[back to top]


BSc in Computer Science (Laurea di I livello in Informatica) - University of Bari

BSc in Computer Science and Software Production Technologies (Laurea di I livello in Informatica e Tecnologie per la Produzione del Software) - University of Bari

MSc in Computer Science (Laurea Magistrale in Informatica) - University of Bari

[back to top]

Public engagement

See also: my LinkedIn profile