papers | Michele Stingo

2026

LLM-TEXT2KG
xch-MIND: Multi-Agent Interpretive Knowledge Graph Augmentation over Cultural Heritage Linked Data

Michele Stingo

In Joint Proceedings of the Fifth International Workshop on LLM-Integrated Knowledge Graph Generation From Text (TEXT2KG) and the Third International BiKE Challenge co-located with the Extended Semantic Web Conference (ESWC 2026), Dubrovnik, Croatia, 2026

CEUR Workshop Proceedings, Vol. 4233, pp. 65–83

Abs Bib HTML PDF

Institutional Linked Data repositories for cultural heritage, such as the Italian ArCo knowledge base, provide rich factual descriptions but lack interpretive layers connecting entities through spatial, temporal, and typological relationships. We present xch-MIND (eXtended Cultural Heritage — Multi-agent Interpretive Nexus Discovery), a multi-agent system that combines LLM-powered interpretive assertion generation with hybrid confidence scoring and non-invasive ontology layering over existing Linked Data. Four specialised agents analyse entities from ArCo, proposing clusters and relations that are validated through a hybrid score combining LLM self-assessment with deterministic domain-specific checks. All assertions are serialised as RDF triples in a purpose-built xch: ontology aligned with PROV-O, with an incremental multi-run strategy for progressive knowledge accumulation. Evaluated on 53 dolmen entities over six runs, xch-MIND discovers 175 unique interpretive assertions at 0.92 average confidence, serialised as 24,788 RDF triples with zero SHACL validation errors. Designed as a decision-support tool, the system provides full provenance and a tiered validation workflow that preserves expert authority over knowledge commitments.
@inproceedings{stingo2026xch, title = {xch-MIND: Multi-Agent Interpretive Knowledge Graph Augmentation over Cultural Heritage Linked Data}, author = {Stingo, Michele}, booktitle = {Joint Proceedings of the Fifth International Workshop on LLM-Integrated Knowledge Graph Generation From Text (TEXT2KG) and the Third International BiKE Challenge co-located with the Extended Semantic Web Conference (ESWC 2026)}, editor = {Tiwari, Sanju and Mihindukulasooriya, Nandana and D'Souza, Jennifer and Osborne, Francesco and Marx, Edgard}, series = {CEUR Workshop Proceedings}, volume = {4233}, pages = {65--83}, publisher = {CEUR-WS.org}, address = {Aachen, Germany}, issn = {1613-0073}, location = {Dubrovnik, Croatia}, urn = {urn:nbn:de:0074-4233-x}, note = {CEUR Workshop Proceedings, Vol. 4233, pp. 65–83}, year = {2026}, }

2020

INCOS
Mina: SeMantic vIrtual Assistant for Domain oNtology Based Question-Answering

Nicola Fiore, Gaetano Parente, Michele Stingo, and 1 more author

In International Conference on Intelligent Networking and Collaborative Systems, 2020

Abs Bib HTML

Aim of this article is to present the methodological nucleus of a custom Question-Answering system named Mina: a software solution with the ability to perform automated search over an ontological representation of localized knowledge domains, in order to extract fine-grained information so as to respond to user queries expressed as natural language questions. First the pipeline adopted by Mina will be presented, focusing on the multi-agent nature of the system Mina exists within, consisting of a distributed and dynamic environment on which several autonomous software (the so called agents) coexist and cooperate to perform specific tasks. Second we will focus on the strategies Mina exploits for the extraction of the correct answer which needs to be provided, starting with the introduction of the NC_Common_Lang library, a collection of linguistic analysis tools, adopted for the extraction of intents and entities contained within user queries; following up we’ll show the ontological representation of a specific knowledge sphere - the telecommunication domain in our case - describing the structured scheme of concepts and relationships according to which the answer retrieval operation is performed, querying the ontology through Description Logic inferences. As a final instance, three different test scenarios are presented with the intent to provide solid evidence about the correct functioning of Mina.
@inproceedings{fiore2020mina, title = {Mina: SeMantic vIrtual Assistant for Domain oNtology Based Question-Answering}, author = {Fiore, Nicola and Parente, Gaetano and Stingo, Michele and Polito, Massimiliano}, booktitle = {International Conference on Intelligent Networking and Collaborative Systems}, pages = {183--193}, year = {2020}, organization = {Springer International Publishing Cham}, }
AIDA_b
Assisted Contracts Validation through LCS problem resolution

Francesca Martinucci, Michele Stingo, and Massimiliano Polito

AIDAinformazioni, 2020

Abs Bib HTML

Verbal stipulation of contracts over the phone has become increasingly important in terms of sales volume for both b2b (business to business) and b2c (business to consumers) companies, especially when involved in main industrial branches such as those of telecommunication and energy. So, in order to be considered legally valid, a contract over the phone has to be spoken out correctly in its entirety — especially for those words concerning sensitive data — by the marketing outbound operator in charge with the stipulation. As a means of retrieving an esteem of re–liability for the stipulation outcomes and, in turn, as a useful and handy tool at disposal of the internal marketing analysts for the validation of concluded contracts, the present paper describes a software pipeline which exploits the resolution of the Longest Common Subsequence Problem in order to compare audio transcription of contracts stipulated over the phone against the certified written source of the same.
@article{martinucci2020assisted, title = {Assisted Contracts Validation through LCS problem resolution}, author = {Martinucci, Francesca and Stingo, Michele and Polito, Massimiliano}, journal = {AIDAinformazioni}, number = {3-4}, pages = {129--142}, year = {2020}, }

2019

WAINA
Automatic text preprocessing for intelligent dialog agents

Alessandro Maisto, Serena Pelosi, Massimiliano Polito, and 1 more author

In Workshops of the International Conference on Advanced Information Networking and Applications, 2019

Abs Bib HTML

The paper describes a new Text Preprocessing Pipeline based on a Hybrid approach which involve rule-based and stochastic approaches. The presented pipeline is part of a larger project titled Big Data for Multi-Agent Specialized System developed by Network Contacts in collaboration with University of Salerno and other institutional partners. The aim of the project is to build an Hybrid Question Answering System composed by sets of Dialog Bots able to process great volumes of data. Due to the importance of unstructured textual data, a particular focus of the project is on automatic processing of Text. The paper will describe the three main modules of the preprocessing pipeline, which involve a Style Correction Module, a Clitic Decomposition Module and a POS Tagging and Lemmatization Module.
@inproceedings{maisto2019automatic, title = {Automatic text preprocessing for intelligent dialog agents}, author = {Maisto, Alessandro and Pelosi, Serena and Polito, Massimiliano and Stingo, Michele}, booktitle = {Workshops of the International Conference on Advanced Information Networking and Applications}, pages = {805--814}, year = {2019}, organization = {Springer International Publishing Cham}, }
3PGCIC
An industrial multi-agent system (MAS) platform

Ariona Shashaj, Federico Mastrorilli, Michele Stingo, and 1 more author

In International Conference on P2P, Parallel, Grid, Cloud and Internet Computing, 2019

Abs Bib HTML

When it comes to address challenges in the area of distributed and parallel computing, Multi-Agent Systems (MAS) are emerging as a key architecture. Characterized as a collection of autonomous software (agents) which are able to cooperate in a distributed environment, MAS-based applications have proven capabilities when using cognitive processes, reasoning and knowledge representation in order to develop functionality related to complex and dynamic scenarios where the contribution of a single agent is computationally limited. In this paper, we propose an industrial platform which fully supports the development, deployment and maintenance cycle of MAS-based applications.
@inproceedings{shashaj2019industrial, title = {An industrial multi-agent system (MAS) platform}, author = {Shashaj, Ariona and Mastrorilli, Federico and Stingo, Michele and Polito, Massimiliano}, booktitle = {International Conference on P2P, Parallel, Grid, Cloud and Internet Computing}, pages = {221--233}, year = {2019}, organization = {Springer International Publishing Cham}, }

2018

AIDA_a
OPI: Ontologia dei prodotti italiani con denominazione DOP, IGP, STG.

Michele Stingo

In XXVIII° Convegno I nternazionale Ass.I.Term “Terminologie e risorse lingui stiche per comunicare in Europa ” Università degli Studi di Salerno , Fisciano - 17 e 18 maggio 2018 Città di Salerno - 19 maggio 2018, 2018

Abs Bib HTML

The paper describes the research efforts carried out for the implementation of OPI, the Ontology of Italian Protected names. The attempt has been made to provide a resource — on the fifth level of the deployment scheme for Open Data — for the ontological representation of the protected names domain with focus on the normative dimension which regulates the designation of European quality logos. The initial section frames the protected names domain, while a state of the art description of ontologies and database on the same domain is described throughout the second part. Third chapter comes with the complete description of classes and properties modelling OPI, plus a consistency test and description logic queries submitted to the ontology. The fourth and final section gives account of both necessary and possible future activities to ameliorate the resource.
@inproceedings{stingo2018opi, title = {OPI: Ontologia dei prodotti italiani con denominazione DOP, IGP, STG.}, author = {Stingo, Michele}, booktitle = {XXVIII° Convegno I nternazionale Ass.I.Term “Terminologie e risorse lingui stiche per comunicare in Europa ” Universit{\`a} degli Studi di Salerno , Fisciano - 17 e 18 maggio 2018 Citt{\`a} di Salerno - 19 maggio 2018}, year = {2018}, }
CLIC-it
Multi-Word Expressions in spoken language: PoliSdict

Daniela Trotta, Teresa Albanese, Michele Stingo, and 2 more authors

In CLiC-it 2018 Fifth Italian Conference on Computational Linguistics, Torino, December 10-12 , 2018, 2018

Abs Bib HTML

The term multiword expressions (MWEs) is referred-to a group of words with a unitary meaning, not inferred from that of the words that compose it, both in current use and in technical-specialized languages. In this paper, we describe PoliSdict an Italian electronic dictionary composed of multi-word expressions (MWEs) automatically extracted from a multimodal corpus grounded on political speech language. We introduce the methodology of creation and the first results of a systematic analysis which considered terminological labels, frequency labels, recurring syntactic patterns, further proposing an associated ontology.
@inproceedings{trotta2018multi, title = {Multi-Word Expressions in spoken language: PoliSdict}, author = {Trotta, Daniela and Albanese, Teresa and Stingo, Michele and Guarasci, Raffaele and Elia, Annibale}, booktitle = {CLiC-it 2018 Fifth Italian Conference on Computational Linguistics, Torino, December 10-12 , 2018}, year = {2018}, }

2017

ESE
A Hybrid Method for the Extraction and Classification of Product Features from User Generated Contents.

Alessandro Maisto, Serena Pelosi, Michele Stingo, and 1 more author

Lingue e Linguaggi, 2017

Abs Bib HTML

The research we present in this paper focuses on the automatic management of the knowledge about experience goods and services and their features, starting from real texts generated online by internet users. The details about an experiment conducted on a dataset of product reviews, on which we tested a set of rule-based and statistical solutions, will be described in the paper. The main goals are the review classification, the extraction of relevant product features and their systematization into product-driven ontologies. Feature extraction is performed through a rule-based strategy grounded on SentIta, an Italian collection of subjective lexical resources. Features and Reviews are classified thanks to a Distributional Semantic algorithm. In the end, we face the problem of the extracted knowledge organization by integrating the subjective information produced by the internet users within a product-driven ontology. The Natural Language Processing (NLP) tool exploited in the work is LG-Starship, a hybrid framework for Italian texts processing based on the Lexicon-Grammar theory.
@article{maisto2017hybrid, title = {A Hybrid Method for the Extraction and Classification of Product Features from User Generated Contents.}, author = {Maisto, Alessandro and Pelosi, Serena and Stingo, Michele and Guarasci, Raffaele}, journal = {Lingue e Linguaggi}, volume = {22}, year = {2017}, }

2016

ICCI
Detecting satire in italian political commentaries

Rodolfo Delmonte and Michele Stingo

In International Conference on Computational Collective Intelligence, 2016

Abs Bib HTML

This paper presents computational work to detect satire/sarcasm in long commentaries on Italian politics. It uses the lexica extracted from the manual annotation based on Appraisal Theory, of some 30 K word texts. The underlying hypothesis is that using this framework it is possible to precisely pinpoint ironic content through the deep semantic analysis of evaluative judgement and appreciation. The paper presents the manual annotation phase realized on 112 texts by two well-known Italian journalists. After a first experimentation phase based on the lexica extracted from the xml output files, we proceeded to retag lexical entries dividing them up into two subclasses: figurative and literal meaning. Finally more fine-grained Appraisal features have been derived and more experiments have been carried out and compared to results obtained by a lean sentiment analysis. The final output is produced from held out texts to verify the usefulness of the lexica and the Appraisal theory in detecting ironic content.
@inproceedings{delmonte2016detecting, title = {Detecting satire in italian political commentaries}, author = {Delmonte, Rodolfo and Stingo, Michele}, booktitle = {International Conference on Computational Collective Intelligence}, pages = {68--77}, year = {2016}, organization = {Springer International Publishing Cham}, }
NLPJ
Annotating satire in italian political commentaries with appraisal theory

Michele Stingo, Rodolfo Delmonte, and others

In NLPMJ-Natural Language Processing meets Journalism, 2016

Abs Bib HTML

This paper presents annotation work to manually classify satire/sarcasm in long commentaries on Italian politics. It is based on Appraisal Theory and uses some 30K word texts. The underlying hypothesis is that using this framework it is possible to pinpoint precisely the deep semantic contents of evaluative judgements and appreciations making up an ironic comment. We performed a high level annotation using major categories, and then refined the classification starting from the lexica derived from the xml annotated files. In this way we succeeded in differentiating texts by the two authors we chose, one of which is characterized by a sharp cutting ironic/almost sarcastic style.
@incollection{stingo2016annotating, title = {Annotating satire in italian political commentaries with appraisal theory}, author = {Stingo, Michele and Delmonte, Rodolfo and others}, booktitle = {NLPMJ-Natural Language Processing meets Journalism}, pages = {74--79}, year = {2016}, publisher = {AAAI Press/International Joint Conferences on Artificial Intelligence}, }