Elisa Di Nuovo
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Elisa Di Nuovo has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Bertrand De Longueville
European Commission, Joint Research Centre
2 shared publications - Emmanuel Cartier
European Commission, Joint Research Centre
2 shared publications - Alessandro Mazzei
University of Turin
1 shared publication Cristian Bernareggi
Dipartimento di Matematica, Laboratorio Polin, Universit degli Studi di Torino, Turin, Italy
1 shared publication- Ilenia Tinnirello
Department of Engineering, University of Palermo, Palermo, Italy
1 shared publication - Luca Anselma
Università degli Studi di Torino
1 shared publication
Latest Publications
- Open Access
- 25 February 2025
- Published by Springer Nature
- p. C1-C2
Correction to: Meet XLM-RLnews-8: Not Just Another Sentiment Analysis Model
- 25 February 2025
- Published by Springer Nature
- p. C1-C2
Open AccessCorrection to: Meet XLM-RLnews-8: Not Just Another Sentiment Analysis Model
- Open Access
- 14 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 150502-150519
An Educational Dialogue System for Visually Impaired People
Finite State Automata (FSA) serve as a fundamental mathematical model of computation, commonly explored in computer science degree programs. Traditionally, FSA are represented through state diagrams or tables, which, in educational settings necessitating assistive technologies, pose accessibility challenges, particularly for Visually Impaired People (VIP). This work is part of a broader initiative focused on creating tools for inclusive access to scientific information. Its main goal is to explore the effectiveness of a dialogue system (DS) as an alternative method for conveying FSA information to VIP, aiming to enhance their comprehension and user experience. To achieve this, a rule-based DS tailored for facilitating FSA access, with a primary focus on VIP as end users, was developed. The research involved an A/B test comparing participants’ comprehension of FSA using the rule-based DS versus standard tabular representations. Statistical analysis was also carried out to evaluate the performance differences between the two methods. The findings indicate that communication through the DS significantly outperforms the tabular representation, establishing it as a viable and effective alternative for both VIP and non-VIP. Although VIP participants displayed slightly varying performance depending on the questions, their feedback favoured the DS for its ease of use and overall user satisfaction. Additionally, the study analysed the capabilities and limitations of popular Large Language Models (LLMs) in describing graphical structures like FSA. Despite their general effectiveness in language tasks, LLMs proved inadequate for accurately and consistently describing FSA, highlighting the need for more controlled and explainable DS approaches in educational contexts.
- 14 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 150502-150519
Open AccessAn Educational Dialogue System for Visually Impaired People
Finite State Automata (FSA) serve as a fundamental mathematical model of computation, commonly explored in computer science degree programs. Traditionally, FSA are represented through state diagrams or tables, which, in educational settings necessitating assistive technologies, pose accessibility challenges, particularly for Visually Impaired People (VIP). This work is part of a broader initiative focused on creating tools for inclusive access to scientific information. Its main goal is to explore the effectiveness of a dialogue system (DS) as an alternative method for conveying FSA information to VIP, aiming to enhance their comprehension and user experience. To achieve this, a rule-based DS tailored for facilitating FSA access, with a primary focus on VIP as end users, was developed. The research involved an A/B test comparing participants’ comprehension of FSA using the rule-based DS versus standard tabular representations. Statistical analysis was also carried out to evaluate the performance differences between the two methods. The findings indicate that communication through the DS significantly outperforms the tabular representation, establishing it as a viable and effective alternative for both VIP and non-VIP. Although VIP participants displayed slightly varying performance depending on the questions, their feedback favoured the DS for its ease of use and overall user satisfaction. Additionally, the study analysed the capabilities and limitations of popular Large Language Models (LLMs) in describing graphical structures like FSA. Despite their general effectiveness in language tasks, LLMs proved inadequate for accurately and consistently describing FSA, highlighting the need for more controlled and explainable DS approaches in educational contexts.
- 20 September 2024
- Published by Springer Nature
- p. 24-35
Meet XLM-RLnews-8: Not Just Another Sentiment Analysis Model
Although automatic sentiment analysis has been widely studied in the past decade, multilingualism remains an issue impairing real-life applications. This paper describes the development and evaluation of XLM-RLnews-8, a model based on XLM-RoBERTa-Large, domain-adapted using a novel dataset of multilingual news articles and fine-tuned for the tripartite sentiment analysis task. In addition, it provides a quantitative analysis of the Unified Multilingual Sentiment Analysis Benchmark, the dataset used for the fine-tuning and the in-domain evaluation. The model is also out-of-domain evaluated, on the IMDb dataset and a new multilingual news headlines silver dataset, and its performance is compared with current State-of-the-Art multilingual models. Models, notebooks and datasets developed for this publication are available at https://github.com/ElisaDiNuovo/XLM-RLnews-8/.
- 20 September 2024
- Published by Springer Nature
- p. 24-35
Meet XLM-RLnews-8: Not Just Another Sentiment Analysis Model
Although automatic sentiment analysis has been widely studied in the past decade, multilingualism remains an issue impairing real-life applications. This paper describes the development and evaluation of XLM-RLnews-8, a model based on XLM-RoBERTa-Large, domain-adapted using a novel dataset of multilingual news articles and fine-tuned for the tripartite sentiment analysis task. In addition, it provides a quantitative analysis of the Unified Multilingual Sentiment Analysis Benchmark, the dataset used for the fine-tuning and the in-domain evaluation. The model is also out-of-domain evaluated, on the IMDb dataset and a new multilingual news headlines silver dataset, and its performance is compared with current State-of-the-Art multilingual models. Models, notebooks and datasets developed for this publication are available at https://github.com/ElisaDiNuovo/XLM-RLnews-8/.
- Open Access
- 9 September 2022
- Published by MDPI AG in Information
- Vol. 13 (9) , 426
Fake News Spreaders Detection: Sometimes Attention Is Not All You Need
Guided by a corpus linguistics approach, in this article we present a comparative evaluation of State-of-the-Art (SotA) models, with a special focus on Transformers, to address the task of Fake News Spreaders (i.e., users that share Fake News) detection. First, we explore the reference multilingual dataset for the considered task, exploiting corpus linguistics techniques, such as chi-square test, keywords and Word Sketch. Second, we perform experiments on several models for Natural Language Processing. Third, we perform a comparative evaluation using the most recent Transformer-based models (RoBERTa, DistilBERT, BERT, XLNet, ELECTRA, Longformer) and other deep and non-deep SotA models (CNN, MultiCNN, Bayes, SVM). The CNN tested outperforms all the models tested and, to the best of our knowledge, any existing approach on the same dataset. Fourth, to better understand this result, we conduct a post-hoc analysis as an attempt to investigate the behaviour of the presented best performing black-box model. This study highlights the importance of choosing a suitable classifier given the specific task. To make an educated decision, we propose the use of corpus linguistics techniques. Our results suggest that large pre-trained deep models like Transformers are not necessarily the first choice when addressing a text classification task as the one presented in this article. All the code developed to run our tests is publicly available on GitHub.
- 9 September 2022
- Published by MDPI AG in Information
- Vol. 13 (9) , 426
Open AccessFake News Spreaders Detection: Sometimes Attention Is Not All You Need
Guided by a corpus linguistics approach, in this article we present a comparative evaluation of State-of-the-Art (SotA) models, with a special focus on Transformers, to address the task of Fake News Spreaders (i.e., users that share Fake News) detection. First, we explore the reference multilingual dataset for the considered task, exploiting corpus linguistics techniques, such as chi-square test, keywords and Word Sketch. Second, we perform experiments on several models for Natural Language Processing. Third, we perform a comparative evaluation using the most recent Transformer-based models (RoBERTa, DistilBERT, BERT, XLNet, ELECTRA, Longformer) and other deep and non-deep SotA models (CNN, MultiCNN, Bayes, SVM). The CNN tested outperforms all the models tested and, to the best of our knowledge, any existing approach on the same dataset. Fourth, to better understand this result, we conduct a post-hoc analysis as an attempt to investigate the behaviour of the presented best performing black-box model. This study highlights the importance of choosing a suitable classifier given the specific task. To make an educated decision, we propose the use of corpus linguistics techniques. Our results suggest that large pre-trained deep models like Transformers are not necessarily the first choice when addressing a text classification task as the one presented in this article. All the code developed to run our tests is publicly available on GitHub.