A5026350868- Topic Modeling
- Natural Language Processing Techniques
- Hate Speech and Cyberbullying Detection
- Text Readability and Simplification
- Bullying, Victimization, and Aggression
- Internet Traffic Analysis and Secure E-voting
- Advanced Text Analysis Techniques
- Social Media and Politics
- Sentiment Analysis and Opinion Mining
- Olfactory and Sensory Function Studies
- Advanced Chemical Sensor Technologies
- Computational and Text Analysis Methods
- Spam and Phishing Detection
- Wikis in Education and Collaboration
- Digital Humanities and Scholarship
- Authorship Attribution and Profiling
- Misinformation and Its Impacts
- Artificial Intelligence in Healthcare
- Adversarial Robustness in Machine Learning
- Semantic Web and Ontologies
- Online Learning and Analytics
- Swearing, Euphemism, Multilingualism
- Geochemistry and Geologic Mapping
- Library Science and Information Systems
- Intelligent Tutoring Systems and Adaptive Learning
Fondazione Bruno Kessler
2014-2024
Kessler Foundation
2021
University of Trento
2014-2018
This paper presents the task on evaluation of Compositional Distributional Semantics Models full sentences organized for first time within SemEval-2014.Participation was open to systems based any approach.Systems were presented with pairs and evaluated their ability predict human judgments (i) semantic relatedness (ii) entailment.The attracted 21 teams, most which participated in both subtasks.We received 17 submissions subtask (for a total 66 runs) 18 entailment (65 runs).
The increasing popularity of social media platforms such as Twitter and Facebook has led to a rise in the presence hate aggressive speech on these platforms. Despite number approaches recently proposed Natural Language Processing research area for detecting forms abusive language, issue identifying at scale is still an unsolved problem. In this article, we propose robust neural architecture that shown perform satisfactory way across different languages; namely, English, Italian, German. We...
Abstract Massive open online courses (MOOCs) provide hundreds of students with teaching materials, assessment tools, and collaborative instruments. The activity, in particular, is demanding terms both time effort; thus, the use artificial intelligence can be useful to address reduce effort required. This paper reports on a system related experiments finalised improve performance quality formative summative assessments specific data science courses. developed automatically grade assignments...
Since state-of-the-art approaches to offensive language detection rely on supervised learning, it is crucial quickly adapt them the continuously evolving scenario of social media. While several have been proposed tackle problem from an algorithmic perspective, so reduce need for annotated data, less attention has paid quality these data. Following a trend that emerged recently, we focus level agreement among annotators while selecting data create datasets, task involving high subjectivity....
Although WhatsApp is used by teenagers as one major channel of cyberbullying, such interactions remain invisible due to the app privacy policies that do not allow ex-post data collection. Indeed, most information on these phenomena rely surveys regarding self-reported data. In order overcome this limitation, we describe in paper activities led creation a dataset study cyberbullying among Italian students aged 12-13. We present only collected chats with annotations about user role and type...
In this work, we apply argumentation mining techniques, in particular relation prediction, to study political speeches monological form, where there is no direct interaction between opponents. We argue that kind of technique can effectively support researchers history, social and sciences, which must deal with an increasing amount data digital form need ways automatically extract analyse patterns. test discuss our approach based on the analysis documents issued by R. Nixon J. F. Kennedy...
Social media platforms like Twitter and Instagram face a surge in cyberbullying phenomena against young users need to develop scalable computational methods limit the negative consequences of this kind abuse. Despite number approaches recently proposed Natural Language Processing (NLP) research area for detecting different forms abusive language, issue identifying at scale is still an unsolved problem. This because couple language detection on textual message with network analysis, so that...
The datasets most widely used for abusive language detection contain lists of messages, usually tweets, that have been manually judged as or not by one more annotators, with the annotation performed at message level. In this paper, we investigate what happens when hateful content a is also based on context, given messages are often ambiguous and need to be interpreted in context occurrence. We first re-annotate part dataset English two conditions, i.e. without context. Then, compare...
Abstract This study aims to advance the discussion around linkages between public funding for scientific research and new technology development. We develop test a methodology that matches patents publications, latter of which stemming from publicly funded research. specifically focus on projects were by European Research Council (ERC) in Life Sciences Physical Science Engineering sectors during FP7 Programme. also compare this method’s results with directly reported PIs ERC-funded at end...
We present a topic-based analysis of agreement and disagreement in political manifestos, which relies on new method for topic detection based key concept clustering. Our approach outperforms both standard techniques like LDA state-of-the-art graph-based method, provides promising initial results this task computational social science.
Stefano Menini, Rachele Sprugnoli, Giovanni Moretti, Enrico Bignotti, Sara Tonelli, Bruno Lepri. Proceedings of the Software Demonstrations 15th Conference European Chapter Association for Computational Linguistics. 2017.
Recent studies have demonstrated the effectiveness of cross-lingual language model pre-training on different NLP tasks, such as natural inference and machine translation. In our work, we test this approach social media data, which are particularly challenging to process within framework, since limited length textual messages irregularity make it harder learn meaningful encodings. More specifically, propose a hybrid emoji-based Masked Language Model (MLM) leverage common information conveyed...
In this paper we present our submission to sub-task A at SemEval 2020 Task 12: Multilingual Offensive Language Identification in Social Media (OffensEval2). For Danish, Turkish, Arabic and Greek, develop an architecture based on transfer learning relying a two-channel BERT model, which the English multilingual one are combined after creating machine-translated parallel corpus for each language task. English, instead, adopt more standard, single-channel approach. We find that, scenario, with...
Abstract The same individuals can express very different emotions in online social media with respect to face-to-face interactions, partially because of intrinsic limitations the digital environments and their algorithmic design, which is optimized maximize engagement. Such differences become even more pronounced for topics concerning socially sensitive polarizing issues, such as massive pharmaceutical interventions. Here, we investigate how emotional responses change during large-scale...
Stefano Menini, Teresa Paccosi, Sara Tonelli, Marieke Van Erp, Inger Leemans, Pasquale Lisena, Raphael Troncy, William Tullett, Ali Hürriyetoğlu, Ger Dijkstra, Femke Gordijn, Elias Jürgens, Josephine Koopman, Aron Ouwerkerk, Sanne Steen, Inna Novalija, Janez Brank, Dunja Mladenic, Anja Zidar. Proceedings of the 3rd Workshop on Computational Approaches to Historical Language Change. 2022.
Stefano Menini, Teresa Paccosi, Serra Sinem Tekiroğlu, Sara Tonelli. Proceedings of the 7th Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature. 2023.