Category: Talks

Where: Sala conferenze (3th floor)
When: 07/12/2023 11:30

Wolfgang S. Schmeisser-Nieto is a PhD student in Linguistic, Literary and Cultural Studies and a member of the Centre de Llenguatge i Computació (CLiC).at the Universitat de Barcelona, Spain. He also participates in the European project “STERHEOTYPES: STudying European Racial Hoaxes and sterEOTYPES”. His thesis, directed by Dr Mariona Taulé and Dr Simona Frenda, focuses on the automatic detection of stereotypes toward immigrants spread through social media. Among other activities, he has co-organized the shared evaluation task DETESTS at IberLEF 2022, held at the conference of the Spanish Society for Natural Language Processing (SEPLN). 

Abstract:  
Stereotypes are the result of human cognition that allows us to classify others into social groups in accordance with supposed shared attributes. While it allows us to organize our reality, it also leads us to wrongly attribute characteristics to certain people that we perceive as belonging to some social category, causing prejudice and discrimination against them. One of the ways in which stereotypes are manifested is through language, therefore, the increasing presence of social media has facilitated the spread and reinforcement of these stereotypes, especially when they work in favor of political and social interests. My research aims at getting a better understanding of the phenomenon of stereotypes toward immigrants through the analysis of language. From psychological, linguistic and computational approaches, I explore proposed and new taxonomies on stereotypes and the linguistic characteristics found in the texts that make difficult the agreement among annotators, and also its automatic detection by models, with a special emphasis on the implicit expressions of stereotypes and the role of context.

On December 22, Viviana Patti organized the two-stage four-handed seminar by Chiara Zanchi (University of Pavia) and Gosse Minnema (University of Groningen), in conjunction with the Linguistic Resources for Natural Language Processing course.

Abstract

Different linguistic expressions can conceptualize the same event from different viewpoints by emphasizing certain participants over others. In this joint presentation, Chiara Zanchi will try to show that different theoretical constructs of cognitive linguistics, including construction grammar (Goldberg 1995) and frame semantics (Fillmore 1985), are useful for analyzing social perspective-taking on news events, in particular relating to societal power imbalances (e.g., between men and women, Italians and refugees, car drivers and pedestrians). Then, based on his own PhD work, Gosse Minnema will show how several recent Natural Language Processing methodologies can be used to perform large-scale frame-based corpus analysis, predict how linguistic choices in news text influence people’s perception of events, and even help to suggest alternative perspectives on an event. 

Bios

Gosse Minnema: Gosse Minnema is a fourth-year PhD student at the University of Groningen in the Netherlands. In his research, he tries to build bridges between computational, cognitive, and social approaches to language, focusing specifically on the semantics of events and its relation to society.

Chiara Zanchi: Chiara Zanchi is assistant professor in linguistics at the University of Pavia, where she teaches introductory courses in linguistics (BA) and Laboratory of Linguistic Data Analysis (MA in Theoretical, Applied and Modern Languages Linguistics). Her main research interests are in the fields of Indo-European linguistics (with a focus on Ancient Greek), pragmatics and discourse analysis (of ancient and modern languages), cognitive linguistics, and language resource construction.

When: 22/12/2022

• Part I: Cognitive linguistic background by Chiara Zanchi – in the morning at 10am, in Aula informatica 2 or on WebEx
• Part II: Hands on it! Natural Language Processing methodologies for event framing by Gosse Minnema – in the afternoon at 2pm, in Aula informatica 1 or on WebEx

Location: Palazzetto Aldo Moro – Via Sant’Ottavio – 12 – Torino

For more info: viviana.patti@unito.it

In this seminar, Oscar Araque will present his research group (Intelligent Systems Group) from the Universidad Politécnica de Madrid, as well as his research interests and recent projects. In the talk, he will briefly describe some previous works in semantic modeling, sentiment and emotion analysis, distributed representations, radicalization and propaganda detection.

Short Bio:

Oscar Araque received the graduate and master’s degrees in telecommunication engineering from the Technical University of Madrid (Universidad Politécnica de Madrid), Spain, and the Ph.D. degree in the same university in 2020 with the thesis titled “A Distributional Semantics Perspective of Lexical Resources for Affect Analysis: An application to Extremist Narratives”. He is currently an Assistant Professor in Universidad Politécnica de Madrid. His research interest includes the application of machine-learning techniques for Natural Language Processing. More concretely, his interests lie in the introduction of specific domain knowledge into machine learning systems in order to enhance sentiment and emotion analysis techniques, and their applications to other domains such as radicalization narratives, moral value assessment and hate speech. He is currently working in the H2020 PARTICIPATION project, where he is using NLP/ML techniques for radical propaganda detection.

When: 09/05/2022 at 15.30

Where: in presentia at Sala Riunioni at the 3° floor

Adventures in Annotation for NLP

Most natural language processing tasks rely on human-labelled data in order to train supervised learning systems.
However, tasks can be subjective, and the “ground truth” labels may be
difficult or even impossible to ascertain. In this seminar, I will describe work on collecting, creating, analysing, and deploying labelled datasets for tasks including sarcasm detection, sentiment analysis, topic identification, and abuse detection in domains as diverse as social media, parliamentary
debates, and conversational AI.

Gavin Abercrombie is a postdoctoral research associate at Heriot-Watt University (Edinburgh, Scotland), where he is working on the ESPRC-funded project “Designing Conversational Assistants to Reduce Gender Bias”. He holds a PhD from the University of Manchester and an MSc from the University of Copenhagen, and is currently a visiting researcher at Bocconi University, Milan.

When: 18/11/2021 at 11.00

Where: in presentia (032_A_P03_3140)

Proper Question Forms Disentangling Inconsistencies in Dialogue Systems

This talk reports on the description of a specific class of clarification requests, adopted for the negotiation of grounded information in argumentation-based dialogue systems. Two studies are carried out to prove the adequateness of a specific form of polar question when a presupposition is contradicted by a new evidence. Whereas the first one proves the appropriateness of the negative form, the second one also demonstrates how the use of such a form can affect the principle of robustness, in terms of observability and recoverability, important in human–machine interaction applications. The two studies show that dialogue systems with such capabilities can lead to improved usability and naturalness in conversation. For this reason, I present here a system capable of detecting conflicts and of using argumentation strategies to signal them consistently with previous observations.

Maria Di Maro is a Post-Doc at the University of Naples ‘Federico II’ working on the interaction design for the project BRILLO (Bartending Robot for Interactive Long-Lasting Operations). She got her Ph.D. in linguistics at the University of Naples ‘Federico II’ in 2021 with the dissertation “Shouldn’t I use a polar question? Proper Question Forms Disentangling Inconsistencies in Common Ground”. Her research interests range from corpus collection and pragmatic annotations to computational pragmatics and the modeling of grounding processes in spoken dialogue systems. She is also passionate for Artificial Intelligence in general and graph databases as the vessel for twisted pragmatic reasonings.

When:  13/07/2021 at 11.00

Unaddressed challenges in persuasive dieting chatbots

Diet coaching gathered lots of interest in research. Recently, chatbots have been leveraged to address this task, with a focus on persuasion to motivate people towards behaviour change. In this talk we will take a look at current approaches in building persuasive dieting chatbots and expose a number of major unsolved challenges. We will motivate them with evidences from previous works and show that current chatbots don’t approach certain scenarios properly, hence limiting their communication and persuasion capabilities.

Simone Balloccu is a PhD student in Natural Language Generation (NLG) at the University of Aberdeen, UK. He received his BSc and MSc in computer science from the University of Cagliari. His research initially focused on unsupervised Natural Language Understanding (NLU) for the Singleton Expansion and Hypernym Discovery tasks. He is now working in the PhilHumans project (H2020) on the development of innovative healthcare AI technology. His current research focus is on user profiling, text comprehension enhancement and stress-based tailoring in the context of diet coaching.

When:  15/04/2021 at 10.00

Where: https://unito.webex.com/webappng/sites/unito/meeting/info/910eaf7ad0534d1ba92c5dde0a66a9a7

Language variation: challenges and future avenues

Recent developments in deep learning have shown striking performance improvements on a wide array of natural language processing (NLP) tasks. Despite the great progress, current approaches to NLP typically assume language is homogeneous, and thus fail to account for the intra- and extra-linguistic variations encoded in language, such as domains, genres, languages, and social factors. The main consequence is a dramatic drop in performance on out-of-distribution data, and unpredictable behaviors on new distributions, including the magnification of harmful stereotypes, and unfair and discriminatory decision making. In this talk, I will dig into the topic, introducing the theoretical notion of the variety space and the challenges which language variation entails. Current efforts and future avenues for research will be discussed, including transfer learning approaches, and the need for living benchmarks and transparent model and data statements.

Alan Ramponi is a PhD candidate in natural language processing at the University of Trento, Italy. He was a fellow at the Microsoft Research COSBI centre, Italy (2017-20), and a visiting PhD fellow in the NLPnorth group at the IT University of Copenhagen, Denmark (2019-20). He received the MSc in computer science cum laude from the University of Trento in 2017. His research focuses on language variation, and specifically in making natural language processing (NLP) robust and ultimately aware of variation across domains, social factors, and languages.

When:  09/04/2021 at 11.00

Where: https://unito.webex.com/webappng/sites/unito/meeting/info/910eaf7ad0534d1ba92c5dde0a66a9a7

Title: Embarrassingly Simple Unsupervised Aspect Extraction

In this talk, he will discuss a new unsupervised model for aspect identification, a subtask of aspect-based sentiment analysis. The motivation behind an unsupervised method is that aspects are highly domain-specific, which makes it difficult to reuse models on other datasets. For example, an aspect extractor trained on the restaurant domain is unlikely to transfer well to the laptop domain. This simple model, which we call Contrastive Attention (CAt) only requires word embeddings and a set of POS tags, requires no training, and achieves state-of-the-art unsupervised performance on a corpus of restaurant reviews. He will introduce the model, explain why it works, and discuss some of its up- and downsides, as well as discuss a recent extension of the model.

Paper: https://www.aclweb.org/anthology/2020.acl-main.290.pdf

Bio: Stéphan Tulkens recently graduated from the University of Antwerp with a Phd on the topic of orthographic representation and word recognition. Aside from word reading, his research interests include Named Entity Recognition, Word Sense Disambiguation, Aspect-Based Sentiment Analysis, and weakly supervised or unsupervised learning. He currently works as a Machine Learning Engineer at Slimmer A.I., a Groningen-based company, where he makes AI products to reduce overhead in the scientific publishing industry.

When: On December 22nd, 2020 at 3.00 pm