Talks and Presentations

Visual Language Models and its applications to social good

October 04, 2024

Talk, IIMAS-UNAM and ITAM Universities in Mexico City.,

Can artificial intelligence models understand a story told through images? In this talk, we will explore how systems that combine language and vision learn to reason visually by connecting information across sequences of images. I will focus on two recent projects, ImageChain and MuSeD, which demonstrate how these advances enable the analysis of complex visual content and support applications aimed at social good. Finally, I will discuss the open challenges in developing models capable of producing reliable explanations.

Beyond Words: Analyzing Social Media with Text and Images

October 04, 2024

Talk, University of Gothenburg,

People express their opinions and experiences through text and images in social media platforms. Large-scale analysis of the content that people use in social media has diverse applications spanning linguistics, geography, political science and advertising. Three specific studies will be presented (1) Point-of-interest: studying the relationship between the content of a post and the posting location type (e.g. park, restaurant); (2) Political Ads: identifying the political ideology of ad sponsors and distinguishing between official political parties and third-party organizations; (3) Influencer Content: examining influencer content, with a focus on multimodal methods to detect monetized posts.

Exploring Transformers and Limitations in Language Modeling

October 04, 2023

Talk, Mexican NLP Summer School,

This tutorial explores language modeling techniques in Natural Language Processing (NLP), covering key concepts from traditional approaches to Transformer architectures. Beginning with an introduction to NLP and language modeling, it delves into probabilistic language models and progresses to neural language models, emphasizing the significance of embeddings for semantic representation. Moving on to Transformer models, we will discuss key concepts such as multi-head attention mechanisms, masked language modeling, and encoder models. Additionally, the tutorial addresses the limitations of large language models, providing insights into challenges and considerations for leveraging these models effectively in practical applications. Mexican NLP Summer School

Multimodal Models for Computational Social Science

October 04, 2023

Talk, University of Copenhagen,

People express their opinions and experiences through text and images in social media platforms. Large-scale analysis of the content that people use in social media has diverse applications spanning linguistics, geography, political science and advertising. Three specific studies will be presented (1) Point-of-interest: studying the relationship between the content of a post and the posting location type (e.g. park, restaurant); (2) Political Ads: identifying the political ideology of ad sponsors and distinguishing between official political parties and third-party organizations; (3) Influencer Content: examining influencer content, with a focus on multimodal methods to detect monetized posts. [Slides]

Annotation hackathon for Law & Technology in Europe LLM

April 21, 2022

Talk, , Utrecht University

Legal annotations on an Instagram influencer dataset for law&tech LLM students. Presented an introduction to machine learning and annotations, and organized an annotation task on an Instagram dataset of posts by mega and micro influencers. This is a legal task that is becoming increasingly important for consumer and media authorities, but the scale of the creator economy is beyond their capabilities.

NLP+ for Computational Social Science

April 20, 2022

Talk, Institute of Data Science, Maastricht University

Large-scale analysis of the language people use in social media have applications in a variety of fields such as linguistics, political science, journalism, and geography. At the same time, language in social media presents challenges to the development of NLP models. In this talk, I will discuss recent work which attempts to overcome some of these challenges by jointly modeling related tasks or by combining relevant information such as visual content. We explore the task of automatically identifying political parody on social media and how we can combine parody with other relevant figurative devices such as humor and sarcasm. Finally, I’ll discuss work on integrating text and images for point-of-interest type prediction: studying the relationship between social media posts and the type of location from which it was posted (e.g. park, restaurant).

AI4Dignity Counterathon, LMU

July 26, 2021

Talk, Workshop AI, Extreme Speech and Disinformation, Online

Multidisciplinary discussion on datasets and NLP models for extreme speech detection. [More Information]

Machine Learning Models for Analyzing Social Media Text (Spanish)

March 06, 2021

Talk, ITAM Research Seminar, Mexico City (Online)

El análisis a gran escala del lenguaje utilizado por la gente en redes sociales tiene aplicaciones en diversas áreas como ciencia política, periodismo y geografía. En esta plática, hablaré de cómo podemos analizar la información de redes sociales con diferentes modelos como redes neuronales artificiales y modelos “pre-entrenados” (BERT, RoBERTa).

• Parodia política: ¿es posible identificar automáticamente una cuenta de parodia política de una cuenta genuina? Por ejemplo: @realDonaldTrFan (parodia) vs. @realDonaldTrump (genuina).
• Clasificación de lugares de interés: ¿cómo analizar la relación entre el contenido de una publicación y el tipo de lugar (e.g. parque, restaurante) desde el cual se publicó?

NLP for Computational Social Science

December 03, 2020

Talk, Sheffield NLP Seminar, Sheffield, UK (Online)

Large-scale analysis of the language people use in social media have applications in a variety of fields such as linguistics, political science, journalism, and geography. In this talk, I concentrate on 3 studies (1) Political Parody: researching political parody as a figurative device in social media; (2) Political Ads: identifying the political ideology of the ad sponsor, and whether the ad sponsor was an official political party or a third-party organization; (3) Point-of-interest: studying the relationship between the text of social media and the type of location from which it was posted (e.g. park, restaurant).

Analyzing Political Parody in Social Media

August 05, 2020

Poster Session, Online Poster Session - Sheffield University, Sheffield, UK (Online)

Online Poster Session hosted by Sheffield University for Women in Computer Science. Participants could navigate around the Mozilla courtyard and view PhD students work. More information [Slides] [Virtual Courtyard]