Keynote Els Lefever
About Els Lefever
Els Lefever is Associate Professor at the LT3 language and translation technology team at Ghent University, which conducts fundamental and applied research for NLP. She started her career as a computational linguist at the R&D-department of Lernout & Hauspie Speech products and holds a PhD in computer science from Ghent University on ParaSense: Parallel Corpora for Word Sense Disambiguation (2012). Els has a strong expertise in machine learning of natural language and multilingual natural language processing, with a special interest for computational semantics, language modeling of lower-resourced languages and multilingual terminology extraction. She currently supervises PhD research on complex reasoning in large language models, argumentation mining in social media, the automatic detection of irony in online text, multimodal emotion detection and generation, and NLP approaches for low(er)-resourced languages, such as cuneiform, Byzantine Greek, or historical travelogues. She teaches Terminology and Translation Technology, Language Technology, Localisation, Digital Text Analysis, Digital Humanities and feature-based machine learning courses.
The Keynote
Title: From Manuscripts to Neural Networks: Unlocking the Past with Natural Language Processing
Abstract
Natural Language Processing (NLP) is revolutionizing how we interact with text, not only in modern contexts, but also in how we interpret the deep past. This keynote explores how cutting-edge NLP methods are being applied to some of the most challenging and data-scarce areas in the humanities, namely historical and ancient texts.
We begin with a high-level overview of NLP and recent advances in machine learning, before addressing the unique challenges of applying computational models to fragmented, noisy, and often low-resourced data. From inconsistent orthography to multilingual corpora and diverse media formats, historical documents provide a rich and demanding testbed for current NLP methodologies.
Through three case studies, including cuneiform correspondence, Byzantine epigrams, and historical travelogues, we will examine how core NLP methods for linguistic annotation, text classification, and sentiment analysis perform on these specialized domains. These examples illustrate how NLP opens new perspectives for understanding the past, but also how extreme linguistic scenarios sharpen the methods and tools we use across the broader field of Digital Humanities.