UPM|Mediateca

Jan Hajic, Charles University, Czech Republic Symposium Artificial Intelligence and Ceremony of Awards. 21 and 22 October 2019. Language Resources have a long history, which in fact predates their role as a training data for machine learning approaches to Natural Language Processing. After briefly mentioning their history from the Brown Corpus to Universal Dependencies, the talk will focus on how various linguistically relevant data contribute to both linguistic research as well as applications, and how their role is changing alongside new developments in machine learning, and specifically in today’s prevalent Deep Learning approaches. There are three types of Language Resources: naturally occurring resources (such as large monolingual or parallel corpora, both for text and speech, often correlated to some real world knowledge or events), and linguistically structured and/or annotated resources, such as dictionaries or lexicons and linguistically annotated corpora. Each of them has different use for different purposes. the talk will argue that despite the latest progress in Deep Learning and high quality end-to-end practical systems, annotated and structured linguistic resources play a very important role not only in basic linguistic research, but also in developing new methods for natural language understanding as well as an indispensable ingredient in the quest for true Artificial Intelligence.