The Rise of Natural Language Processing:
and New Challenges
11 November 2021 - 12h15-14h00
Sciences III, Room 1S059 - Online, Zoom
Registration mandatory - Under this link
The Data Science Competence Center (CCSD) of the University of Geneva is pleased to invite you to the fifth edition of the Data Science Seminars, exploring natural language processing.
In one decade, Natural Language Processing (NLP) has grown from a niche field to a leading discipline. Its applications have multiplied in many disciplines, while its scientific community has grown accordingly. At the crossroads of linguistics, computer science, and artificial intelligence, this science and technology is indeed extremely promising, due to its ability to accurately extract information and insights contained in language data (documents, audio files) as well as categorize and organize these language files. By continuously refining the ways to process and analyze large amounts of natural language data, NLP specialists aims to make computers capable of "understanding" the contents of documents, including the contextual nuances of the language within them. As such, this technology raises ethical issues, which are become more important as its use is now rising to an unprecedented level.
Through concrete examples drawn from their research, the speakers at this seminar will present how they use NLP in their field to open up new avenues for research, but also what are the challenges they have encountered in using it. These presentations will notably highlight how the use of NLP can contribute to offer a better understanding of the Switzerland Philanthropic Organizations, paving the way to the establishment of synergies between diverse actors sharing the goal of achieving the SDGs and ultimately to boost the role of philanthropy. They will also underline how important and fruitful is the collaboration between philologists and computer scientists, specialised in NLP, to achieve valuable discoveries for the history of French. Finally, they will also address the ethical dimensions of NLP, exploring the issue of bias, but also, on a cognitive level, the fascinating similarities between humans and neural networks when it comes to their knowledge of the meaning of words.
Mapping the Swiss Philanthropic Landscape through Big-Data
Lucia Gomez Teijeiro, Assistant Research Scientist at the Department of Basic Neurosciences, Faculty of Medicine and Geneva Finance Research Institute (GFRI), University of Geneva.
Giuseppe Ugazio, Assistant Professor at the Geneva Finance Research Institute (GFRI), Geneva School of Economics and Management.
Switzerland occupies a starring position in the philanthropic ecosystem, counting with over 12000 philanthropic organizations (POs). However, the absence of a digital, comprehensive, and publicly available data resource describing Switzerland POs and their activities makes it challenging to comprehend the complexity of the sector. Partially due to this lack of knowledge, only a minority of Switzerland POs engage in cooperative networks to coordinate their activities. This, in turn, prevents the sector from achieving its maximum potential to generate a meaningful impact (Pfitzer, M. et al., 2010). Using state of the art Natural Language Processing and Machine Learning tools, in the present work we propose a novel approach to fill this gap. First, we analyzed currently active POs in Switzerland to provide a detailed view on the actions in which they engage and their operative modes; secondly, relying on existing topical lexicons we identified the presence of references to emotions and moral-values in order to determine potential motives for engaging in philanthropic activities. Finally, we measured the extent to which philanthropic action overlaps with the United Nations’ Sustainable Development Goals (SDGs) (GSDR 2019). This research is key to increase the visibility of the philanthropic motives and activities, for promoting synergies between diverse actors sharing the goal of achieving the SDGs and ultimately to boost the role of philanthropy in shaping a better future.
Studying linguistic change: on mutual benefits of collaborations between philology and NLP
Simon Gabay, Lecturer in Digital Humanities, Faculté des Lettres.
The quest for better digital editions is becoming increasingly important among philologists, but the conception of the best possible tools can only be achieved with the help of NLP specialists. Reversely, computer scientists can only benefit from collaborations with researchers specialised in data curation, who are able to prepare the necessary corpora for future state of the art solutions. Using the example of linguistic normalisation with automatic translation techniques, we will try to show preliminary results, but also unexpected outcomes such as valuable discoveries for the history of French.
Do current neural network know the meaning of words like people do ?
Paola Merlo, Professor of Computational Linguistics, Director of the Computational Learning and Computational Linguistics lab, Head of the unit for computer science for the humanities, Faculté des Lettres
To develop systems that process language in a way that is compatible with human expectations, we need computational representations of lexical and grammatical properties that form the basis of human knowledge of words and sentences and we need methods that require realistic amounts of computational and linguistic resources. In this brief presentation, I will investigate the lexical representations that are currently learnt by deep learning models, the most up-to-date machine learning techniques in artificial intelligence. While these methods are unlike humans in that they require unrealistic amount of data to learn, I will show that they reach a knowledge of the meaning of words that is similar to humans, both for a single language and in bilingual settings.