9h00-9h15: Welcome and opening presentation (Miguel Couceiro & Pierre-Alexandre Murena) [slides]
9h15-10h15: Plenary Talk by Yves Lepage [slides]
Title: Analogy on text data
Abstract: The notion of analogy moved immediately from mathematics (natural numbers) to the study of the form of words and the study of their meaning. It was also used in a hidden form in comparative linguistics, and then explicitly at the birth of modern linguistics. In natural language processing, it has been used in computational morphology and distributional semantics. This talk will discuss the application of analogy to the processing of text data, be it characters, word forms, chunks, N-grams or sentences, in string or vector representations.
10h30-11h00: Masked prompt learning for formal analogies beyond words (Liyan Wang*, Yves Lepage) [paper] [slides]
11h00-11h30: Theoretical study and empirical investigation of sentence analogies (Stergos Afantenos*, Suryani Lim, Henri Prade, Gilles Richard) [paper] [slides]
11h30-12h00: Solving Morphological Analogies Through Generation (Esteban Marquer*, Shane Peter Kaszefski-Yaschuk, Kevin Chan, Camille Saran, Miguel Couceiro) [paper]
13h30-14h00: A Galois Framework for the Study of Analogical Classifiers (Miguel Couceiro*, Erkko Lehtonen) [paper] [slides]
14h00-14h30: Measuring the Feasibility of Analogical Transfer using Complexity (Pierre-Alexandre Murena) [paper] [slides]
14h30-15h00: Towards a Model of Visual Reasoning (Ekaterina Y. Shurkova*; Leonidas Doumas, Invited talk) Cancelled.
15h30-16h00: Exploring Analogical Inference in Healthcare (Safa Alsaidi*, Miguel Couceiro, Sophie Quennelle, Anita Burgun, Nicolas Garcelon, Adrien Coulet) [paper] [slides]
16h00-16h30: Analogical Proportions (Christian Antic, Invited talk)
17h00-18h00: Plenary Talk by Kenneth Forbus [slides]
Title: Analogy as a Technology for Machine Learning
Abstract: There is ample evidence that analogy, as described by Gentner’s structure-mapping theory, is a core operation in human cognition, used everywhere from visual reasoning through conceptual change. Analogical learning operates over relational representations and is incremental, inspectable, and data-efficient. This talk will describe the analogy stack that my group has developed around these ideas and surveys a few of the ways it has been used. Some example explorations in hybridizing analogy and other ML techniques will also be described, which seems to be a promising frontier for research.
18h00-18h30: Closing discussion