Solving analogies on words based on minimal complexity transformation
| Autor: | Jean-Louis Dessalles, Marie Al-Ghossein, Antoine Cornuéjols, Pierre-Alexandre Murena |
|---|---|
| Přispěvatelé: | Dessalles, Jean-Louis, Bessiere, Christian, Department of Computer Science, Institut Polytechnique de Paris, AgroParisTech, Aalto-yliopisto, Aalto University, Data, Intelligence and Graphs (DIG), Laboratoire Traitement et Communication de l'Information (LTCI), Institut Mines-Télécom [Paris] (IMT)-Télécom Paris-Institut Mines-Télécom [Paris] (IMT)-Télécom Paris, Département Informatique et Réseaux (INFRES), Télécom ParisTech, Institut Polytechnique de Paris (IP Paris), Mathématiques et Informatique Appliquées (MIA-Paris), AgroParisTech-Institut National de la Recherche Agronomique (INRA), Aalto University, Finland |
| Jazyk: | angličtina |
| Rok vydání: | 2020 |
| Předmět: |
[INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI]
0209 industrial biotechnology Computer science [SCCO.COMP]Cognitive science/Computer science MDL Computer Science::Computation and Language (Computational Linguistics and Natural Language and Speech Processing) 02 engineering and technology analogy Transformation (music) [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI] Algebra 020901 industrial engineering & automation [SCCO.COMP] Cognitive science/Computer science 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing complexity induction |
| Zdroj: | IJCAI International Joint Conference on Artificial Intelligence (IJCAI-2020) International Joint Conference on Artificial Intelligence (IJCAI-2020), Jan 2021, Kyoto, Japan. pp.1848-1854 |
| Popis: | Due to the Covid outbreak, the Conference has been postponed to January, 2021, but the proceedings have been published in 2020.; International audience; Analogies are 4-ary relations of the form “A is to Bas C is to D”. When A, B and C are fixed, we callanalogical equation the problem of finding the correctD. A direct applicative domain is Natural LanguageProcessing, in which it has been shown successfulon word inflections, such as conjugation ordeclension. If most approaches rely on the axiomsof proportional analogy to solve these equations,these axioms are known to have limitations, in particularin the nature of the considered flections.In this paper, we propose an alternative approach,based on the assumption that optimal word inflectionsare transformations of minimal complexity.We propose a rough estimation of complexity forword analogies and an algorithm to find the optimaltransformations. We illustrate our method ona large-scale benchmark dataset and compare withstate-of-the-art approaches to demonstrate the interestof using complexity to solve analogies onwords. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|