%0 Conference Proceedings %T Improving Patent Mining and Relevance Classification using Transformers %+ EuroMov - Digital Health in Motion (Euromov DHM) %+ Total Research and Technology Feluy ( [TRTF]) %A Ding, Théo %A Vermeiren, Walter %A Ranwez, Sylvie %A Xu, Binbin %< avec comité de lecture %B APIA 2021 - Conférence Nationale sur les Applications Pratiques de l’Intelligence Artificielle (événement affilié à PFIA 2021) %C Bordeaux, France %P p. 81-90 %8 2021-06-28 %D 2021 %Z 2105.03979 %K Transformers %K fine-tuning %K patent classification %K naturallanguage processing %K language models %Z Computer Science [cs]Conference papers %X Patent analysis and mining are time-consuming and costly processes for companies, but nevertheless essential if they are willing to remain competitive. To face the overload induced by numerous patents, the idea is to automatically filter them, bringing only few to read to experts. This paper reports a successful application of fine-tuning and retraining on pre-trained deep Natural Language Processing models on patent classification. The solution that we propose combines several state-of-the-art treatments to achieve ourgoal : decrease the workload while preserving recall and precision metrics. %G English %L hal-03277767 %U https://imt-mines-ales.hal.science/hal-03277767 %~ INSTITUT-TELECOM %~ UNIV-PARIS13 %~ EM-ALES %~ IRISA %~ UR1-MATH-STIC %~ UR1-UFR-ISTIC %~ UNIV-MONTPELLIER %~ INRIA-300009 %~ INSTITUTS-TELECOM %~ SORBONNE-PARIS-NORD %~ UR1-MATH-NUM %~ EUROMOV-DHM %~ UM-2015-2021 %~ APIA-2021