GAINON DE FORSAN DE GABRIAC Clara
Forschungsgruppe : MLIA
Datum, an dem das LIP6 verlassen wurde : 31.12.2021
https://lip6.fr/Clara.Gainon-de-Forsan-de-Gabriac
Forschungsleitung (Direction de recherche) : Patrick GALLINARI
Co-Betreuung : GUIGUE Vincent
Deep Natural Language Processing for User Representation
The last decade has witnessed the impressive expansion of Machine Learning ( ML ) and in particular Deep Learning ( DL ) methods, both in academic research and the private sector. This success can be explained by the ability DL to model ever more complex entities. In particular, Representation Learning ( RL ) methods focus on building latent representations from heterogeneous data that are versatile and re-usable, namely in Natural Language Processing ( NLP ). In parallel, the ever-growing number of systems relying on user data (social media, recommendation systems ... ) brings its own lot of challenges. This work proposes methods to leverage the representation power of NLP in order to learn rich and versatile user representations.
Firstly, we detail the works and domains associated with this thesis. We study Recommendation; a field where User Representations ( URs) has long been at the heart of researches. We then go over recent NLP advances and how they can be applied to leverage user-generated texts, before detailing Generative models, that we feel are one of the purest expressions of RL.
Secondly, we present a Recommender System ( RS ) that is based on the combination of a traditional Matrix Factorization ( MF ) representation method and a sentiment analysis model. The association of those modules forms a dual model that is trained on user reviews for rating prediction. Experiments show that, on top of improving accuracy performances, the model allows us to better understand what the user is really interested in in a given item, as well as to provide explanations to the suggestions made.
Finally, we introduce a new task-centered on UR: Professional Profile Learning. User profiles are often composed of several fields or attributes of different nature, and it is our thesis that user-generated textual data are not only rich enough to uniquely represent a user, but also to predict the other attributes of the profile. We thus propose an NLP -based framework, Resumé, to learn and evaluate professional profiles on different tasks, including next job generation.
Verteidigung einer Doktorarbeit : 13.12.2021
Mitglieder der Prüfungskommission :
Anne Boyer, Professeure, Université de Lorraine - LORIA [Rapporteur]
Julien Velcin, Professeur, Université Lyon 2-- ERIC Lab [Rapporteur]
Mohamed Chetouani, Professeur, Sorbonne Université -- ISIR
Alejandro Bellogín, Associate Professor, Universidad Autónoma de Madrid - IRG@UAM
Patrick Gallinari, Professeur, Sorbonne Université -- LIP6
Vincent Guigue, Maître de Conférence, Sorbonne Université - LIP6
Publikationen 2018-2021
-
2021
- C. Gainon de Forsan de Gabriac : “Deep Natural Language Processing for User Representation”, these, verteidigung einer doktorarbeit 13.12.2021, forschungsleitung (direction de recherche) Gallinari, Patrick, co-betreuung : Guigue, Vincent (2021)
-
2020
- C. Gainon de Forsan de Gabriac, V. Guigue, P. Gallinari : “Resume: A Robust Framework for Professional Profile Learning & Evaluation”, ESANN 2020 - 28th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, Bruges, Belgium (2020)
-
2019
- Ch. Dias , C. Gainon de Forsan de Gabriac, V. Guigue, P. Gallinari : “RNN & modèle d’attention pour l’apprentissage de profils textuels personnalisés”, Document numérique - Revue des sciences et technologies de l'information. Série Document numérique, vol. 22 (3), pp. 9-27, (Hermès) (2019)
-
2018
- Ch. Dias , C. Gainon de Forsan de Gabriac, V. Guigue, P. Gallinari : “RNN & modèle d’attention pour l’apprentissage de profils textuels personnalisés”, CORIA 2018 - 15e COnférence en Recherche d'Informations et Applications, Rennes, France (2018)