Weakly Supervised One-shot Classification using Recurrent Neural Networks with Attention: Application to Claim Acceptance Detection

Charles Condevaux (1) , Sébastien Harispe (2) , Stéphane Mussard (1) , Guillaume Zambrano (3)

1 CHROME - Détection, évaluation, gestion des risques CHROniques et éMErgents (CHROME) / Université de Nîmes
2 LGI2P - Laboratoire de Génie Informatique et Ingénierie de Production
3 UNIMES - Université de Nîmes

Determining if a claim is accepted given judge arguments is an important non-trivial task in court decisions analyses. Application of recent efficient machine learning techniques may however be inappropriate for tackling this problem since, in the Legal domain, labelled datasets are most often small, scarce and expensive. This paper presents a deep learning model and a methodology for solving such complex classification tasks with only few labelled examples. We show in particular that mixing one-shot learning with recurrent neural networks and an attention mechanism enables obtaining efficient models while preserving some form of inter-pretability and limiting potential overfit. Results obtained on several types of claims in French court decisions, using different vectorization processes, are presented.

Mots clés: Classification; legal analysis; one-shot learning; deep learning

Domaines: Intelligence artificielle [cs.AI]; Réseau de neurones [cs.NE]

Charles Condevaux, Sébastien Harispe, Stéphane Mussard, Guillaume Zambrano. Weakly Supervised One-shot Classification using Recurrent Neural Networks with Attention: Application to Claim Acceptance Detection. JURIX 2019 32nd International Conference on Legal Knowledge and Information Systems, Dec 2019, Madrid, Spain. ⟨10.3233/faia190303⟩. ⟨hal-02407405⟩