The UOB-Telecom ParisTech Arabic Handwriting Recognition and Translation Systems for the OpenHart 2013 Competition

Olivier Morillot (1) , Cristina Oprean (1) , Laurence Likforman-Sulem (1) , Chafic Mokbel (2) , Edgar Chammas (2) , Emmanuèle Grosicki (3)

1 LTCI - Laboratoire Traitement et Communication de l'Information
2 UOB - University of Balamand [Liban]
3 DGA/CTA/DT/GIP - CEP Arcueil

This article is a description of the two systems proposed for the recognition of Arabic handwritten text lines and for the automatic translation of text-line and sentence images into English text. The recognition systems are based on HMMs (Hidden Markov Models) and BLSTMs (bi-directional long short term memory) recurrent networks. Two SMT (Statistical Machine Translation) systems based on MOSES [1] were built for the evaluation system: one on text-line translation and one for sentence translation.

Domaines: Intelligence artificielle [cs.AI]; Traitement des images [eess.IV]; Traitement du texte et du document

Olivier Morillot, Cristina Oprean, Laurence Likforman-Sulem, Chafic Mokbel, Edgar Chammas, et al.. The UOB-Telecom ParisTech Arabic Handwriting Recognition and Translation Systems for the OpenHart 2013 Competition. 12th International Conference on Document Analysis and Recognition (ICDAR), 2013, Aug 2013, Washington DC, United States. pp.NIST. ⟨hal-00948985⟩