Skip to Main content Skip to Navigation
Master Thesis

Traduction automatique de documents manuscrits et typographiés en arabe par couplage étroit entre systèmes

Abstract : During the last few years, the field of automatic processing of natural languages has seen rapid developments, especially in research on scanned documents, which involves two interrelated fields: Optical character recognition (OCR) and machine translation (MT). In this research paper, we investigate machine translation of scanned documents, whether handwritten or typed. First of all, we created a Arabic-French machine translation system. Next we improved our system by translating a words lattice constructed from noisy OCR outputs (N-best). Finally, we added a specific preprocessing for out-of-vocabulary (OOV) words using word-embeddings (word2vec).
Document type :
Master Thesis
Complete list of metadatas

Cited literature [17 references]  Display  Hide  Download

https://dumas.ccsd.cnrs.fr/dumas-01494465
Contributor : Uga - Bulles <>
Submitted on : Thursday, March 23, 2017 - 2:57:03 PM
Last modification on : Friday, July 17, 2020 - 11:10:23 AM
Long-term archiving on: : Saturday, June 24, 2017 - 2:24:42 PM

Identifiers

  • HAL Id : dumas-01494465, version 1

Citation

Kamel Bouzidi. Traduction automatique de documents manuscrits et typographiés en arabe par couplage étroit entre systèmes. Sciences de l'Homme et Société. 2016. ⟨dumas-01494465⟩

Share

Metrics

Record views

285

Files downloads

532