• Aucun résultat trouvé

The ACCEPT Academic Portal: Bringing Together Pre-editing, MT and Post-editing into a Learning Environment

N/A
N/A
Protected

Academic year: 2022

Partager "The ACCEPT Academic Portal: Bringing Together Pre-editing, MT and Post-editing into a Learning Environment"

Copied!
2
0
0

Texte intégral

(1)

Proceedings Chapter

Reference

The ACCEPT Academic Portal: Bringing Together Pre-editing, MT and Post-editing into a Learning Environment

BOUILLON, Pierrette, et al.

BOUILLON, Pierrette, et al . The ACCEPT Academic Portal: Bringing Together Pre-editing, MT and Post-editing into a Learning Environment. In: İIknur Durgar El‐Kahlout, Mehmed Özkan, Felipe Sánchez‐Martínez, Gema Ramírez‐Sánchez, Fred Hollowood, Andy Way. Proceedings of the 18th Annual Conference of the European Association for Machine Translation . European Association for Machine Translation, 2015. p. 212

Available at:

http://archive-ouverte.unige.ch/unige:77673

Disclaimer: layout of this document may differ from the published version.

1 / 1

(2)

The ACCEPT Academic Portal: Bringing Together Pre-editing, MT and Post-editing into a Learning Environment

Pierrette Bouillon, Johanna Gerlach, Asheesh Gulati, Victoria Porro, Violeta Seretan Université de Genève FTI/TIM, [email protected]

www.accept-portal.unige.ch

Description

The ACCEPT Academic Portal is a user-centred online platform specifically designed to offer a complete machine translation workflow including pre-editing and post-editing steps for teaching purposes. The platform leverages technology developed in the ACCEPT1 European Project (2012-2014) devoted to improving the translatability of user-generated content. Origi- nally available as a series of plug-ins and demonstrators on the ACCEPT portal2, the various software components have been interconnected into an easy-to-use platform reproducing all phases of a real MT workflow. The platform provides a unique environment to study the inter- action between MT-related processes and to assess the contribution of new technologies to translation. It will be useful for research and teaching purposes alike.

The platform allows a user to select existing data (or supply their own data, in plain text for- mat) and to subject it to a sequence of processes, until the desired output is reached. Source and target content reformulation can be performed automatically or interactively, the interface allowing experimentation with specific editing rules, visual comparison, on-the-fly translation, XLIFF-based recording of post-editing actions, and one-click export of results. The steps can be executed in a flexible manner according to the desired scenario; users may, for instance, upload their own machine-translated texts and perform post-editing only. Users can stop at any step in the workflow and download their results. In-context documentation (tool tips, user guide) is available at all steps. The platform has a minimalistic app-like look-and-feel for op- timised user experience. It is designed for the non-expert, and can therefore bring the MT benefits to a larger community of users.

The platform’s main modules and functionalities are briefly described below.

- Start page: Selection of data (existing; own data entered in text area; own data uploaded as text file); selection of language pairs (en–fr, fr–en and en–de are currently supported);

selection of processing scenario (different combinations of pre-editing, MT and post-editing).

- Pre-editing module: Automatic or interactive checking using the ACCEPT rules devel- oped for user-generated content using the Acrolinx technology (www.acrolinx.com).

- MT module: Translation using the ACCEPT phrase-based Moses system adapted to user-generated content.

- Post-editing module: Free post-editing; interactive checking using ACCEPT post-editing rules for user-generated content; final check with pre-editing rules; XLIFF report.

- Statistics page: Final summary; editing statistics; XLIFF-based report (keystrokes, time);

results download.

The ACCEPT Academic Portal can be freely accessed at: www.accept-portal.unige.ch.

1ACCEPT: Automated Community Content Editing PorTal, FP7 grant agreement 288769 (www.accept- project.eu; accessed: March, 2015).

2 www.accept-portal.com (Accessed: March, 2015).

212

Références

Documents relatifs

• (Ginzburg et al., 2014) propose a universal concerning negative EPs—if NEG is a lan- guage’s word that can be used as a negation and in cross-turn correction, then NEG can be used

Oracle scores indicate that gains in terms of BLEU score are still possible, even with a PBMT system built on in-domain data and with- out introducing new data during the

An MTPE app has to live with the constraints of mobile devices. 147) states, “serious usability problems are the result of a poor decision about a trade-off

The impact of such model was examined by the evaluation of the quality of post-edited segments and throughput rates demonstrated by the participants as well as the

Instead of training two SMT systems PE → SRC and SRC → MT on huge amounts of paral- lel data, and doing a round-trip translation of the monolingual data, we train two small PE → SRC

We run our experiments on English-Italian data, by comparing the performance of different neural APE models (batch and online) used to correct the output of NMT systems of

Recent approaches to the Au- tomatic Post-editing (APE) of Machine Translation (MT) have shown that best re- sults are obtained by neural multi-source models that correct the raw

Recent advances in automatic post-editing (APE) have shown that it is possible to automatically correct system- atic errors made by machine translation systems. However, most of