Home

MTUOC: Neural and Statistical Machine Translation

The MTUOC project distributes a series of components (scripts and programs) that allow for tranining Neural and Statistical Machine Translation systems, as well as translating with these systems. The porject makes all this available regardless the operating system you are using. The project distributes the following components:

• Python modules providing basic functionalities as tokenization, truecasing, etc.
• Bash an Pyhton scripts and several configuration files for corpus preprocessing and MT system taining.
• MTUOC server: this component receives a segment to translate from a client and sents it to the MT Server (Moses, Marian, ONMT or ModernMT). It performs the required preprocessing steps. When it receivess the translated segment, it postprocess it and sends it back to the client. This server can mimetize several kinds of MT servers and can act as MTUOC, Moses, Marian, ONMT, NMTWizard and ModernMT. This makes MTUOC server compatible with a lot of CAT Tools.
• MTUOC client (the so-called MTUOC-Translator). This program can deal with text and XLIFF files, connect with a MT server and translate the files. The program can connect with MTUOC, Moses, Marian, ONMT, NMTWizard and ModernMT servers.
• MTUOC-eval: a program for automatic evaluation of machine translation systems. It allows to calculate: BLEU, NIST, WER, EditDistance and TER (using tercom).
• MTUOC Virtual Machines: most of MT toolskits works only with Linux. We provide several Virtual Box virtual machines to run the MTUOC server under any operating system.
• Pre-trained Neural Machine Translation Systems ready-to-use with MTUOC-server.

The wiki uses Markdown syntax.

Project Members:

• Antoni Oliver (admin)