We are happy to announce TM Cleaner , a software for identifying the translation units in translation memories or parallel corpora that contains segments that are not translations of each other.
TM Cleaner has the following features :
- It is written in python and java.
- It is hosted at github : https://github.com/SoimulPatriei/TMCleaner
- Free for research and commercial use: it is released under LGPL license.
- It uses a combination of machine learning and rule-based classification to identify wrong translation segments.
- It is fully integrated with scikit-learn: this means that all classification algorithms in scikit-learn can be easily integrated.
- Can be used standalone or as a web service.
- It works in three modalities : with Bing translation engine, with Hunalign sentence aligner and with Fastalign word aligner .
- It contains a module for evaluating the accuracy of classification
- It has easy to follow tutorials.
For comments, bug reports or suggestions please use the email address firstname.lastname@example.org.