Flyer

Apertium is a free software (GPL) machine translation platform; it was initially designed to translated between the Romance languages of the Iberian peninsula, but is now being used for more distant pairs.

Who is developing it ?

The Apertium engine is being developed in the Transducens research group at the Department de Llenguatges i Sistemes Informàtics within the Universitat d'Alacant. Linguistic data are being developed by Transducens, the Seminario de Lingüística Informàtica of the Universidade de Vigo, the Institut Universitari de Lingüística Aplicada at the Universitat Pompeu Fabra in Barcelona, along with a number of companies and independent free software developers both in Spain and abroad.

Funding

The Spanish Ministry of Industry, Tourism and Commerce funded the development of the engine and three initial language pairs: Spanish-Catalan, Spanish-Galician and Spanish-Portuguese. The project has also received funding from the Generalitat de Catalunya (Government of Catalonia) to develop language pairs such as English-Catalan, Occitan-Catalan and Occitan-Spanish, and from the Romanian Ministry of Foreign Affairs to develop translators between Spanish-Romanian and Catalan-Romanian.

There are currently seven supported translation pairs published using the Apertium platform. These are:

• Spanish-Catalan
• Spanish-Portuguese
• Spanish-Galician
• Catalan-French
• Catalan-Romanian
• Spanish-Romanian
• English-Catalan

Other pairs currently under active development, but without a stable release include: French-Spanish, English-Afrikaans, English-Welsh, Catalan-Romanian, Spanish-Basque and English-Polish.

How good is it?

The quality of the final translations depends greatly on the amount of time spent in development, and the closeness of the languages. For example Spanish-Catalan has approximately 95% accuracy, but Spanish-Portuguese has around 90%. For less related and unreleased pairs such as English-Afrikaans, the accuracy, excluding unknown words is somewhere around 70%.

Downloading

Current versions of the engine, linguistic data and documentation can be found on our SourceForge project page ([1]). Further documentation and discussion can be found both on our wiki ([2]) and mailing list ([3]).

Development

The project is always looking for developers who are interested in improving the engine, working on new language pairs (especially those involving less-used or under-resourced languages), creating interfaces, or adapting the software to fit your needs.

Applications

• Multilingual management of web content such as media
• Rapid localisation of free software
• Translation of documentation between a more resourced language and a less resourced language