COPPA V2. 0: Corpus Of Parallel Patent Applications Building Large Parallel Corpora with GNU Make

WIPO seeks to help users and researchers to overcome the language barrier when searching patents published in different languages. Having collected a big multilingual corpus of translated patent applications, WIPO decided to share this corpus in a product called COPPA (Corpus Of Parallel Patent Applications) to stimulate research in Machine Translation and in language tools for patent texts. A first version was released in 2011 but contained only French and English languages. It has been decided to release a major update of this product containing newer data (from 2011 up to 2014) but also other languages (German, English, French, Japanese, Korean, Portuguese, Spanish, Russian and Chinese). This corpus can be used for terminology extraction, cross-language information retrieval or statistical machine translation. With the new version a huge number of files (more than 26 million) has to be processed. We describe the technical process in details.