Lost in Translation – The Parallel Meaning Bank
In this talk I will present the objectives, challenges, and first results of a project that investigates the role of meaning in translation following a data-driven and computational approach. In this project we construct a large corpus of translations (for four languages: English, Dutch, German, and Italian) annotated with formal meaning descriptions based on an extended version of Hans Kamp’s Discourse Representation Theory. The meanings are produced in a semi-automatic way for English, manually corrected, and then projected on Dutch, German, and Italian. This enables us to build and evaluate semantic parsers for all four languages, and compare meanings of source and target to verify if anything got lost in translation. But this is easier said than done, and in this talk I will also address the challenges that we are facing and possible solutions.
The Parallel Meaning Bank: http://pmb.let.rug.nl