engleski

Statistical machine translation of Croatian weather forecast: How much data do we need?

This research is a first step towards a system for translating Croatian weather forecast into multiple languages. This steps deals with the Croatian-English language pair. The parallel corpus consists of a one-year sample of the weather forecasts for the Adriatic consisting of 7, 893 sentence pairs. Evaluation is performed by best known automatic evaluation measures BLUE, NIST and METEOR, as well as by evaluating manually a sample of 200 translations. In this research we have shown that with a small-sized training set and the state-of-the art Moses system, decoding can be done with 96% accuracy concerning adequacy and fluency. Additional improvement is to be expected by increasing the training set size.

statistical machine translation; weather forecast; automatic evaluation; human evaluation

nije evidentirano