The Numb er of Proportional Analogies between Marker-based Chunks in 11 European Languages
Kota Takeya, Jing Sun, Yves Lepage · 2011
An example-based machine translation (EBMT) system based on proportional analogies requires numerous proportional analogies between linguistic units to work properly. Consequently, long sentences cannot be handled directly in such a framework. Cutting sentences into chunks would be a solution. Using different markers, we count the number of proportional analogies between chunks in 11 European languages. As expected, the number of proportional analogies between chunks is very high. Whereas samples of thousand English sentences from the Europarl corpus do not lead to any analogy between sentences, we obtain several tens of thousands of analogies between the chunks extracted from these sentences using 10 markers. These results are very promising for the EBMT system that we intend to build.