Experimental Work

In this section we attempt to confirm our hypothesis that both corpus-based and knowledge-based methods can improve the accuracy of each other. The first subsection shows the results of preprocessing the test data with the maximum entropy method (ME) in order to help the specification marks method (SM). Next, we test the opposite, if preprocessing the test data with the domain heuristic can help the maximum entropy method to disambiguate more accurately.

The last experiment combines the vME system (the majority voting system) and SM method. Actually, it relies on simply adding the SM as one more heuristic to the voting scheme.



Subsections