previous next Title Contents

6 Conclusions

We feel that we are on the right track for evaluation of these fairly simple NLP tools.

Although we only carried out tests on two different products, we could see how they differ in lexical coverage and in handling errors, as well as in suggestion adequacy.

The unexpected behaviour that we encountered in one of the products, although relatively "minor" (for instance, it does not cause the machine to crash, nor corrupts the file in any noticeable way) showed that products still lack robustness in this field.


previous next Title Contents