First page Back Continue Last page Overview Graphics
Future Work
The method presented here can easily be applied to many data-sets, and can empirically buttress theses
There is a lot of room for improvement
- Add smoothing to missing data, like Good-Turing-estimation
- Filter out more troublesome POS n-grams, like those mainly related to sentence-length (or only keep those of interest...)
And for fine-tuning and experimenting
- Find out how big the data-set must be for good results
- Try and evaluate with different measures