Comments to the Project B

Here are some additional remarks to the obligatory assignment 4 (Project B):

1. In some classifiers, categorical features with high cardinality (like lemmas for predicates/arguments) can greatly increase the training time after being converted to sparse binary vectors. Thus, it is not obligatory to use lemma of the predicate (the first feature from section 2.2 of the assignment) in your baseline system. However, you can decide to somehow cope with the sparsity problem or use other classifiers, if you want.
2. The global scoring part (section 2.4 of the assignment) is an advanced-level task (*). It is possible to get a passing grade without fully completing this part. Still, you should at least try after you are done with other parts.
3. Some helper functions to get you started can be found on the course Github page.