Exercise 1 - subclasses of verbs
What do we mean by subclasses of verbs?
We may say
- Mary slept.
- Mary saw John.
but not
- *Mary slept John
- *Mary saw
Hence 'slept' and 'saw' belong to different classes of verbs. We would like to have at least 5 verbs that behaves like *slept' and similarly for the other verb classes.
Exercise 1 - Lexicon size
What do we have to include so many words in each category? Aren't we interested in the grammar and the principles.
A characteristic property of natural languages compared to the general class of CFGs is that natural languages have large lexica compared to syntactic rules. CFGs for naturla languages should therefore use general rules as far as possible, e.g.
- VP -> TV NP
and not
- VP -> 'saw' NP
If the lexicon is too small we don't see why this is important.
Also in exercise 5 we consider the effect of separating out lexical insertion and that becomes clearer if the lexicon isn't too small.
Exercise 2 - Ambiguities
It is hard to see the ambiguities in (a) and (b).
this may be easier to see in other examples. For (b) consider the difference between
- Mary ate pasta with parmesan
- Mary ate pasta with a fork
For (a) it may be more difficult to construct the least likely reading, but what about
- a car from Japan with automatic transmission
Exercise 2 - Answering the questions
The text says "...explain in one or two sentences...", but this is not mentioned under "submission".
I have now included "the explanations" under submission and uploaded corrected exercis set.
Exercise 5 - middle case
What is meant by "the middle case" in ex 5?
The original program carries the structure
if a: do_1
elseif b: do_2
else: do_3
We describe that do_1 (the first case) and do_3 (the last case) only need smaller changes. The middle case is the do_2-part.