Module 1B Exam
Seemed to go well.
Question 1 - 17/20
Information retrieval question. Define "precision at rank 10" - full marks. Formula and explanation for Mean Average Precision - full marks. 11-point average precision - should be full marks. Might have lost a mark or two on TREC evaluations, but pretty much a model answer.
Question 2 - 17/20
Stochastic Context-Free Grammars. Three derivations (one of which was ambiguous... tsk tsk) - full marks. Two problems with SCFGs: discrimination in N->N N and NP->NP PP rules with ambiguity - full marks. How to add in UB PSG features - full marks? Lastly use of a stochastic UB PSG as a language model - probably dropped a couple of marks.
Question 3 - 17/20
Typed Feature-structure Grammars. Certain about two of my answers to the first part, almost certain about another and reasonably confident about the fourth - maybe a couple of marks dropped? Nailed the issues in the second part - miscoordination between adverb/PP (want them to unify, but different HEAD tags), pluralisation ("the X and the Y" is plural, but unifies as singular), implicit coordination ("X, Y or Z" vs "X or Y or Z") and over coordination of determiners ("the and that or a and the" = determiner?!). Full marks on that part.
Question 6 - 14/20
Semantics, quantifiers and scope. Generalised quantifier description - full marks. Variable binding limiting scope - full marks. Reification of events vs event-free logic - maybe a mark lost? Scope of quantifiers - full marks. Last part, worth 8 marks, is a bit random - I don't think I explained any of it, but I clearly had some relevant stuff about qeqs and MRS... maybe a couple of marks?
So, estimating 60/80 for 1A and 65/80 for 1B, that'd be 78% overall...
Results are due at the end of the month. Hopefully the markers will be generous.
Now, beer!
Question 1 - 17/20
Information retrieval question. Define "precision at rank 10" - full marks. Formula and explanation for Mean Average Precision - full marks. 11-point average precision - should be full marks. Might have lost a mark or two on TREC evaluations, but pretty much a model answer.
Question 2 - 17/20
Stochastic Context-Free Grammars. Three derivations (one of which was ambiguous... tsk tsk) - full marks. Two problems with SCFGs: discrimination in N->N N and NP->NP PP rules with ambiguity - full marks. How to add in UB PSG features - full marks? Lastly use of a stochastic UB PSG as a language model - probably dropped a couple of marks.
Question 3 - 17/20
Typed Feature-structure Grammars. Certain about two of my answers to the first part, almost certain about another and reasonably confident about the fourth - maybe a couple of marks dropped? Nailed the issues in the second part - miscoordination between adverb/PP (want them to unify, but different HEAD tags), pluralisation ("the X and the Y" is plural, but unifies as singular), implicit coordination ("X, Y or Z" vs "X or Y or Z") and over coordination of determiners ("the and that or a and the" = determiner?!). Full marks on that part.
Question 6 - 14/20
Semantics, quantifiers and scope. Generalised quantifier description - full marks. Variable binding limiting scope - full marks. Reification of events vs event-free logic - maybe a mark lost? Scope of quantifiers - full marks. Last part, worth 8 marks, is a bit random - I don't think I explained any of it, but I clearly had some relevant stuff about qeqs and MRS... maybe a couple of marks?
So, estimating 60/80 for 1A and 65/80 for 1B, that'd be 78% overall...
Results are due at the end of the month. Hopefully the markers will be generous.
Now, beer!