iit-cs585 / main Goto Github PK

Can anyone share the implementation of viterbi; who got precise values for sentence probability ?

(of-course with professor's consent @aronwc)

Am a bit confused on the structure of dicts (one of the return parameter in make_feature_dicts())

method defines as - A list of dicts, one per token, containing the features for that token.

If its a list of dict per token, would dicts be a list of list of dicts where each sublist has all feature dicts for a token -
[[{feature1 : 1}, {feature2 : 1}, {feature3 : 1}], [feature1 : 1], {feature3 : 1},{feature4 : 1}] ?

But if we go by test_a3 , dicts seem to be just a list of dict something like
[{feature1 : 1}, {feature2 : 1}, ....]

If this is the case, how do we associate token to features ?

rules = [('S', ['NP', 'VP']),
('NP', ['ProperNoun']),
('ProperNoun', ['John', 'Mary']),
('VP', ['V', 'ProperNoun']),
('V', ['likes', 'hates'])]
if rules are specified as above will the following be considered as valid production or not
1.('S', ['NP'])
2.('S', ['VP'])

Hi professor,
Can you please clarify on these:
a)run_fsa
description of input_symbols:
input_symbols...list of strings representing the input string
I am assuming input symbols as a list of characters. Also in the test case it is taken as list of characters by converting one single string. Is it list of strings?

b) get_productions corner case
In the doc string of Tree class, children can be empty. Suppose we define Tree as Tree(“S”), in this case, there are no children, then by assuming RHS of the production rule can be empty and considered as valid, I am considering output as [('S',[])]. Is this correct?

Just got confused about what "n_params" exactly is. Can anyone explain it?
Thanks

I tried installing math module on version 3.5.2, it failed. Upgraded python to 3.6, am still getting the following error -
$pip3 install math
Collecting Math
Downloading Math-0.5.tar.gz
Complete output from command python setup.py egg_info:
Traceback (most recent call last):
File "", line 1, in
File "/private/tmp/pip-build-yoomnugu/Math/setup.py", line 2, in
raise RuntimeError("Package 'Math' must not be downloaded from pypi")
RuntimeError: Package 'Math' must not be downloaded from pipit

I have got the same values for train and test data, but confusion matrix is slightly deviating from what's given in Log.txt. Has anyone else faced this... I wonder where and what could have gone wrong. Any pointers will be appreciated.

training data shape: (27867, 18290)

testing data shape: (28033, 18290)

confusion matrix:
Predicted I-LOC I-MISC I-ORG I-PER O
Actual
I-LOC 861 13 58 131 87
I-MISC 54 334 43 40 98
I-ORG 155 21 405 261 173
I-PER 66 10 42 1304 134
O 58 15 33 97 23540

evaluation matrix:
I-LOC I-MISC I-ORG I-PER O
precision 0.721106 0.849873 0.697074 0.711402 0.979527
recall 0.748696 0.586995 0.399015 0.838046 0.991450
f1 0.734642 0.694387 0.507519 0.769549 0.985453

I just want to confirm the syllabus for Test 1 , it is till L14 right ?

Hi professor,
Q3 score is not added to the total score in the blackboard. In the schedule, Q4 is missing now.

Professor,

a4 is due on 4/13(in 3 days). Can it please be uploaded ?

Thanks

Hi Professor,
Could you please do exam review this week. A brief review of what's covered will help.

For example, we are training a corpus that doesn't contain a given word "EU".

Folks,
I am in search for a group/group-mate for CS-585 Project Work. let me know if anyone still up for it.
email me: [email protected]

Regards,
Sunny

Links to reading in Schedule.md for lectures 2,3,4 does not lead to the desired section of the book.