#2

Question 1

Intrinsic evaluation

Answer 1

define a metric and check which system does the best

Question 2

Extrinsic evaluation

Answer 2

check how much the performance is improved when u use the output as an input to a larger system

Question 3

issues with NB Classifier

Answer 3

a missing feature
out-of-vocabulary words
stop words

Question 4

how you can deal with a missing feature in the dataset

Answer 4

smoothing

Question 5

stop words

Answer 5

very frequent, uninformative words

Question 6

smoothing

Answer 6

add 1 to the count of features

Question 7

generative classifier

Answer 7

learn a model how the data is generated

Question 8

discriminative classifier

Answer 8

learn which features best predict a certain class

Question 9

How to prevent overfitting

Answer 9

by fine-tunning, K-fold-cross validation

Question 10

Prior probability

Answer 10

represents your initial belief about the likelihood of the hypothesis

Question 11

likelihood

Answer 11

the evidence that an instance has been generated by a given class

Question 12

Posterior probability

Answer 12

represents your updated belief about the likelihood of the hypothesis after taking into account the evidence

Question 13

How NB treats a text document

Answer 13

Bag of Words

Question 14

Advantage of Rule systems

Answer 14

1. Robust
2. no need of large dataset

Question 15

Disadvantage of Rule systems

Answer 15

1. Expensive to write
2. Require domain knowledge
3. Rigit when dealing with ambiguity

Question 16

4 terms of learning

Answer 16

generalize (understanding new info) ,

infer( use it within different situations),

analogize (interpret the results),

adaptive ( the moving system adapts quickly when the environment is changing)

Question 17

Discriminative classifier

Answer 17

Learn which features best predict a certain class

Question 18

Text simplification

Answer 18

Mapping a text into another text

Question 19

main difference between intrinsic and extrinsic evaluation

• what was your algorithm trained to do

Answer 19

Intrinsic: how well the system is doing what’s trained to do

Extrinsic: using the system in a larger environment, how well the algorithm is performing on a bigger environment

Question 20

Does bag of words crack co-occurence

Answer 20

No, bag of words does not crack co-occurrence

Question 21

Why NBC is a “naive”

Answer 21

because assumes that the features are conditionally independent of each other

Question 22

out-of-vocabulary words

Answer 22

words that appear at the test set, but not in the training set

Question 23

how to treat OOV words?

Answer 23

-ignore them
- make a dedicated feature in the training set