w7 l2 sequence labeling

Question 1

what are named entity recognition

Answer 1

named entitiy = proper noun

named enetity recognition is an importnatnt task

Question 2

what is do we use for NER word-by word sequelce labelling task

Answer 2

BIO scheme

beginning of the NE experession

insided for all the word tokens NE

O for all words outside the NEs

*E is end of expression and S is for signle token entities

BIOES

2n+1 tags for n entity types

Question 3

what is a weakness of genrative hmms for sequential modeling, whats the alternative

Answer 3

can take into accoutn unseen events

instead we can directly take the features of the words and surrounding words

Question 4

what is an alternative way for seuqence labeling

Answer 4

linear chain conditional random fiedls

CRF

Question 5

what is the main difference between hmm and crf

Answer 5

hmm is generative and crf is discrimiative

Question 6

what is the goal of the crf

Answer 6

given input described by its features,

discriminate between all possible tag sequences and select the most probable among them

Question 7

what can named entities be

Answer 7

person location organisation geopoltical entities

also dates times and prices

Question 8

whtas a problem with hmm

Answer 8

it cant consider unseen events