AI NLP
NLP is a field of CS, AI, and _____ ______ - answer computational linguistics
MLP applications leverage deep neural networks to handle ______ ______ and natural
language understanding - answer speech recognition
machine translation (historically motivated translate Russian to English); first systems
using mechanical translation failed. Vagueness, discourse structure, auto correction,
ambiguity - answer
(he brought a brand new car and drove it home) is an example of? - answer anaphora
i will be there. What did you do? (example of?) - answer indexicality
she learned how to play Mozart at a very young age - answer metonymy (Mozart's
music)
Progress in NLP: _____ (part of speech and name entity recognition); text ______
(spam filtering); parsing: text to tree, sentiment analysis; machine translation; info
extraction - answertagging, classification
is summarization, q/a, and dialog systems a work in process? (Joakim Nivre) -
answeryes
text classification: Naive bayes. A training data (xi, yi), xi is a feature vector and yi is a
discrete label. d features, n examples. Consider Document classification: A new
example with feature values x_new = (a1, a2, ... ad).
we want to predict the label y_new of the new example so y_new will be -
answerargmax p(y | a1, a2, ... ad)
the naive bayes alg is learning based on the _____ counts in the dataset; classification
use the same formula y_new = argmax(p(y) MULT-all p(y|a) - answerfrequency
given a doc; define an attribute for each word position in the doc. The value of the
attribute is the english word in that position. To reduce the # of probabilitys that need to
be estimated, besides NB indep assumption, we assume that the probability of a given
word w_k occurrence is independent of the word position within the text. That is, p(x1 =
w_k|c_j) = ______ where nj = total # word positions in all training examples of class cj;
n_k is # times the word w_k is found in amont these n_j word positions. - answer(n_k +
1) / (n_j + |Vocab| )
NLP is a field of CS, AI, and _____ ______ - answer computational linguistics
MLP applications leverage deep neural networks to handle ______ ______ and natural
language understanding - answer speech recognition
machine translation (historically motivated translate Russian to English); first systems
using mechanical translation failed. Vagueness, discourse structure, auto correction,
ambiguity - answer
(he brought a brand new car and drove it home) is an example of? - answer anaphora
i will be there. What did you do? (example of?) - answer indexicality
she learned how to play Mozart at a very young age - answer metonymy (Mozart's
music)
Progress in NLP: _____ (part of speech and name entity recognition); text ______
(spam filtering); parsing: text to tree, sentiment analysis; machine translation; info
extraction - answertagging, classification
is summarization, q/a, and dialog systems a work in process? (Joakim Nivre) -
answeryes
text classification: Naive bayes. A training data (xi, yi), xi is a feature vector and yi is a
discrete label. d features, n examples. Consider Document classification: A new
example with feature values x_new = (a1, a2, ... ad).
we want to predict the label y_new of the new example so y_new will be -
answerargmax p(y | a1, a2, ... ad)
the naive bayes alg is learning based on the _____ counts in the dataset; classification
use the same formula y_new = argmax(p(y) MULT-all p(y|a) - answerfrequency
given a doc; define an attribute for each word position in the doc. The value of the
attribute is the english word in that position. To reduce the # of probabilitys that need to
be estimated, besides NB indep assumption, we assume that the probability of a given
word w_k occurrence is independent of the word position within the text. That is, p(x1 =
w_k|c_j) = ______ where nj = total # word positions in all training examples of class cj;
n_k is # times the word w_k is found in amont these n_j word positions. - answer(n_k +
1) / (n_j + |Vocab| )
