I was going through a paper on Sarcasm Detection which proposed a Hybrid Attention Based LSTM. The Model included some linguistic features to be extracted first, which included Entropy of a word in a sentence. Unfortunately, the paper does not explain it very well. The explanation given is:
The entropy of a word is defined as the degree of randomness of a word concerning the sentence. It is calculated using the following equation:
S(A/B) = (p(ai / bj) ∗ log(p(ai))) / p(bj)
where S(A/B) represents the probability of the word ‘A’ with respect to the sentence ‘B’, and p(ai / bj) means the probability of ai given bi.
If someone understands the terms, kindly explain these.
Title of Paper: Hybrid attention-based Long Short-Term Memory network for
sarcasm identification
Link to Paper