Fig 1.
Model architecture of (A) CBOW and (B) Skip-gram.
Fig 2.
The general model of the CNN.
Table 1.
Hardware specification of our deep learning computer.
Fig 3.
Processing steps of Word2vec.
Table 2.
Training datasets for word2vec.
Table 3.
List for disease related keywords.
Table 4.
Parameter of word2vec.
Fig 4.
Sequence(M), word vector rate, and learning rate of (a) CBOW and (b) Skip-gram.
Table 5.
Parameter values of CNN.
Fig 5.
Our CNN architecture with CBOW, Skip-gram, and random learning algorithms.
Fig 6.
(A) Accuracy and (B) F1 score of CNN with CBOW as a function of epoch for various training volumes in news articles.
Fig 7.
(A) Accuracy and (B) F1 score of CNN with Skip-gram as a function of epoch for various training volumes in news articles.
Fig 8.
(A) Accuracy and (B) F1 score of CNN with the random vector as a function of epoch for various training volumes in news articles.
Fig 9.
(A) Accuracy and (B) F1 score of CNN with CBOW as a function of epoch for various training volumes in tweets.
Fig 10.
(A) Accuracy and (B) F1 score of CNN with Skip-gram as a function of epoch for various training volumes in tweets.
Fig 11.
(A) Accuracy and (B) F1 score of CNN with the random vector as a function of epoch for various training volumes in tweets.
Fig 12.
(A) Accuracy and (B) F1 score of CNN with CBOW, CNN with Skip-gram, and CNN with the random vector as a function of training volume when the epoch is fixed to 100 in news articles.
Fig 13.
(A) Accuracy and (B) F1 score of CNN with CBOW, CNN with Skip-gram, and CNN with the random vector as a function of training volume when the epoch is fixed to 100 in tweets.
Table 6.
Experiments analysis.