Page 159 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications
P. 159
148 Chapter 5 Depression discovery in cancer communities using deep learning
Out
Out
Out
Out
Out
Forward Backward
LSTM LSTM
Figure 5.10 Bidirectional LSTM model. LSTM, long short-term memory.
• mul: The forward and backward outputs are multiplied
together.
• Concat: The forward and backward outputs are concatenated
together, providing double the number of outputs to the next
layer. This is the by-default method and is rottenly used in
Bi-LSTM studies.
• ave: The average of the outputs is taken.
Now at the output layer, the sequence prediction is done.
5. Conclusion
In this chapter, we introduced an approach for depression
detection in cancer communities on Twitter. We integrate NLP
with deep learning for this task. We used optimized word embed-
ding for depression detection. Furthermore, we proposed the use
of different deep learning models for depression detection from
tweets on the user level. We also highlighted the benefitof
deep learning models that they are good for feature engineering
using word embedding to find well the context of the sentence.
The model learns itself on the features and helps in giving correct
prediction.
