Skip to main content
. Author manuscript; available in PMC: 2019 Dec 3.
Published in final edited form as: KDD. 2019 Aug;2019:2547–2555. doi: 10.1145/3292500.3330770

Figure 1:

Figure 1:

Joint Model with document representation. The first attention layer attends to the words in a sentence. The second attention layer attends to the sentence and it’s neighbouring sentences which act as the local context for the sentence. The document vector is created separately which provides global context to the model. The BiLSTM parameters shared between different parts of the model are color-coded so that the same color represents shared parameters