Analysis of User’s Opinion using Deep Neural Network Techniques

Through many research and discoveries it has been widely accepted that aspect-level sentiment classification is achieved effectively by using Long Short-Term Memory (LSTM) network combined with attention mechanism and memory module. As existing approaches widely depend on the modeling of semantic relatedness of an aspect, at the same time we ignore their syntactic dependencies which are already a part of that sentence. This will result in undesirably an aspect on textual words that are descriptive of other aspects. So, in this paper, to offer syntax free contexts as well as they should be aspect specific, so we propose a proximity-weighted convolution network. To be more precise, we have one way of determining proximity weight which is dependency proximity. The construction of the model includes bidirectional LSTM architecture along with a proximity-weighted convolution neural network