Aspect Level Sentiment Classification with Deep Memory Network
This work addresses sentiment analysis at the aspect level, which is important for applications like product reviews, but it is incremental as it builds on existing neural attention and memory network ideas.
The paper tackles aspect-level sentiment classification by introducing a deep memory network that explicitly captures the importance of context words for sentiment polarity, achieving performance comparable to state-of-the-art SVM and substantially better than LSTM-based models on laptop and restaurant datasets, with a 9-layer network being 15 times faster than LSTM on CPU.
We introduce a deep memory network for aspect level sentiment classification. Unlike feature-based SVM and sequential neural models such as LSTM, this approach explicitly captures the importance of each context word when inferring the sentiment polarity of an aspect. Such importance degree and text representation are calculated with multiple computational layers, each of which is a neural attention model over an external memory. Experiments on laptop and restaurant datasets demonstrate that our approach performs comparable to state-of-art feature based SVM system, and substantially better than LSTM and attention-based LSTM architectures. On both datasets we show that multiple computational layers could improve the performance. Moreover, our approach is also fast. The deep memory network with 9 layers is 15 times faster than LSTM with a CPU implementation.