Corpus ID: 215814449

How recurrent networks implement contextual processing in sentiment analysis

@article{Maheswaranathan2020HowRN,
  title={How recurrent networks implement contextual processing in sentiment analysis},
  author={Niru Maheswaranathan and David Sussillo},
  journal={ArXiv},
  year={2020},
  volume={abs/2004.08013}
}
  • Niru Maheswaranathan, David Sussillo
  • Published 2020
  • Computer Science, Mathematics
  • ArXiv
  • Neural networks have a remarkable capacity for contextual processing--using recent or nearby inputs to modify processing of current input. For example, in natural language, contextual processing is necessary to correctly interpret negation (e.g. phrases such as "not bad"). However, our ability to understand how networks process context is limited. Here, we propose general methods for reverse engineering recurrent neural networks (RNNs) to identify and elucidate contextual processing. We apply… CONTINUE READING

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 69 REFERENCES

    Understanding Hidden Memories of Recurrent Neural Networks

    VIEW 2 EXCERPTS

    Linguistically Regularized LSTMs for Sentiment Classification

    VIEW 15 EXCERPTS
    HIGHLY INFLUENTIAL