Neural Associative Memory for Dual-Sequence Modeling
This work addresses NLP tasks like textual entailment with a novel associative memory approach, though it is incremental with mixed performance.
The authors tackled dual-sequence modeling in NLP by proposing a new architecture based on associative memory, achieving competitive results on textual entailment, but found limitations in sequence-to-sequence tasks requiring additional supervision.
Many important NLP problems can be posed as dual-sequence or sequence-to-sequence modeling tasks. Recent advances in building end-to-end neural architectures have been highly successful in solving such tasks. In this work we propose a new architecture for dual-sequence modeling that is based on associative memory. We derive AM-RNNs, a recurrent associative memory (AM) which augments generic recurrent neural networks (RNN). This architecture is extended to the Dual AM-RNN which operates on two AMs at once. Our models achieve very competitive results on textual entailment. A qualitative analysis demonstrates that long range dependencies between source and target-sequence can be bridged effectively using Dual AM-RNNs. However, an initial experiment on auto-encoding reveals that these benefits are not exploited by the system when learning to solve sequence-to-sequence tasks which indicates that additional supervision or regularization is needed.