Improving Neural Language Models with a Continuous Cache
This work addresses the challenge of improving language model predictions for natural language processing tasks, representing an incremental advancement over existing memory-augmented methods.
The authors tackled the problem of adapting neural language models to recent context by proposing a continuous cache mechanism, which significantly outperformed recent memory-augmented networks on several datasets.
We propose an extension to neural network language models to adapt their prediction to the recent history. Our model is a simplified version of memory augmented networks, which stores past hidden activations as memory and accesses them through a dot product with the current hidden activation. This mechanism is very efficient and scales to very large memory sizes. We also draw a link between the use of external memory in neural network and cache models used with count based language models. We demonstrate on several language model datasets that our approach performs significantly better than recent memory augmented networks.