Reference-Aware Language Models
This addresses the challenge of incorporating predictable external information into language models for applications requiring entity and attribute mentions, though it appears incremental as it builds on existing attention mechanisms.
The authors tackled the problem of language models lacking explicit reference mechanisms by proposing a class of models that treat reference as a stochastic latent variable, enabling access to external databases and internal state for tasks like dialogue and recipe generation, with experiments showing performance on three tasks using deterministic attention variants.
We propose a general class of language models that treat reference as an explicit stochastic latent variable. This architecture allows models to create mentions of entities and their attributes by accessing external databases (required by, e.g., dialogue generation and recipe generation) and internal state (required by, e.g. language models which are aware of coreference). This facilitates the incorporation of information that can be accessed in predictable locations in databases or discourse context, even when the targets of the reference may be rare words. Experiments on three tasks shows our model variants based on deterministic attention.