Dynamic Entity Representations in Neural Language Models
This addresses the challenge of entity tracking in long documents for natural language processing applications, representing a novel method rather than an incremental improvement.
The paper tackles the problem of tracking entity evolution in long documents by introducing EntityNLM, a language model that dynamically updates entity representations and contextually generates mentions. The model consistently outperforms strong baselines and prior work across multiple tasks including language modeling, coreference resolution, and entity prediction.
Understanding a long document requires tracking how entities are introduced and evolve over time. We present a new type of language model, EntityNLM, that can explicitly model entities, dynamically update their representations, and contextually generate their mentions. Our model is generative and flexible; it can model an arbitrary number of entities in context while generating each entity mention at an arbitrary length. In addition, it can be used for several different tasks such as language modeling, coreference resolution, and entity prediction. Experimental results with all these tasks demonstrate that our model consistently outperforms strong baselines and prior work.