# Tracking Discrete and Continuous Entity State for Process Understanding

## TL;DR

This paper introduces a neural architecture that models both discrete and continuous aspects of entity states in procedural text, improving process understanding and achieving state-of-the-art results on QA tasks.

## Contribution

The paper presents a structured neural model combining recurrent tracking of entity states with a neural CRF to enforce global constraints, a novel approach for process understanding.

## Key findings

- Achieves state-of-the-art results on ProPara dataset.
- Effectively models entity state constraints over time.
- Improves accuracy in process-related QA tasks.

## Abstract

Procedural text, which describes entities and their interactions as they undergo some process, depicts entities in a uniquely nuanced way. First, each entity may have some observable discrete attributes, such as its state or location; modeling these involves imposing global structure and enforcing consistency. Second, an entity may have properties which are not made explicit but can be effectively induced and tracked by neural networks. In this paper, we propose a structured neural architecture that reflects this dual nature of entity evolution. The model tracks each entity recurrently, updating its hidden continuous representation at each step to contain relevant state information. The global discrete state structure is explicitly modeled with a neural CRF over the changing hidden representation of the entity. This CRF can explicitly capture constraints on entity states over time, enforcing that, for example, an entity cannot move to a location after it is destroyed. We evaluate the performance of our proposed model on QA tasks over process paragraphs in the ProPara dataset and find that our model achieves state-of-the-art results.