Improving Document Image Understanding with Reinforcement Finetuning

09/26/2022
by   Bao-Sinh Nguyen, et al.
0

Successful Artificial Intelligence systems often require numerous labeled data to extract information from document images. In this paper, we investigate the problem of improving the performance of Artificial Intelligence systems in understanding document images, especially in cases where training data is limited. We address the problem by proposing a novel finetuning method using reinforcement learning. Our approach treats the Information Extraction model as a policy network and uses policy gradient training to update the model to maximize combined reward functions that complement the traditional cross-entropy losses. Our experiments on four datasets using labels and expert feedback demonstrate that our finetuning mechanism consistently improves the performance of a state-of-the-art information extractor, especially in the small training data regime.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
06/04/2023

A Topological Approach to Measuring Training Data Quality

Data quality is crucial for the successful training, generalization and ...
research
03/04/2019

Complement Objective Training

Learning with a primary objective, such as softmax cross entropy for cla...
research
08/04/2021

Human-In-The-Loop Document Layout Analysis

Document layout analysis (DLA) aims to divide a document image into diff...
research
06/10/2019

Neural Keyphrase Generation via Reinforcement Learning with Adaptive Rewards

Generating keyphrases that summarize the main points of a document is a ...
research
10/09/2019

Integrating Behavior Cloning and Reinforcement Learning for Improved Performance in Sparse Reward Environments

This paper investigates how to efficiently transition and update policie...
research
03/25/2016

Improving Information Extraction by Acquiring External Evidence with Reinforcement Learning

Most successful information extraction systems operate with access to a ...
research
08/08/2023

Adapting Foundation Models for Information Synthesis of Wireless Communication Specifications

Existing approaches to understanding, developing and researching modern ...

Please sign up or login with your details

Forgot password? Click here to reset