Neurosymbolic Information Extraction from Transactional Documents

TL;DR

This paper introduces a neurosymbolic framework for extracting information from transactional documents, combining language models with symbolic validation to improve accuracy and enable zero-shot learning.

Contribution

It presents a schema-based approach with symbolic validation, relabeled datasets, and high-quality label generation for knowledge distillation in transactional document extraction.

Findings

Significant improvements in F1-scores and accuracy

Effective zero-shot extraction enabled by symbolic validation

Enhanced knowledge distillation through high-quality labels

Abstract

This paper presents a neurosymbolic framework for information extraction from documents, evaluated on transactional documents. We introduce a schema-based approach that integrates symbolic validation methods to enable more effective zero-shot output and knowledge distillation. The methodology uses language models to generate candidate extractions, which are then filtered through syntactic-, task-, and domain-level validation to ensure adherence to domain-specific arithmetic constraints. Our contributions include a comprehensive schema for transactional documents, relabeled datasets, and an approach for generating high-quality labels for knowledge distillation. Experimental results demonstrate significant improvements in $F_1$-scores and accuracy, highlighting the effectiveness of neurosymbolic validation in transactional document processing.