DeepCPCFG: Deep Learning and Context Free Grammars for End-to-End Information Extraction

Abstract

We combine deep learning and Conditional Probabilistic Context Free Grammars (CPCFG) to create an end-to-end system for extracting structured information from complex documents. For each class of documents, we create a CPCFG that describes the structure of the information to be extracted. Conditional probabilities are modeled by deep neural networks. We use this grammar to parse 2-D documents to directly produce structured records containing the extracted information. This system is trained end-to-end with (Document, Record) pairs. We apply this approach to extract information from scanned invoices achieving state-of-the-art results.

Artificial Intelligence for the Automated Creation of Multi-scale Digital Twins of the Built World—AI4TWINNING

“…CFGs approaches from image analysis [27] and document analysis [19] were explored and applied to a facade context in order to generate grammars. For the analysis of grammars, the framework Agglomerator [22] was used to extract a This is a pre-print for personal use only.…”

MORTY: Structured Summarization for Targeted Information Extraction from Scholarly Articles

“…TRIE [42] uses end-to-end system to jointly perform document reading and information extraction on everyday documents such as invoices, tickets, and resumes. Chua and Duffy [4] proposes a method for finding the suitable grammar set for the parsing and the extraction of information. Specifically for scholarly context, various systems has been created to extract and retrieve information from publications and scholarly articles.…”