Lights, Camera, Action! A Framework to Improve NLP Accuracy over OCR documents

Autor:	Gupte, Amit, Romanov, Alexey, Mantravadi, Sahitya, Banda, Dalitso, Liu, Jianjie, Khan, Raza, Meenal, Lakshmanan Ramu, Han, Benjamin, Srinivasan, Soundar
Rok vydání:	2021
Předmět:	Computer Science - Computation and Language Computer Science - Machine Learning
Druh dokumentu:	Working Paper
Popis:	Document digitization is essential for the digital transformation of our societies, yet a crucial step in the process, Optical Character Recognition (OCR), is still not perfect. Even commercial OCR systems can produce questionable output depending on the fidelity of the scanned documents. In this paper, we demonstrate an effective framework for mitigating OCR errors for any downstream NLP task, using Named Entity Recognition (NER) as an example. We first address the data scarcity problem for model training by constructing a document synthesis pipeline, generating realistic but degraded data with NER labels. We measure the NER accuracy drop at various degradation levels and show that a text restoration model, trained on the degraded data, significantly closes the NER accuracy gaps caused by OCR errors, including on an out-of-domain dataset. For the benefit of the community, we have made the document synthesis pipeline available as an open-source project. Comment: Accepted to the Document Intelligence Workshop at KDD 2021. The source code of Genalog is available at https://github.com/microsoft/genalog
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/2108.02899 Zobrazit plný text záznamu View this record from Arxiv