An Artificial Intelligence Based Approach to Automate Document Processing in Business Area
Name
chen-brandonc-sm-sdm-2021-thesis.pdf
Description
Thesis PDF
Size
3.44 MB
Format
Adobe PDF
Checksum (MD5)
8038285914f3241c76d79a218e8c5147
Author(s)
Chen, Ta Hang
Advisor(s)
Gupta, Amar
Szolovits, Peter
Rhodes, Donna H.
Date Issued
June 2021
Publisher
Massachusetts Institute of Technology
Abstract
Automatic document processing is always a strategy for business executives to improve operational efficiency. With Optical Character Recognition (OCR) and machine learning techniques, businesses are able to apply Artificial Intelligence (AI) to automate the process. However, introducing an AI application to business is challenging; it is easy to fail because of the complexity between the technical and organizational components. This thesis considers document processing from a sociotechnical system perspective and leverages a four-step system analysis approach to identify the critical components.
This research also proposes a machine learning model using Support Vector Machine (SVM) as the classifier and Word2vec embeddings as document features to classify business documents. The proposed model reaches a 0.872 Macro F1-score using scanned business documents from the RVL-CDIP dataset. The proposed model outperforms the other commonly used rule-based algorithms, RIPPER and PART, showing that the proposed model is potentially suitable to be deployed into business to classify the
documents.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
System Design and Management Program.
Terms of Use
In Copyright - Educational Use Permitted
Copyright MIT
Persistent DSpace Link
