Document Capturing Automation: Case Study

• Published in: Baltic Journal of Modern Computing Vol. 8; no. 2; pp. 259 - 274
• Main Authors: Rāts, Juris, Pede, Inguna
• Format: Journal Article
• Language: English
• Published: Riga University of Latvia 2020
• Subjects: Algorithms; Automation; Classification; Datasets; Deep learning; Discriminant analysis; Enterprise content management; Machine learning; Neural networks; Performance evaluation; Principal components analysis; Process parameters; Recommender systems; Sentiment analysis; Support vector machines; User interface
• ISSN: 2255-8950; 2255-8942; 2255-8950
• DOI: 10.22364/bjmc.2020.8.2.04

Capturing of new documents into the enterprise document repository involves some kind of (usually labour intensive) task of metadata attributing. Attributes include document types, folders, cases and others. This is, in fact, a classification process and may be addressed by Machine Learning based classification methods. The research presented in this paper aims to introduce a flexible framework for document capturing automation employing Machine Learning based document classification bots and providing an intuitive user interface featuring analysis of the document classification performance, creation of the domain specific rules, and configuration or the process parameters. The paper outlines important aspects of the framework - channels, rules, handling imbalanced data and continuous training. The prototype DOBO of the framework is presented and its performance evaluation results are discussed.