A Document-based Knowledge Discovery with Microservices Architecture

TL;DR

This paper presents a microservices-based approach for knowledge discovery from digitized organizational data, focusing on keyword extraction, document similarity, and natural language queries, with an emphasis on scalability and resilience.

Contribution

It introduces a novel microservices architecture for document-based knowledge discovery, including design guidelines and an implementation demonstrator.

Findings

The approach effectively extracts keywords and calculates document similarities.

The system supports natural language queries and language-independent information provision.

The demonstrator meets scalability and resilience requirements.

Abstract

The first step towards digitalization within organizations lies in digitization - the conversion of analog data into digitally stored data. This basic step is the prerequisite for all following activities like the digitalization of processes or the servitization of products or offerings. However, digitization itself often leads to 'data-rich' but 'knowledge-poor' material. Knowledge discovery and knowledge extraction as approaches try to increase the usefulness of digitized data. In this paper, we point out the key challenges in the context of knowledge discovery and present an approach to addressing these using a microservices architecture. Our solution led to a conceptual design focusing on keyword extraction, similarity calculation of documents, database queries in natural language, and programming language independent provision of the extracted information. In addition, the conceptual design provides referential design guidelines for integrating processes and applications for semi-automatic learning, editing, and visualization of ontologies. The concept also uses a microservices architecture to address non-functional requirements, such as scalability and resilience. The evaluation of the specified requirements is performed using a demonstrator that implements the concept. Furthermore, this modern approach is used in the German patent office in an extended version.