KnowText: Auto-generated Knowledge Graphs for custom domain applications

Bojan Bozic, Jayadeep Kumar Sasikumar, Tamara Matthews

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

While industrial Knowledge Graphs enable information extraction from massive data volumes creating the backbone of the Semantic Web, the specialised, custom designed knowledge graphs focused on enterprise specific information are an emerging trend. We present “KnowText”, an application that performs automatic generation of custom Knowledge Graphs from unstructured text and enables fast information extraction based on graph visualisation and free text query methods designed for non-specialist users. An OWL ontology automatically extracted from text is linked to the knowledge graph and used as a knowledge base. A basic ontological schema is provided including 16 Classes and Data type Properties. The extracted facts and the OWL ontology can be downloaded and further refined. KnowText is designed for applications in business (CRM, HR, banking). Custom KG can serve for locally managing existing data, often stored as “sensitive” information or proprietary accounts, which are not on open web access. KnowText deploys a custom KG from a collection of text documents and enable fast information extraction based on its graph based visualisation and text based query methods.
Original languageEnglish
Title of host publicationiiWAS2021: The 23rd International Conference on Information Integration and Web Intelligence
EditorsEric Pardede
Pages350
Number of pages358
DOIs
Publication statusPublished - 1 Dec 2021

Keywords

  • Knowledge Graphs
  • Semantic Web
  • enterprise specific information
  • automatic generation
  • unstructured text
  • information extraction
  • graph visualisation
  • free text query
  • OWL ontology
  • business applications
  • CRM
  • HR
  • banking
  • sensitive information
  • proprietary accounts

Fingerprint

Dive into the research topics of 'KnowText: Auto-generated Knowledge Graphs for custom domain applications'. Together they form a unique fingerprint.

Cite this