Projectdetails
Runtime DiCoMe: 01.05.24-31.05.24
Runtime PreDiCoMe: 01.10.23-31.01.24
Funded by: FFG Innovationscheck AI.engineer
The aim was to develop a knowledge system that abstracts specialist documents in a knowledge graph and acts as an intelligent advisor for experts.
The predecessor project PreDiCoMe was about laying the foundations for an advanced knowledge system. The focus is on developing a methodology for the detailed linguistic analysis of specialist documents in order to use this information effectively in DiCoMe.
The basis of the knowledge system is a graph-based approach that semantically captures the terms and statements of a text and transparently depicts their relationships. For this purpose, underlying linguistic information is extracted from text corpora using symbolic and non-symbolic AI methods. Methods such as dependency graphs, entity disambiguation and semantic role labeling were used to obtain this information. Research was then carried out into how they can be used to build a knowledge graph that transparently maps the content of the texts. In an iterative process, rules were developed for a software prototype that generates the graph step by step.
Projectdetails
Runtime DiCoMe: 01.05.24-31.05.24
Runtime PreDiCoMe: 01.10.23-31.01.24
Funded by: FFG Innovationscheck AI.engineer
Demonstrator: PreDiCoMe: Software-Prototype for creating knowledge-graphs in natural language
The aim was to develop a knowledge system that abstracts specialist documents in a knowledge graph and acts as an intelligent advisor for experts.
The predecessor project PreDiCoMe was about laying the foundations for an advanced knowledge system. The focus is on developing a methodology for the detailed linguistic analysis of specialist documents in order to use this information effectively in DiCoMe.
The basis of the knowledge system is a graph-based approach that semantically captures the terms and statements of a text and transparently depicts their relationships. For this purpose, underlying linguistic information is extracted from text corpora using symbolic and non-symbolic AI methods. Methods such as dependency graphs, entity disambiguation and semantic role labeling were used to obtain this information. Research was then carried out into how they can be used to build a knowledge graph that transparently maps the content of the texts. In an iterative process, rules were developed for a software prototype that generates the graph step by step.
