AI Reading, or Automatic Semantic Decomposition into Knowledge Graphs and Symbolic reasoning through Marker Passing

Abstract

Marker passing algorithms have been applied to solve problems in artificial intelligence related to the semantic of written words. Such approaches could a also prove to be useful in digital forensics, e.g., to reduce the effort of extracting evidence from confiscated data. We call creating semantic or even pragmatic understanding of text: AI reading. With this, we show that the aggregation of knowledge out of heterogeneous information sources can be a combination of symbolic and connectionist approaches. With that, the extraction of knowledge graphs can be automated. This approach has the benefit, when used correctly, that both the creation and the use of the knowledge graph through Marker Passing stay explainable. In this paper we describe a tool chain of a Marker Passing approach from the point of view of digital forensics and discuss challenges and opportunities arising from the application of such an approach.