We are happy to announce that our paper “Lifecycle of Semantic Web Machine Learning Systems” has been accepted at the DEXA workshop “Machine Learning and Knowledge Graphs”- MLKGraphs 2022.
Following up on our research on Semantic Web Machine Learning Systems (SWeMLS) which combine a symbolic and subsymbolic component, our paper investigates the combined lifecycle of such systems.
Until now, most hybrid systems are looked at from one angle (ML or SW) and lifecycles are mostly designed for one field, which is also due to the specialization of the fields. However, with the growth of hybrid systems and increased needs for traceability and auditability a holistic overview is needed. Therefore, we introduced a framework based closely on the widely-accepted CRISP- DM combining ML and SW to provide a unified representation and a communication means for interaction points on an architectural level. The overall SWeMLS lifecycle connects the dedicated ML and SW lifecycle and then both can be drilled down for the specific activities executed for development and operation of such resources (cf. Figure 1 for SW lifecycle).
We then use the framework to describe three reported cases of hybrid systems to show the applicability of our model.
This framework is a key deliverable of the OBARIS project (Deliverable 4.2 – Blocks and Functions of Semantic AI).
We are looking forward to attending DEXA in person again in Vienna after only online conferences for quite some while now.
Fig. 1: Semantic Web component lifecycle view