Using RML for Knowledge Graph generation from IoT platform

Hello everyone,

I’m currently working on a research project where I plan to use RML to export knowledge graphs from different IoT platforms. My goal is not to write all the available data into a knowledge graph but rather only to convey the semantic information of the IoT system.

Thanks the RMLMapper, I've already conducted a proof of concept (PoC) using this approach and it seems to be feasible. However, I still have one question before scaling this up for further usage: What are the fundamental advantages of using RML to generate knowledge graphs? Or more specifically, how can the use of RML reduce manual effort while integrating heterogeneous data sources?

In my current workflow, I need to create a separate RML mapping file for each IoT platform, as the data models differ between platforms. To be honest, RML's syntax is somewhat unintuitive, and the mappings files are relatively verbose. As a result, creating each RML mapping file still requires significant manual effort and time, and the resulting mapping files can become quite large.

My concern is that, with the same time and effort, I can (actually I really did) write an ad-hoc mapping program that performs the same function for each IoT platform. Moreover, such a program might even be more compact and "readable" than the equivalent RML mapping file. Although I still believe RML offers a more elegant solution, I’m struggling to answer this question. I would greatly appreciate any insights or discussions on this matter!

Here is an example of my PoC.

source (8 KB): https://gist.github.com/djs0109/edffa7e19dc833f76247a67e816386a2
mapping file (14 KB): https://gist.github.com/djs0109/bea004feb28498bef8ec859186387a60
output (5 KB): https://gist.github.com/djs0109/92bc62b8b9f8f3cb20a4d430edf89735

RMLio / rmlmapper-java

Using RML for Knowledge Graph generation from IoT platform #242