Thank you for your quick reply. Yes, I’ve tried the GitHub repository (both the main and refactor branches) to build a new index with all DBpedia labels. The problem is that the original DBpedia collection dataset is huge, and building a complete DBpedia index is resource-consuming. I’m encountering an out-of-memory issue when using the BUILD_DISK option. I am wondering if there is a pre-built DBpedia index available for download?
Hm, probably not. Foremost, we are working on getting the Webservice running. This will take a while since we are restructuring the infrastructure to make it more stable (Kubernetes and such). If it is up again, then you can hammer it with as many requests as it can handle.
Then we also need to work on the refactor branch. In the future, we plan to have a jenkins, where the community can add such build jobs via docker to tackle gaps like missing pre-built indexes.
You could create a new topic in the forum kindly asking someone from the community to build it for you. I don’t know exactly, but maybe 32 or 64gb ram are enough.
I ran into the same issue even with a strong machine. I added a new config variable called commitInterval that let’s you tweak the amount of writes that can be made until a commit (write to disk) to the index is forced. This will clear up memory which prevents the issue of running out.
I have also built a new lookup index. It still needs testing but once that is done, I will publish it via the Databus and add docu to run it f rom there.
Is there a way I can contribute to help get the app running? I have some capacity to volunteer. I hold a PhD in KG reasoning and have some experience with DevOps, Kubernetes, and related areas.