Saw this article from 2018 but curious if there are any other (fully) reproducible studies from the last few years?
2 Likes
Hey Sophia,
I did a blog posts just a few days ago that does Named Entity Linking and Relationship Extraction before storing results as a knowledge graph:
https://towardsdatascience.com/from-text-to-knowledge-the-information-extraction-pipeline-b65e7e30273e
2 Likes
There has been substantial research in this area for over 20 years which has led to a lot of systems.
Search for “ontology learning from text”
Wikipedia page: Ontology learning - Wikipedia
ACM Survey papers are always a great way to start:
https://dl.acm.org/doi/10.1145/2333112.2333115
In the past years, this type of work is core to the Automated Knowledge Base Construction community. AKBC 2021
These should be great starting points
2 Likes
Awesome! Thank you. Separately, I’ve been told that completely automated development of ontologies is an ongoing area of research, and it is not something that AI can handle right now because it requires understanding of common sense facts that the best examples are still far away from solving.
Not sure if that was particularly pessimistic or an accurate state of the world/field. Will take a look at the survey paper from 2012 and the AKBC community to get a current grasp of the field.
What are your thoughts? I’ve seen some enterprise examples of automated knowledge graph creations. like DiffBot, but not sure how “well-designed” those ultimately area.