Enterprise Knowledge, LLC’s Post

The most common feature request we received in May: Mapping 2 taxonomies together ✨ We'll help you build a many-to-many mapping between taxonomy A and taxonomy B based on lexical and semantic similarity. What that means for our users: When your classifier labels your text, we'll also map it to your secondary taxonomy. And if there are multiple mapped labels, your models will pick the best option.

We aim nowadays to use more and more semantic data, and for that we need ontologies, taxonomies, schemas, etc. These can be simply called semantic artefacts. But how well these semantic artefacts are reposited? What are the emerging qualities of catalogues for FAIR semantic artefacts? In this paper, we present a maturity model for assessing catalogues of semantic artefacts, one of the keystones that permit semantic interoperability of systems. https://lnkd.in/gCEyhe6H

We have built a new service which allows anyone to see who has cited data held in the Environmental Data Service (EDS). This is possible if a dataset has been published with a Digital Object Identifier (DOI) by any EDS data centre. This new development allows anyone to find how many citations and links are connected to the DOI’d resource - allowing greater understanding about who is re-using archived data for the first time.  https://lnkd.in/eYhg2RZ3 NERC: Natural Environment Research Council GEBCO - General Bathymetric Chart of the Oceans DataCite

Interesting view posted on MIT Techology Review on storing digital information for the future, questioning what to save and how to save it. #piql #piqlfilm #arcticworldarchive

"In applications where categorical labels follow a natural hierarchy, classification methods that exploit the label structure often outperform those that do not.... we investigate a class of methods that induce a hierarchy that can similarly improve classification performance..." https://lnkd.in/evHZgMq5 #clustering #hierarchy #classification #computervision #statistics #bayesian

The article discusses advancements in Retrieval-Augmented Generation (RAG) systems, which have become increasingly popular for generating informed and accurate responses by retrieving and utilizing external knowledge, especially in the form of knowledge graphs. The author examines the evolution from early RAG systems that used unstructured text to more sophisticated approaches involving structured knowledge representation, addressing the challenges and benefits of reasoning over complex multi-relational data at scale https://lnkd.in/gD2gvRP3

By definition, the ‘truly novel things’ exist outside (!) of the current data space. Truly new things get named novel words or existing words get a dramatically new meaning, hence existing outside of the current semantic space. They’re not a mixture of existing things or a new pattern.

Top Biggеst Tеchnological Blundеrs of 2023 https://lnkd.in/gpRdpgmM #biggеst #biggеsttеchnologicalblundеrsin #blundеrs #top #topbiggеsttеchnologicalblundеrsof #tеchnological

Knowledge graphs are pretty. Ontologies are valuable 🔆 People describe object types as the “nouns” of the Ontology, that’s apt but it leaves so much unsaid. By indexing data into the Ontology and digitizing a piece of your enterprise, you automatically get access to a rich set of building blocks for doing operationally useful things.

It's crucial to take due diligence while using cosine-similarity as distance metrics on embedding. A recent paper highlights why. It serves as a reminder to properly assess the model before deploying it to the real world. Check it out here: https://lnkd.in/gwZPqF2v