@NatureComms

@NatureComms

Most recents (8)

Matthias Samwald (@matthiassamwald@sigmoid.social)

We suffered through curating and analysing thousands of benchmarks -- to better understand the (mis)measurement of AI! 📏🤖🔬

We cover all of #NLProc and #ComputerVision.

Now live at @NatureComms! nature.com/articles/s4146…

1/

Benchmarks are crucial to measuring and steering AI progress.

Their number has become astounding.

Each has unique patterns of activity, improvement and eventual stagnation/saturation. Together they form the intricate story of global progress in AI. 🌐
2/

We found a sizable portion of benchmarks have kind of reached saturation ("can't get better than this") or stagnation ("could get better, but we don't know how / nobody tries"). But still a lot of dynamic benchmarks as well!
3/

Read 9 tweets

Raymond D Sims

@rsims

I fell into the #KnowledgeGraph (KG) and Personal Knowledge Graph (PKG) rabbit hole. After too many hours of reading and watching, I'm now smarter than before; however, I am still learning.

Some stops along the way:
⬇️

Towards Definition of Knowledge Graphs ceur-ws.org/Vol-1695/paper… resonated re "Graph-based knowledge representation has been researched for decades and the term knowledge graph does not constitute new technology. Rather, it is buzzword reinvented by Google and adopted by other[s]"

However, Ehrlinger and Wöß's definition "A knowledge graph acquires and integrates information into an ontology and applies a reasoner to derive new knowledge." lands flat.

Read 9 tweets

Kingsley Uyi Idehen

@kidehen

@timberners_lee

You can't keep a good thing down.

#LinkedData is a classic example, despite confusion that has swirled around it for years.

I was watching @timberners_lee's @TEDTalks presentation on the subject earlier today, and it has aged very well.

#Web30

🧵

@YouTube

#LinkedData principles add a powerful dimension to #StructuredData creation, courtesy of a #Hyperlink functioning as an unambiguous Entity Name.

Here's a @YouTube video titled "How the hyperlink changed everything" about this "deceptively simple" tool.

Once you the power of #LinkedData settles in, its importance to #KnowledgeGraph creation becomes clearer.

Here's a nice explainer video.

#SemanticWeb #Web30

Read 7 tweets

Gaurav Ketkar

@ketcomp

What is common between product docs, Slack channels, and Wikis? Why, everyone ignoring all of those and instead asking the subject matter expert the same question over and over again, of course!

A 🧵on the next big thing in tech: effective knowledge management: #knowledgeispower

@SlackHQ

1/ It is no exaggeration when @SlackHQ calls itself the Digital HQ of a company - our chat apps accrue staggering amounts of information. This information can be tapped to unlock tremendous value! #chatops #KnowledgeManagement #projectmanagement

2/ Answers v. Information: Answers are often derived by putting together information from multiple sources. These sources are often not well documented or rely on the user knowing the system to find them. New hires are disadvantaged. #information #KnowledgeGraph

Read 6 tweets

Michael Galkin

@michael_galkin

@michael_galkin

We are glad to announce that our paper "Message Passing for Hyper-Relational Knowledge Graphs" by @michael_galkin @shape_mismatch @__gauravm @Ricardo_Usbeck @JLehmann82 has been accepted at @emnlp2020! 🎆 We propose a #GNN architecture for hyper-relational KGs like @wikidata. ⬇️

Traditional KGs are based on triples, whereas new KGs like #wikidata use statements and qualifiers to instantiate each edge further making the graph hyper-relational (img1). We incorporate these qualifiers by modifying existing multi-relational GNN (CompGCN) in the StarE (img 2).

This improves link prediction performance across the board!

Read 7 tweets

M. Lissandrini

@Kuzeko

https://twitter.com/Kuzeko/status/1287826804162605056

Q) What is #KnowledgeGraph #Exploration?

A) a machine-assisted process of analysis of a KG to: (1) understand its structure (2) identify whether it can satisfy the current information need and (3) locate the portion of the KG pertinent to the current need

1/4

https://twitter.com/Kuzeko/status/1287826804162605056

In this work we divide techniques in:
a) Summarization/Profiling
b) Exploratory Analytics
c) Exploratory Search

These 3 are mapped on a spectrum from no interaction and no requirement to domain knowledge, to high interaction and higher requirement of domain knowledge.

2/4

Less interactive methods require less domain knowledge, the system is doing all the work given some general definition of "interestingness". Yet, they also produce less detailed answers.
More interactive approaches require more domain knowledge, but can provide more details.

3/4

Read 5 tweets

Kingsley Uyi Idehen

@kidehen

#LinkedData is crucial to creating a #KnowledgeGraph that manifests as #SemanticWeb, period.

The #Web in all of this is about a #Hyperlink-based Entity Relationship Graph that's navigable by both humans and machines using de-reference (i.e., lookups).

@OpenLink

@OpenLink has opened up a new community forum channel focused on #LinkedData with the sole aim of teaching anyone that's interested about this powerful concept via live examples.

See: community.openlinksw.com/c/linkeddata

#GraphDatabase #KnowledgeGraph #SemanticWeb
#CDO #DataOps

Actually, here is a very basic depiction of an Entity Relationship Graph constructed in line with #LinkedData principles.

#GraphDatabase #KnowledgeGraph #CDO #DataOps

Read 3 tweets

Kingsley Uyi Idehen

@kidehen