An Introduction to Knowledge Graphs | Hacker News Reader

An Introduction to Knowledge Graphs

textmine.com

69 pointslegislate34 comments

Custom JS scrolling in this day and age is a HUGE no-no. Just don't.

I was wondering what was going on. I thought my coffee hadn't kicked in.

Agreed. Came here to say this. Yuck.

I immediately hit the reader-mode button once it started hopping around on its own. Why are they doing so much work just to make things worse?

The scroll on this is horrible. Really hard to read.

Why do people keep doing this? At this point I consider it a sign they'd rather be fancy than usable, which I can only imagine must be how they write their other software as well.

Is there a better alternative? Does something like a wordpress theme support proper scrolling out of the box? I am working on a blog and know next to nothing about this :)

You don't need "support" for scrolling. It is supported by the browser, out of the box. The website hijacks scrolling with JS by introducing "smooth" scroll, which is a big no from the UX perspective.

What they think is "smooth" then feels like "my finger is wet and the screen isn't properly reading swipes" on a phone.

Your readers will have picked a web-browser that supports scrolling the way they like it. The less you try to interfere with the way people choose to visit your webpage the better.

I know next to nothing about websites. Sorry if this is obvious, but what do you mean by that? I was able to scroll and read it fine on an iOS device with Safari. What is the alternative or best practice nowadays for scrolling websites?

The website is using a javascript library to override the default system scrolling behavior. It creates what feels like, at least to me and some others in these comments, an unresponsive scroll animation that does not match the behavior of essentially all other apps on my device. This is called scrolljacking, and is generally discouraged as it tends to ignore the behavior the user has configured at the device or OS level.

Are there any resources on using knowledge graphs and ontologies for RAG?

Tony Seale has an excellent series of articles on Knowledge Graphs and AI here: https://experiencestack.co/embrace-complexity-part-1-39483f1...

Thanks! This was an interesting read.

I was hoping for something a bit more in-depth, though.

This was more in the area of stating/explaining some basic (and useful) concepts.

Knowledge graphs sound like the way you represent knowledge in Prolog and Datalog...and so nothing particularly new. Am I wrong?

Knowledge graphs have been around for a long time (early 70s). You could even argue they're older than that when considering something like an ontological map. I think this post is not meant to demonstrate novel technology or structure; rather, it's simply a self-reported introduction to the topic space.

Can you share some insights into how Datalog does it?

I'm not an expert; but both Datalog and Prolog make use of Horn clauses, e.g.,

``` person(mary) person(joe) parent(joe,mary) child(c,p) :- parent(p,c) ```

(Don't quote me on the syntax.) But you can specify what the OP calls "triples" directly in Datalog, express logical relationships among them, and then query them.

Felt a bit too AI generated.

It's more of an advert highlighting the various use cases for knowledge graphs than an actually intellectually stimulating introduction

Note that the poster only ever posts links to ads on their website.

Thanks, negbot 4000, your own comment added to the discourse. You have received today's ultra beige award.

I just felt that the article does not meet my expectation of an HN entry. No hard feelings. Try again soon OP.

Yeah, I soon as I read “in a world where…” I just stopped reading, it was obviously aI made, I like ai just as much as the next guy, but it’s hard for me to force myself to read stuff that I know has been generated

a thinly veiled advert. yuk

I was surprised the highjacked scrolling didn't slide me right into a full-page modal newsletter signup form.

What they are and how they can accelerate LLM training

"Training" appears once on the page, in the following paragraph:

Generative AI models such as large language models typically leverage huge amounts of unstructured data during their initial training and fine-tuning. Knowledge graphs can be leveraged to identify the structure of the underlying data which can then improve the amount of signal which is extracted from the data during the training.

Which is too vague to be of interest, generally uninformative, and furthermore to me reeks of LLM-generated babble instead of something a human wrote.

That point is also a bit of a stretch.

While there are a few papers on using KG/Ontologies to enhance training, this is really far from the mainstream, and i would be surprised if it would be used anywhere (outside of a research paper)

Statements and axioms. Merge two knowledge graphs, what can you reason about them?

IE people and obesity, people and mortality. Can you find correlations between the two?

The next layer up is to prove the trust in the data and your insights.

Also you are a decade late, google already bought freebase and made schema.org

Yes there was once a hype about these graphs and standards so called Resource Description Framework (RDF). It never took off. The knowledge itself requires hefty curation though, if it's in a field that requires low false positive rate such as pharmacology.

Not my area at all, just passing by and wondering to the extent and how they use knowledge graphs for drug discovery.

Some time back I had a peek at AstraZeneca's GitHub [0] and got me curious. I know in genomics they try to use custom hardware to accelerate the process using FPGAs and others [1].

Curious if anyone can shed light on knowledge graph use at scale is being accelerated.

[0] AstraZeneca; Awesome Drug Discovery Knowledge Graphs https://github.com/AstraZeneca/awesome-drug-discovery-knowle...

[1] Gene sequencing accelerates with custom hardware https://www.mewburn.com/news-insights/gene-sequencing-accele...

I only know in my area you can infer gene and protein interaction network using knowledge graphs to some degree. In drug discovery I've seen graphs of chemical knowledges and what can be drugged and what not. https://pubs.acs.org/doi/10.1021/acs.jpca.2c06408

There might be more publications on mining literature and building such graphs but I'm not following it much since deep learning took over.

Reply on news.ycombinator.com