Show HN: Pulpie – Models for Cleaning the Web

100 points
1/21/1970
2 days ago
by snyy

Comments


zaptheimpaler

So this is tailored towards kind of a "reader view" for models right? Can it handle images, tables, shadow DOMs too? Like there are 3 use cases I have now - one is a simple text view for models to understand it, one is a "web clip" mode which would ideally preserve images and media, and one is to extract tabular data from web pages. Which ones is this good at?

a day ago

snyy

Images pass through as they are considered main content. Same with tables.

Pulpie will return all main content on a page as HTML/Markdown. I’m not sure I fully understand “which one this is good at?”. perhaps you can try the model on hugging face and let me know if the results look good?

https://huggingface.co/spaces/feyninc/pulpie

a day ago

kocamaz

It's good looking, and I liked it. The trial page accessed from the hugging face website is a very inefficient experience when I use Mozilla and the dark theme, FYI.

a day ago

snyy

Fixed. Try again. Let me know if any other issues

a day ago

wiradikusuma

Does it work with ecommerce for product scraping? E.g. Amazon, or Shopee (big in SEA)

a day ago

snyy

Yes. I tried it with https://www.allbirds.com/products/womens-cruiser-canvas on our HF space and Pulpie worked great.

HF Space: https://huggingface.co/spaces/feyninc/pulpie

a day ago

danielmeskin

Well allbirds is an AI company now so I guess that makes sense

a day ago

andrethegiant

Why not use a plain old html → markdown converter? You can easily strip out ads using CSS /jQuery-like selectors. That would cost zero dollars.

a day ago

snyy

We see far better performance with models. Heuristics break on richer content like codeblocks, formulae, quotes, etc. In our testing, our model was 25 F1 points better than Trafilatura.

a day ago

andrethegiant

I think instead of "performance" you must mean "accuracy". Traditional deterministic conversion will always be faster and cheaper than running through a model, even if it is less accurate.

a day ago

snyy

Ah yes, I meant accuracy.

a day ago

spelk

If I had to reckon, it's because the web comes in very many shapes, and outsourcing that work to a generalist LLM/SLM like GPT Nano is expensive, and doing it deterministically will never catch all the edge cases as well as a purpose-built encoder when run at webscale.

a day ago

dracyr

Looks like they are including Trafilatura in the comparison tables, which I've used before with pretty decent results, but it still has trouble with some pages. Looks like the pulpie f1 scores are quite a bit better, especially for the hard cases.

Would be curious how it runs on more modest hardware though, I'm using it for a small bookmark archiving tool and being able to run it on my small mini-pc homelab would be nice.

a day ago

grillermo

I’m implementing this for readitsoon my web to kindle app. Thanks for this!

a day ago

geniium

Amazing I was just looking for something like this to be able to import web page content into Whisperit

a day ago

lnenad

Very nice! Thank you for building this.

2 days ago

emblemapp

This looks really cool

a day ago

cpill

I did some research on this about 10 years ago. I spent 2 days hand labelling data from scraped news sites. Then built a good old fashioned Random Forest model to classify html nodes based on some feature engineering. turns out the P tag and the number-of-words threshold get you 90% of the way there, on news sites anyway. Great thing about RF models is they tell you which features are the most important. fun little project (apart from the 2 days of data labelling).

a day ago

tyzoid

How does this work on pages that require JavaScript in order to render?

a day ago

philipkglass

You'd typically use a headless browser to generate the fully rendered page, then capture the rendered output for use with the model.

a day ago

snyy

Exactly this. Thank you for answering!

a day ago

esafak

Why does the 'Quality vs Cost of Web Content Extraction' chart not have zero cost at the origin? Up to the right does not have to mean better; we can read.

a day ago

snyy

Funnily enough, that wasn't my first choice either. I A/B tested it with a small group and people understood "up and to the right is better" faster.

a day ago

vishalkundar

Very interesting

a day ago