I've built a Bluesky feed that uses machine learning to find data visualisations on Bluesky. It's called DataBot. #dataviz bsky.app/profile/did:... - ThreadSky

olihawkins.com • 314 days ago

I've built a Bluesky feed that uses machine learning to find data visualisations on Bluesky. It's called DataBot. #dataviz

https://bsky.app/profile/did:plc:aljpcrrf7qt6lhlnumjum6oz/feed/databot

Reposted from /home/cyhaxztpeu/threadsky.app/post.php on line 479

Warning: Trying to access array offset on value of type null in /home/cyhaxztpeu/threadsky.app/post.php on line 479

Deprecated: urlencode(): Passing null to parameter #1 ($string) of type string is deprecated in /home/cyhaxztpeu/threadsky.app/post.php on line 479
">
Warning: Undefined array key "author" in /home/cyhaxztpeu/threadsky.app/post.php on line 480

Warning: Trying to access array offset on value of type null in /home/cyhaxztpeu/threadsky.app/post.php on line 480

Deprecated: htmlspecialchars(): Passing null to parameter #1 ($string) of type string is deprecated in /home/cyhaxztpeu/threadsky.app/post.php on line 480

Comments

pixeline.be•314 days ago

Amazing! how did you do it ? I only use @skyfeed.xyz for my custom feeds. Would you consider writing a tutorial ? Or any that you could recommend ?

olihawkins.com•314 days ago

Thanks! It’s a custom application that I built from scratch. To be perfectly honest, it was a bit of a challenge getting all of the different bits of this to work together while keeping the compute costs down. I am going to write it up soon. https://bsky.app/profile/olihawkins.com/post/3kqdvlnx4wa2j

pixeline.be•314 days ago

Thanks! looking forward to reading it :)

pgov.bsky.social•313 days ago

Great idea! Subscribed

olihawkins.com•313 days ago

Thank you!

paulmpax.bsky.social•312 days ago

Thanks Oli this is a v useful feed

olihawkins.com•312 days ago

Thanks!

tinarussell.bsky.social•314 days ago

@scientificdiscovery.dev This seems to me like something you might like ☺️

scientificdiscovery.dev•314 days ago

Ooh thank you!

manos.lol•312 days ago

📌

olihawkins.com•312 days ago

Thanks!

jabsco.fbi.cia.fyi•314 days ago

Is your source code available?

olihawkins.com•314 days ago

Not right now but I will probably make the repo public soon. I want the freedom to play around with the implementation in private for a few more weeks now that the feed is public. I’ll do a full write-up soon, which will look at model architecture and compute costs and I’ll post the link here.

olihawkins.com•314 days ago

It's aimed at people who are interested in data viz, either for its own sake or as a way of finding interesting data stories.

olihawkins.com•314 days ago

Right now, DataBot looks for top-level posts tagged as english language that have images that (probably) contain data visualisations. I might relax those criteria over time, but that's how it currently works.

olihawkins.com•314 days ago

This is an interesting computational challenge. The feed is being hosted on a VPS and inference is being done on the CPU. The challenge is to make a model that is small, fast and accurate. It does okay in this respect but could be improved.

olihawkins.com•314 days ago

There are likely to be teething issues and I may need to adjust the implementation as I figure things out. I am still tuning the model to work well with Bluesky data in particular.

olihawkins.com•314 days ago

I will do a more detailed write-up on my blog soon, but for now please like and subscribe if this is your sort of thing.

olihawkins.com•314 days ago

Big thanks to @aendra.com who gave me loads of help with the AT protocol.

Comments

Posting Rules

Reply