Live data from Hacker News

Show HN: Subreddit Finder - Trained on 4M Reddit Posts from 4K Subreddits

valohai.com

51–60 of 82 posts

Re: Show HN: Subreddit Finder - Trained on 4M Reddit Posts from 4K Subreddits

#51
post #21

Tried stocks, stock options, investing - all kept giving Robinhoodpennystocks as the top option. Not sure if the model is not fully trained? What are some examples where the model does recommend meaningful things?

Here were my two experiences, one I felt would be easy and the other hard: Title: Build recommendations. Message: "I'd like to upgrade some components. My current rig has an old i7 and an RTX 2060. Looking for something midrange that can handle modern games at high settings (but maybe not ultra)." Matches: Nvidia (19%), IndianGaming (8%), GamingLaptops (8%), pcgamingtechsupport (6%) Title: Travel advice. Message: "I'…

Thanks so much for your detailed feedback. There is definitely something going on with the UK thing... Need to dig deeper!

Re: Show HN: Subreddit Finder - Trained on 4M Reddit Posts from 4K Subreddits

#52

Anyone remember /r/reddit.com? That was around the time reddit looked like Hacker News and people were embarrassed to admit they use it.

Thanks for reminding me, I just found gold here.

* After 5 years of surfing reddit, these are my favorite discoveries...

https://old.reddit.com/r/reddit.com/comments/guktv/after_5_y...

A collection of science and computing related links.

Re: Show HN: Subreddit Finder - Trained on 4M Reddit Posts from 4K Subreddits

#53

I tried it with "best time tracking app for iOS?" and "I'm looking for a time tracking app. Any recommendations?" I expected the iPhone or iOS subreddit to be suggested, but it suggested GearVR | 13.0%, ringdoorbell | 9.0%, canadacordcutters | 5.0%, TTVreborn | 5.0%, AusSkincare | 4.0%, sideloaded | 4.0%, FlutterDev | 2.0%, shopify | 2.0%, weightwatchers | 2.0%, crossfit | 2.0%. Congrats on the attempt but it does st…

Weird! I just tried "best time tracking app for iOS?" and got "iOSProgramming | 45.0 % swift | 18.0% ProtonMail | 14.0% ios | 4.0% apple | 3.0% iphone | 3.0% FlutterDev | 2.0% freebies | 1.0% jailbreak_ | 1.0% jailbreak | 1.0%"

For the second case: "I'm looking for a time tracking app. Any recommendations?" and got "GearVR | 43.0% TTVreborn | 14.0% WearOS | 4.0% OculusGo | 3.0% IPTV | 3.0% ApksApps | 2.0% animepiracy | 2.0% RabbitReddit | 2.0% NetflixViaVPN | 1.0% androidapps | 1.0%"

Which is a bit better! But still not perfect ;)

Re: Show HN: Subreddit Finder - Trained on 4M Reddit Posts from 4K Subreddits

#54
post #53

I tried it with "best time tracking app for iOS?" and "I'm looking for a time tracking app. Any recommendations?" I expected the iPhone or iOS subreddit to be suggested, but it suggested GearVR | 13.0%, ringdoorbell | 9.0%, canadacordcutters | 5.0%, TTVreborn | 5.0%, AusSkincare | 4.0%, sideloaded | 4.0%, FlutterDev | 2.0%, shopify | 2.0%, weightwatchers | 2.0%, crossfit | 2.0%. Congrats on the attempt but it does st…

Weird! I just tried "best time tracking app for iOS?" and got "iOSProgramming | 45.0 % swift | 18.0% ProtonMail | 14.0% ios | 4.0% apple | 3.0% iphone | 3.0% FlutterDev | 2.0% freebies | 1.0% jailbreak_ | 1.0% jailbreak | 1.0%" For the second case: "I'm looking for a time tracking app. Any recommendations?" and got "GearVR | 43.0% TTVreborn | 14.0% WearOS | 4.0% OculusGo | 3.0% IPTV | 3.0% ApksApps | 2.0% animepiracy…

In my case the first sentence was the title and the second the body; I guess having the 2 together might’ve thrown it off. I didn’t know only one of the fields was actually required.

Re: Show HN: Subreddit Finder - Trained on 4M Reddit Posts from 4K Subreddits

#55
post #53

Earlier quoted context omitted.

Weird! I just tried "best time tracking app for iOS?" and got "iOSProgramming | 45.0 % swift | 18.0% ProtonMail | 14.0% ios | 4.0% apple | 3.0% iphone | 3.0% FlutterDev | 2.0% freebies | 1.0% jailbreak_ | 1.0% jailbreak | 1.0%" For the second case: "I'm looking for a time tracking app. Any recommendations?" and got "GearVR | 43.0% TTVreborn | 14.0% WearOS | 4.0% OculusGo | 3.0% IPTV | 3.0% ApksApps | 2.0% animepiracy…

In my case the first sentence was the title and the second the body; I guess having the 2 together might’ve thrown it off. I didn’t know only one of the fields was actually required.

Good feedback. Indeed, it is not clear on the UI. The model concatenates the two, so not sure if changing the label "Message" to "Text" to mimic Reddit UI for text posts would be enough... Or rather just have a single "Text" textarea as for the model is the same.

Re: Show HN: Subreddit Finder - Trained on 4M Reddit Posts from 4K Subreddits

#56
post #51
post #21

Earlier quoted context omitted.

Here were my two experiences, one I felt would be easy and the other hard: Title: Build recommendations. Message: "I'd like to upgrade some components. My current rig has an old i7 and an RTX 2060. Looking for something midrange that can handle modern games at high settings (but maybe not ultra)." Matches: Nvidia (19%), IndianGaming (8%), GamingLaptops (8%), pcgamingtechsupport (6%) Title: Travel advice. Message: "I'…

Thanks so much for your detailed feedback. There is definitely something going on with the UK thing... Need to dig deeper!

Thanks for making it, super cool! Btw, if it wasn't clear, my "controversial" comment isn't to be taken seriously

Re: Show HN: Subreddit Finder - Trained on 4M Reddit Posts from 4K Subreddits

#57
post #26

Cool. Last year I created something like this as a Chrome extension so that you could type in your post and it would show up on reddit where to post. You could then just select it by clicking a link. Project is here https://github.com/wesbarnett/insight

Nice work @exegete! The Chrome extension idea is great ;) Nice to also see some metrics comparison. I will review your work. Looks like you achieved 0.6 accuracy with 600 classes. I got 0.4 f1-score on 4000 classes, but I have a ton of posts and subreddits with images and no text ;) For this case, it is also nice to report Recall@k. The current model has Recall@5 of 0.6. Meaning that on the test dataset, 60% of the time the human choice is within the first 5 suggestions. Currently, it is not supposed to automatically post but help the user discover new subreddits :)

I will probably retrain it on more subreddits, and fine tune a few things.

Re: Show HN: Subreddit Finder - Trained on 4M Reddit Posts from 4K Subreddits

#58

Anyone remember /r/reddit.com? That was around the time reddit looked like Hacker News and people were embarrassed to admit they use it.

I for one was embarrassed to admit I used reddit long after /r/reddit.com was defunct.
Post reply on HN