Fri, Dec. 8th, 2023

matt1993: (Default)
As I said in the second part of my first regular entry in a while (I think? What even is a "regular" entry in my journal anymore?), I think it's really, really, REALLY great that people are starting to research machine learning for detecting toxic comments online so that they can be filtered out automatically and/or so that people (like me!) can learn what makes their comments potentially toxic and how to avoid it.

You know what would be even better, though?

If I could actually use it.

Many of the websites and videos I've seen about these tools for detecting toxic comments make it seem like you can just type or paste in a comment and BAM, it will give scores representing the likelihood of it being a toxic comment for various potential reasons. So I thought "oh hey, I can paste in of my LJ/DW entries and comments one by one and it'll tell me which ones are the best and worst, and then I can rank them and sort them into 'cringe levels' like I wanted to a couple years ago!"

But finding a website that lets me do that is harder than I thought...
https://www.perspectiveapi.com/ - Has a note saying to not enter personally identifiable information, so naturally I'm wondering what exactly that entails and if certain things in some my entries count. Seems like this would be the easiest to use, though, if I knew for a fact that doing so would be okay. (EDIT: Well, I tried this one out with a random sentence that isn't from my journal and actually got a rating with it! I guess I'm gonna use this one after all - I'll just save entries with my name in them for last, I suppose. But it would be nice to figure out how to use one of these other tools too in case the scores differ.)
https://www.unitary.ai/product - This page has an interface where you can input your own text and hit "Compute" and it will estimate the toxicity of it... except not because the "interface" on that page is just a mockup apparently! They went out of their way to make it look like you could highlight the example text and replace it and click Compute but it doesn't actually let you do that? :(
https://detoxifai.com/ - This one has a live demo where you can type in a comment (yes, this one actually lets you type into it!) and click "Analyze Toxicity" and... it says "Processing" and then never does anything.

And pretty much everything else I've found requires me to install Python, learn how to code in Python, install Git (or remember whether or not my current laptop is the one I once had Git installed on before), relearn how to use Git when I never understood the first time, and/or learn how to make an API just to use it. :(

It's hard to describe this journal in just a few paragraphs... but I'll try

Just an autistic guy born in 1993 (hence the username) who's had this journal since I was 14. Has trouble phrasing things well especially in earlier entries, so keep that in mind.

My interests have fluctuated in the 17 years since I've had this journal, but the following has remained constant:
- I like Mario games (mainly the platformers, Mario Kart, Mario Party) and Super Smash Bros.
- I like Garfield
- I like Homestar Runner
- I like Enya
- And I've always loved the idea of wacky crossovers between ANY of my interests, whether it's in the form of a short-lived webcomic in 2006-2008, a short-lived backwards lyrics site in 2008, writing down weird dreams in 2009-2011, playing Tomodachi Life a lot in 2014-2016, or what most of my posts are about these days: playing around with Character.AI starting in 2023. I still can't believe I'm actually in a world where an AI of Enya can talk to AIs of Mario or Garfield or even my favorite obscure characters like Mary O. from Super Mario Maker. :O

Go to my sticky entry for more.

Expand Cut Tags

No cut tags

I Ramble Too Much About

Style Credit

Page generated Mon, Sep. 8th, 2025 13:09
Powered by Dreamwidth Studios
June 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 2025