You are using an out of date browser. It may not display this or other websites correctly. You should upgrade or use an alternative browser.
Proving Ethan Writes Pantsu's Tweets and Texts Using Plagiarism Detection Algorithms - Using ghostdetect.com to perform gunt analysis on supposed 'May' posts
see you already walked back one part, before this time next week you'll be walking back the rest of your bullshit. hell you can use my posts in entirely different threads on here and it will break the machine. the "faces" will look more distinct than the ones between ralph and ralph in amanda mode. again its called being competent and not having whatever autism you clearly have
View attachment 3181880
no it stays pretty consistent. Unless you skipped over all the parts of the thread where I pointed out the importance of sample selection because you cant read.
Comparing Ralph saying "FUCK YOU BITCH, FUCK YOU FUCK YOU" in a single tweet isnt a good sample to compare to Ralph calmly typing an essay about why Jim needs to die of cancer.
I understand at this point your IQ is below the level where this all makes sense to you, I've accepted that fact. Just clarifying for readers.
Comparing Ralph saying "FUCK YOU BITCH, FUCK YOU FUCK YOU" in a single tweet isnt a good sample to compare to Ralph calmly typing an essay about why Jim needs to die of cancer.
I understand at this point your IQ is below the level where this all makes sense to you, I've accepted that fact. Just clarifying for readers.
like the others have said here, unless the readers are as autistic and stupid as you they'll understand how insanely idiotic this entire endeavor truly was on your part
like the others have said here, unless the readers are as autistic and stupid as you they'll understand how insanely idiotic this entire endeavor truly was on your part
Its like you and 2 people, the majority of poeple who've replied like the post nigga. What is this weird white girl social pressure game you keep flubbing over trying to use?
Facially blind autism moment....
The faces share the same features with slightly darker hair from greater line density and lower eyebrows. You've only confirm ghostdetect works.
Similarities: Hairline position (ish), eye shape, smile,
Differences: Hair density, ears, eyebrow thickness and position, eye distance and size, and nose
I'm curious how well it works when you don't already know the author of the tweets? More of a blinded test. I spent five minutes grabbing tweets that weren't just links or that didn't crash the ghostdetect site (it chokes on certain combinations of punctuation); they haven't been specially cherrypicked.
So, the question for each picture is: same or different author? Which ones look like Ralph tweets?
Similarities: Hairline position (ish), eye shape, smile,
Differences: Hair density, ears, eyebrow thickness and position, eye distance and size, and nose
You don't really build up an expected face for the writer, it compares similarities and non-similarities thats all. The faces are an algorithm for visualizing the similarities shared between two sets of multivariate data. Also if youre grabbing random tweets, youre likely to be comparing Ralph saying "FUCK YOU BITCH FUCK YOU" to an actually thought out post by someone which wont give meaningful results
You don't really build up an expected face for the writer, it compares similarities and non-similarities thats all. The faces are an algorithm for visualizing the similarities shared between two sets of multivariate data. Also if youre grabbing random tweets, youre likely to be comparing Ralph saying "FUCK YOU BITCH FUCK YOU" to an actually thought out post by someone which wont give meaningful results
this isn't as good as people think; you don't need fancy software to prove Ralph writes the tweets, only basic logic and social skills. Also as other people have mentioned you can use her own tweets pre and post gunting to prove her own writing style changes, for fucks sake we even know her kiwifarms accounts and suspected alts and they write in an unfathomably verbose style that would be inconceivably advanced, at least according to the algorithms presumably coded by some antidisestablishmentarianist wannabe hermaphroditic tranny. but those grade evaluators are extremely easy to game, and aren't exemplary models for examining the intelligence of the writings being put under the microscope. if they were we'd be teaching Russell Brand's books in high schools all over the country right now because of his thesaurus-loved passages like this one "This attitude of churlish indifference seems like nerdish deference contrasted with the belligerent antipathy of the indigenous farm folk, who regard the hippie-dippie interlopers of the denizens of the shimmering tit temples, as one fey step away from transvestites."
TL;DR you wasted your time showing off some bullshit algorithm. you can't rate writings when they're confined to 150 characters or less faggot.
At first I was hoping this was a troll-post by you. Then I read your other posts, and I fear that you are being serious in making this claim.
Read what you wrote, then read what your claim is. I doubt you will see the irony unless it is pointed out to you....
If you by "little lady" mean 13 year old girl, then sure. That would be your level. Suitable really, if you think about it.
So this begs the question; why are you so upset about this? Why are you so upset about an a random stranger's analysis on some text?
It extracts statistical features of the source text (average syllables per word, sentence length, composite features like estimated reading age etc.); the resulting vector is visualized using the faces technique which is easier for human comparison (we are notoriously poor at multivariate reasoning above 2 dimensions, there's a whole chapter in Kahneman's book going through the studies that have shown this in experts like doctors and stock traders who would swear blind they could reliably detect patterns in such data). This allows for more reliable detection of similarities and differences between the two underlying vectors, which would be missed if they were presented as raw numbers.
(actually, they don't describe their exact algorithm on the site and they have a big disclaimer that it's merely suggestive and cannot prove matching/distinct authorship, but we'll ignore that for now)
Now, if this relies on someone who (a) isn't blinded to the authorship of the tweets, (b) is motivated to only find results that support their thesis (that the test is useful, that Ralph=May) and (c) subjectively selects which snippets to use based on "similarity", not just on medium (tweet vs. article) but on the stylistic genre of the tweet itself, then I'm not surprised if you get the results you're trying to find. You could probably do the same thing reading tea-leaves.
All I'm trying to establish is whether the comparison is informative as to authorship when the person examining the faces doesn't already know the answer. It's a small sample but if the informative effect is large it'll do for now. If anything, it'll give people an intuitive idea as to how robust this technique across "different" tweets by the same author.
Take a minute and give some answers, see how it goes, then you can type out an explanation for why the results were/weren't what you'd expect.
Anyone else can have a go too. I'll post the uncropped answers (including the tweets) later.
It extracts statistical features of the source text (average syllables per word, sentence length, composite features like estimated reading age etc.); the resulting vector is visualized using the faces technique which is easier for human comparison (we are notoriously poor at multivariate reasoning above 2 dimensions, there's a whole chapter in Kahneman's book going through the studies that have shown this in experts like doctors and stock traders who would swear blind they could reliably detect patterns in such data). This allows for more reliable detection of similarities and differences between the two underlying vectors, which would be missed if they were presented as raw numbers.
(actually, they don't describe their exact algorithm on the site and they have a big disclaimer that it's merely suggestive and cannot prove matching/distinct authorship, but we'll ignore that for now)
Now, if this relies on someone who (a) isn't blinded to the authorship of the tweets, (b) is motivated to only find results that support their thesis (that the test is useful, that Ralph=May) and (c) subjectively selects which snippets to use based on "similarity", not just on medium (tweet vs. article) but on the stylistic genre of the tweet itself, then I'm not surprised if you get the results you're trying to find. You could probably do the same thing reading tea-leaves.
All I'm trying to establish is whether the comparison is informative as to authorship when the person examining the faces doesn't already know the answer. It's a small sample but if the informative effect is large it'll do for now. If anything, it'll give people an intuitive idea as to how robust this technique across "different" tweets by the same author.
Take a minute and give some answers, see how it goes, then you can type out an explanation for why the results were/weren't what you'd expect.
Anyone else can have a go too. I'll post the uncropped answers (including the tweets) later.
Thats a good summary of how the faces are generated but youre still misunderstanding the process fundamentally. What you're asking is getting really schitzo