As someone who has spent a <i>lot</i> of time working with text-generating neural networks (<a href="https://github.com/minimaxir/textgenrnn" rel="nofollow">https://github.com/minimaxir/textgenrnn</a>), I have a few quick comments.<p>1) The input dataset from Memegenerator is a bit weird. More importantly, <i>it does not distinctly identify top and bottom texts</i> (some have a capital letter to signifify the start of the bottom text, which isn't always true). A good technique when encoding text for these types of things is to use a control token (e.g. a newline) to indicate these types of behaviors. (the conclusion notes this problem: "One example would be to train on a dataset that includes the break point in the text between upper and lower for the image. These were chosen manually here and are important for the humor impact of the meme.")<p>2) The use of GLoVe embeddings don't make as much sense here, even as a base. Generally the embeddings work best on text which follows real-world word usage, which memes do not follow. (in this case, it's better to let the network train the embeddings from scratch)<p>3) A 512-cell LSTM might be too big for a word-level model of that size; since the text follows rules, a 256-cell Bidirectional might work better.
Very silly; best not to alert the media or we'll soon see "AI can now generate memes" clickbait.<p>I thought it was funny though that Richard Socher, one of the authors of GLoVe and NLP researcher is pictured in the generated memes on p. 8. ("the face you make when")
Full paper with some examples here: <a href="https://web.stanford.edu/class/cs224n/reports/6909159.pdf" rel="nofollow">https://web.stanford.edu/class/cs224n/reports/6909159.pdf</a>
I was expecting this to use some formats that aren't from 2012. It would be interesting to see a neural network that could decide text for more complex meme formats that trend on twitter and instagram.
Reminds me a bit of <a href="https://www.reddit.com/r/SubredditSimulator/" rel="nofollow">https://www.reddit.com/r/SubredditSimulator/</a>
It looks like a joke now but I'm fairly convinced that in the not too distant future the most influential social media accounts will be run by some kind of AI.
All their generated examples look like Markov chain generated captions. Pretty random and generally unfunny. I completely disagree with the claim that you can't differentiate between these generated memes and real memes. None of these would make the front page of reddit, for example.