I am trying to build a real-time sentiment analysis project on memes.
I have laid out my plan as follows:
- Getting images from Instagram in real-time using Instaloader (anyone knows for reddit or twitter?). I’ll apply an OCR and extract the text from the image.
- Using a CNN+RNN architecture, I’ll generate captions for the meme.
- Using the extracted text and generated captions, I’ll feed these texts to a sentiment analysis script which will output me the polarity of the sentence (or to say the meme).
Is this idea workable/implementable? I am a beginner, so feel free to add something that would work out much better.