Using Pretrained Models and Transfer Learning in Text Analysis

When i started fast a, the idea of using pretrained models and transfer learning in text analysis was almost entirely unused. The problem i decided to look at was predicting the sentiment of movie reviews. Is this a positive or a negative movie review? After movie reviews from i m d b, this was a problem that had been considered extremely challenging because the movie reviews are a kind of long. There about two or three thousand words. And so this was a task that was considered superhard. Much to my delight, i found that the very first time i tried using computer vision transfer learning techniques on it, i actually broke the state of the art,. which means that i got a better accuracy than