Learning To Generate Reviews and Discovering Sentiment

Alec Radford, Rafal Jozefowicz, Ilya Sutskever

Abstract: We explore the properties of byte-level recurrent language models. When given sufficient amounts of capacity, training data, and compute time, the representations learned by these models include disentangled features corresponding to high-level concepts. Specifically, we find a single unit which performs sentiment analysis. These representations, learned in an unsupervised manner, achieve state of the art on the binary subset of the Stanford Sentiment Treebank. They are also very data efficient. When using only a handful of labeled examples, our approach matches the performance of strong baselines trained on full datasets. We also demonstrate the sentiment unit has a direct influence on the generative process of the model. Simply fixing its value to be positive or negative generates samples with the corresponding positive or negative sentiment.

TL;DR: Byte-level recurrent language models learn high-quality domain specific representations of text.

Keywords: unsupervised learning, representation learning, deep learning

Code: [ 2 community implementations](https://paperswithcode.com/paper/?openreview=SJ71VXZAZ)

Data: [IMDb Movie Reviews](https://paperswithcode.com/dataset/imdb-movie-reviews), [SST](https://paperswithcode.com/dataset/sst), [SST-2](https://paperswithcode.com/dataset/sst-2), [SUBJ](https://paperswithcode.com/dataset/subj)

Community Implementations: [ 2 code implementations](https://www.catalyzex.com/paper/learning-to-generate-reviews-and-discovering/code)