This repository provides code for "Learning to Generate Reviews and Discovering Sentiment," a research project focused on using language models for sentiment analysis and review generation. It is targeted at researchers and practitioners in NLP and sentiment analysis, offering a feature extraction mechanism for sentiment classification.

How It Works

The core approach utilizes a multiplicative LSTM language model, trained on a massive dataset of Amazon product reviews. This model acts as a feature extractor, transforming text into meaningful representations that can be used for downstream tasks like sentiment classification. The advantage lies in leveraging the learned linguistic patterns of the LSTM for nuanced sentiment understanding.

Quick Start & Requirements

• Install via pip install -r requirements.txt.
• Requires Python 3.x.
• Includes a PyTorch port for training from scratch.
• A demo for sentiment classification on SST is provided (sst_binary_demo.py).

Highlighted Details

• Trained on over 82 million Amazon product reviews (38 billion bytes).
• Achieved 12,500 characters/second processing on four NVIDIA Pascal GPUs during a one-month training period.
• Model parameters for a 4,096-unit multiplicative LSTM are included.
• Functionality as a feature extractor is demonstrated.

Maintenance & Community

Status: Archive (code is provided as-is, no updates expected).

Licensing & Compatibility

The repository does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The project is archived and no longer maintained or updated. The provided code is for research purposes and may require significant adaptation for production environments. Licensing is unclear, posing potential restrictions on commercial use.