Opinion Lab

The repository contains codebase for building state-of-the-art deep learning techniques for opinion mining.

Currenly following models have been created and tested:

• Models for IEST @ WASSA-2018 [FNN, CNN, LSTM, CNN-LSTM, LSTM-CNN ++]

Prerequisites

1. Related resources for featureizers (only if you are using them) (ex: lexicons, word-embedding models)
2. Some python knowledge

Setup the Lab

1. Install textkit-learn
2. Configure path to resources in oplab/config.py
3. Happy Experimenting!

How to Start

• To train a model its simple as calling python oplab train -c [config_file] -m [model_path]

• To evaluate just python oplab evaluate -m [model_path] -e [evaluation_metrics] -o [predictions_file]

Licensing

• Code is under Apache License 2.0.
• Model (releases) are under Creative Commons Attribution - Non Commercial 2.0 Generic license.