This is a basic model for Irony Detection Subtask held at SemEval 2018. The model is based on the allennlp library.
The model is composed of three basic components :-
- Reader - Responsible for reading the dataset in .txt files in the dataset folder.
- Model - The model module defines the neural net we want to use for the model. It is currently a sequence to sequence encoder whoose output is fed to a feed forward network for classification.
- Predictor - This part is required for running the demo and evaluate procedure.
The experiments directory contains the config files which contains the hyperparameters and the model configuration.
To train the model run:
allennlp train experiments/averagedencoder.json -s directory_path_to_save_model --include-package irony_model
Requirements:
- Allennlp==0.7.0
TODO
- Add New Models
- Update Documentation and Readme