mtl-disparate

Code for NAACL 2018 paper "Multi-task Learning of Pairwise Sequence Classification Tasks Over Disparate Label Spaces" by Isabelle Augenstein, Sebastian Ruder, Anders Søgaard

Note that this is research code and will not be maintained to e.g. ensure compatibility with more recent library versions.

Requirements:

• Tensorflow 1.5
• Numpy 1.12.1
• sklearn 0.18.1
• scipy

Steps to run:

• run data/download_data.sh to download and extract data
• preproc/data_reader.py tests if all the data readers work
• preproc/fnc_data_splits.py to split the FNC training dataset into a training and dev set
• main.py trains models

Datasets

SemEval 2016 Task 6 Stance detection

• Task website

Fake News Challenge (FNC)

• Task website

Multi-NLI

• Task website

SemEval 2016 Task 4 Subtask B Topic-based Twitter sentiment analysis

• Task website
• Task description paper
• Note: Same dataset was used as rerun in 2017

SemEval 2016 Task 5 Subtask 1 Slot 3 Aspect-based sentiment analysis

• Task website

Clickbait Challenge 2017

• Task website