Multi-Task Learning for Pun Detection and Location with BERT and Graph Convolutional Neural Networks

The Final Model is:

BERT_GNN_MTL_Control_BiLSTM_v4.ipynb

SemEval-2017 Task 7

https://alt.qcri.org/semeval2017/task7/

This is the Final Project of Group 4, COSC-572 Empirical Methods in Natural Language Processing, Spring 2023, Georgetown University

Course Webpage:
https://people.cs.georgetown.edu/nschneid/cosc572/s23/index.html

Data Preprocessing in:

BERT_GNN_MTL_Control_BiLSTM_v1.ipynb

The Project is Including Two Tasks:

Subtask 1: Pun detection

For this subtask, participants are given an entire raw data set. For each context, the system must decide whether or not it contains a pun.

Subtask 2: Pun location

For this subtask, the contexts not containing puns are removed from the data set. For each context, the system must identify which word is the pun.

Environment:

Google Colab (with Python 3)
https://colab.research.google.com/

Usage:

The file name of each file contains the models which it used.
All the files can be run in Google Colab.
For the Evaluation.ipynb, you can load a model (.pt file) from Google Drive and do the evaluation.

Dataset:

1. SemEval-2017 Task 7

https://alt.qcri.org/semeval2017/task7/data/uploads/semeval2017_task7.tar.xz

2. Pun of the Day

From Yang et al. (2015)
(This dataset is not publicly available and was obtained by contacting paper authors directly.)