Sticker-Selection

This repository contains the code for paper "Selecting Stickers in Open-Domain Dialogue through Multitask Learning" accepted to the Findings of ACL 2022.

dependency

python>=3.6

pip install -r requirements.txt

process data

1.download dstc10 data at https://github.com/lizekang/DSTC10-MOD/tree/main/data

2.put the downloaded data in the directory dstc_data (including train.json, validation.json, c_test_easy_task2.json, c_test_hard_task2.json). Note that train.json is renamed from the downloaded c_train.json.

3.put the downloaded data in the directory data(including img2id.json, meme_set).

4.python create_data.py

train

bash run.sh 0

test

bash run.sh 1