Selecting Stickers in Open-Domain Dialogue through Multitask Learning

TL;DR

This paper introduces a multitask learning approach for selecting appropriate stickers in open-domain dialogue, leveraging auxiliary tasks to improve understanding of dialogue context and sticker semantics, resulting in higher accuracy.

Contribution

The paper presents a novel multitask learning framework with three auxiliary tasks specifically designed for sticker selection in open-domain dialogue.

Findings

Model outperforms strong baselines in accuracy

Auxiliary tasks significantly improve understanding of dialogue and stickers

Ablation study confirms effectiveness of each auxiliary task

Abstract

With the increasing popularity of online chatting, stickers are becoming important in our online communication. Selecting appropriate stickers in open-domain dialogue requires a comprehensive understanding of both dialogues and stickers, as well as the relationship between the two types of modalities. To tackle these challenges, we propose a multitask learning method comprised of three auxiliary tasks to enhance the understanding of dialogue history, emotion and semantic meaning of stickers. Extensive experiments conducted on a recent challenging dataset show that our model can better combine the multimodal information and achieve significantly higher accuracy over strong baselines. Ablation study further verifies the effectiveness of each auxiliary task. Our code is available at \url{https://github.com/nonstopfor/Sticker-Selection}