Lazybob at SemEval-2023 Task 9: Quantifying Intimacy of Multilingual Tweets with Multi-Task Learning

Mengfei Yuan, Cheng Chen


Abstract
This study presents a systematic method for analyzing the level of intimacy in tweets across ten different languages, using multi-task learning for SemEval 2023 Task 9: Multilingual Tweet Intimacy Analysis. The system begins with the utilization of the official training data, and then we experiment with different fine-tuning tricks and effective strategies, such as data augmentation, multi-task learning, etc. Through additional experiments, the approach is shown to be effective for the task. To enhance the model’s robustness, different transformer-based language models and some widely-used plug-and-play priors are incorporated into our system. Our final submission achieved a Pearson R of 0.6160 for the intimacy score on the official test set, placing us at the top of the leader board among 45 teams.
Anthology ID:
2023.semeval-1.128
Volume:
Proceedings of the 17th International Workshop on Semantic Evaluation (SemEval-2023)
Month:
July
Year:
2023
Address:
Toronto, Canada
Editors:
Atul Kr. Ojha, A. Seza Doğruöz, Giovanni Da San Martino, Harish Tayyar Madabushi, Ritesh Kumar, Elisa Sartori
Venue:
SemEval
SIG:
SIGLEX
Publisher:
Association for Computational Linguistics
Note:
Pages:
927–933
Language:
URL:
https://aclanthology.org/2023.semeval-1.128
DOI:
10.18653/v1/2023.semeval-1.128
Bibkey:
Cite (ACL):
Mengfei Yuan and Cheng Chen. 2023. Lazybob at SemEval-2023 Task 9: Quantifying Intimacy of Multilingual Tweets with Multi-Task Learning. In Proceedings of the 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 927–933, Toronto, Canada. Association for Computational Linguistics.
Cite (Informal):
Lazybob at SemEval-2023 Task 9: Quantifying Intimacy of Multilingual Tweets with Multi-Task Learning (Yuan & Chen, SemEval 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.semeval-1.128.pdf