Views
No views yet
1@inproceedings{miaschi-etal-2024-linguistic-knowledge,
2 title = "Linguistic Knowledge Can Enhance Encoder-Decoder Models (If You Let It)",
3 author = "Miaschi, Alessio and
4 Dell{'}Orletta, Felice and
5 Venturi, Giulia",
6 editor = "Calzolari, Nicoletta and
7 Kan, Min-Yen and
8 Hoste, Veronique and
9 Lenci, Alessandro and
10 Sakti, Sakriani and
11 Xue, Nianwen",
12 booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)",
13 month = may,
14 year = "2024",
15 address = "Torino, Italy",
16 publisher = "ELRA and ICCL",
17 url = "https://aclanthology.org/2024.lrec-main.922",
18 pages = "10539--10554",
19 abstract = "In this paper, we explore the impact of augmenting pre-trained Encoder-Decoder models, specifically T5, with linguistic knowledge for the prediction of a target task. In particular, we investigate whether fine-tuning a T5 model on an intermediate task that predicts structural linguistic properties of sentences modifies its performance in the target task of predicting sentence-level complexity. Our study encompasses diverse experiments conducted on Italian and English datasets, employing both monolingual and multilingual T5 models at various sizes. Results obtained for both languages and in cross-lingual configurations show that linguistically motivated intermediate fine-tuning has generally a positive impact on target task performance, especially when applied to smaller models and in scenarios with limited data availability.",
20}Abstract: In this paper, we explore the impact of augmenting pre-trained Encoder-Decoder models, specifically T5, with linguistic knowledge for the prediction of a target task. In particular, we investigate whether fine-tuning a T5 model on an intermediate task that predicts structural linguistic properties of sentences modifies its performance in the target task of predicting sentence-level complexity. Our study encompasses diverse experiments conducted on Italian and English datasets, employing both monolingual and multilingual T5 models at various sizes. Results obtained for both languages and in cross-lingual configurations show that linguistically motivated intermediate fine-tuning has generally a positive impact on target task performance, especially when applied to smaller models and in scenarios with limited data availability.