Towards Linguistically Informed Multi-objective Transformer Pre-training for Natural Language Inference
We introduce a linguistically enhanced combination of pre-training methods for transformers. The pre-training objectives include POS-tagging, synset prediction based on semantic knowledge graphs, and parent prediction based on dependency parse trees. Our approach achieves competitive results on the Natural Language Inference task, compared to the state of the art. Specifically for smaller models, the method results in a significant performance boost, emphasizing the fact that intelligent pre-training can make up for fewer parameters and help building more efficient models. Combining POS-tagging and synset prediction yields the overall best results.
- Published in:
European Conference on Information Retrieval - Type:
Inproceedings - Authors:
Pielka, Maren; Schmidt, Svetlana; Pucknat, Lisa; Sifa, Rafet - Year:
2023
Citation information
Pielka, Maren; Schmidt, Svetlana; Pucknat, Lisa; Sifa, Rafet: Towards Linguistically Informed Multi-objective Transformer Pre-training for Natural Language Inference, European Conference on Information Retrieval, 2023, https://link.springer.com/chapter/10.1007/978-3-031-28238-6_46, Pielka.etal.2023a,
@Inproceedings{Pielka.etal.2023a,
author={Pielka, Maren; Schmidt, Svetlana; Pucknat, Lisa; Sifa, Rafet},
title={Towards Linguistically Informed Multi-objective Transformer Pre-training for Natural Language Inference},
booktitle={European Conference on Information Retrieval},
url={https://link.springer.com/chapter/10.1007/978-3-031-28238-6_46},
year={2023},
abstract={We introduce a linguistically enhanced combination of pre-training methods for transformers. The pre-training objectives include POS-tagging, synset prediction based on semantic knowledge graphs, and parent prediction based on dependency parse trees. Our approach achieves competitive results on the Natural Language Inference task, compared to the state of the art. Specifically for smaller models,...}}
