Optimizing the Performance of Text Classification Models by Improving the Isotropy of the Embeddings Using a Joint Loss Function
- 19 August 2023
- book chapter
- Published by Springer Nature
- p. 121-136
Abstract
No abstract availableKeywords
This publication has 16 references indexed in Scilit:
- Learning to Remove: Towards Isotropic Pre-trained BERT Embedding2021
- How Does Fine-tuning Affect the Geometry of Embedding Space: A Case Study on Isotropy2021
- All Bark and No Bite: Rogue Dimensions in Transformer Language Models Obscure Representational Quality2021
- A Cluster-based Approach for Improving Isotropy in Contextual Embedding Space2021
- Too Much in Common: Shifting of Embeddings in Transformer Language Models and its Implications2021
- On the Sentence Embeddings from Pre-trained Language Models2020
- Revisiting Representation Degeneration Problem in Language Modeling2020
- How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings2019
- Predicting the Type and Target of Offensive Posts in Social Media2019
- GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding2018