Zhijing Jin
Max Planck Society
42 Papers
25 Citations
Zhijing Jin is an academic researcher from Max Planck Society. The author has contributed to research in topics: Computer science & Natural language processing. The author has an hindex of 10, co-authored 28 publications. Previous affiliations of Zhijing Jin include Massachusetts Institute of Technology & University of Hong Kong.
Chat about Author
Papers
Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and Entailment
Di Jin,Zhijing Jin,Joey Tianyi Zhou,Peter Szolovits +3 more
- 03 Apr 2020
TL;DR: TextFooler as discussed by the authors is a baseline to generate adversarial text for text classification and textual entailment tasks, and it outperforms previous attacks by success rate and perturbation rate, preserving semantic content, grammaticality, and correct types.
•Posted Content
Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and Entailment
TL;DR: TextFooler is presented, a simple but strong baseline to generate adversarial text that outperforms previous attacks by success rate and perturbation rate, and is utility-preserving and efficient, which generates adversarialtext with computational complexity linear to the text length.
•Posted Content
Is BERT Really Robust? Natural Language Attack on Text Classification and Entailment
Di Jin,Zhijing Jin,Joey Tianyi Zhou,Peter Szolovits +3 more
- 27 Jul 2019
TL;DR: The TextFooler is presented, a general attack framework, to generate natural adversarial texts that outperforms state-of-the-art attacks in terms of success rate and perturbation rate.
249
Deep Learning for Text Style Transfer: A Survey
TL;DR: Text style transfer is an important task in natural language generation, which aims to control certain attributes in the generated text, such as politeness, emotion, humor, and many others as mentioned in this paper .
•Posted Content
GraphIE: A Graph-Based Framework for Information Extraction
TL;DR: Evaluation on three different tasks shows that GraphIE consistently outperforms the state-of-the-art sequence tagging model by a significant margin, and generates a richer representation that can be exploited to improve word-level predictions.
114