Papers by Shengxuan Luo
Label Refinement via Contrastive Learning for Distantly-Supervised Named Entity Recognition (2022.findings-naacl)
Copied to clipboard
| Challenge: | Existing methods to locate and classify entities using knowledge bases and unlabeled corpus are expensive and limited application. |
| Approach: | They propose to use a method to directly learn the distant label refinement knowledge by imitating annotations of different qualities and comparing them in contrastive learning frameworks. |
| Outcome: | The proposed method can give modified suggestions on distant data without additional supervised labels and thus reduces the requirement on the quality of the knowledge bases. |
An Accurate Unsupervised Method for Joint Entity Alignment and Dangling Entity Detection (2022.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for knowledge graph integration lack dangling entities that can be manually extracted. |
| Approach: | They propose a Unsupervised method for joint Entity alignment and Dangling entity detection that uses literal semantic information to generate pseudo entity pairs and globally guided alignment information for EA. |
| Outcome: | The proposed method outperforms state-of-the-art methods in the EA and DED tasks and achieves comparable results without supervision. |