Re15: Read the paper LEVEN: A Large-Scale Chinese Legal Event Detection Dataset

Gods are silent-personal CSDN blog directory

Paper Title: LEVEN: A Large-Scale Chinese Legal Event Detection Dataset

This article is a 2022 ACL paper and the author is from Tsinghua University.A legal event extraction dataset is presented.
This paper proposes a legal text fact extraction dataset with the largest amount of data considering comprehensive event types, including case-related events and ordinary events.Experiments show that using legal events as side information can enhance the accuracy of downstream legal adjudication prediction tasks.
This and EPM¹ are both 2022 ACLs, really cool…...I don't even understand how EPM's kind of fantasy stacking music can be successful, there is no simple beauty at all... and EPM's data volume is much smaller than LEVEN.

This article is only a brief overview, the details are to be added.

1. Background

Event extraction: extract event triggers and classify them into event types

2. Dataset LEVEN

Contains charge-related events and general events

3. baseline and experimental results

3.1 Event Extraction

1. token classification: BiLSTM, BERT
2. Dynamic max-pooling: DMCNN, DMBERT
3. Sequence labeling: BiLSTM+CRF, BERT+CRF
   

3.2 Downstream Tasks

Experimental results: directly use event information as downstream tasksside information, improved in low-resource judgment prediction tasks and unsupervised case retrieval tasks

---

1. Re11: Read the paper EPM Legal Judgment Predictionvia Event Extraction with Constraints︎