- This repo is for code developed during TUM practrical course: Legal NLP practical lab @2023
for more detailed description about each experiment, check experiments_catalogue.txt
| Path | Description | CL on CL | CL on IT | IT on IT | IT on CL |
|---|---|---|---|---|---|
| code/baseline_BERT_only | naive baseline model. model is bert-base-uncased with feed forward layer on top for classification. BERT is not forzen. BERT generate embeddings. | 0.56 | 0.42 | 0.62 | 0.54 |
| code/baseline_hierarichal_BiLSTM/Hierarichal... | transform the documents of sentences into documents of embeddings using BERT bert-base-uncased. | 0.62 | - | - | - |
| code/baseline_hierarichal_BiLSTM/Hierarichal... | transform the documents of sentences into documents of embeddings using BERT bert-base-uncased. | 0.39 | - | - | - |
| code/copied/hier-bilstm-crf-baseline.ipynb | the experiment is from here. | 0.38 | - | - | - |
| code/prototypical/prototypical-CL.ipynb | a prototypical network with LSTM encoder over the BERT embeddings. | F1-score for k=2: 49.23% (+-10.41%) F1-score for k=4: 51.08% (+-9.09%) F1-score for k=8: 51.74% (+-8.35%) F1-score for k=16: 52.54% (+-7.76%) F1-score for k=32: 53.31% (+-7.41%) |
- | - | - |
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent