https://github.com/thunlp/HATT-Proto

Code and dataset of AAAI2019 paper Hybrid Attention-Based Prototypical Networks for Noisy Few-Shot Relation Classification
https://github.com/thunlp/HATT-Proto

few-shot relation-extraction

Last synced: about 2 months ago
JSON representation

Code and dataset of AAAI2019 paper Hybrid Attention-Based Prototypical Networks for Noisy Few-Shot Relation Classification

• Host: GitHub
• URL: https://github.com/thunlp/HATT-Proto
• Owner: thunlp
• License: mit
• Created: 2018-11-13T11:37:03.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2019-01-24T17:25:09.000Z (almost 6 years ago)
• Last Synced: 2024-07-31T23:45:11.352Z (4 months ago)
• Topics: few-shot, relation-extraction
• Language: Python
• Homepage:
• Size: 5.36 MB
• Stars: 185
• Watchers: 8
• Forks: 35
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-few-shot-meta-learning - code - official (PyTorch)

README

        
# Hybrid Attention-Based Prototypical Networks for Noisy Few-Shot Relation Classification

Code and data for AAAI2019 paper [Hybrid Attention-Based Prototypical Networks for Noisy Few-Shot Relation Classification](https://gaotianyu1350.github.io/assets/aaai2019_hatt_paper.pdf).

Author: Tianyu Gao*, Xu Han*, Zhiyuan Liu, Maosong Sun. (\* means equal contribution)

## Dataset and Word Embedding

We evaluate our models on [FewRel](https://thunlp.github.io/fewrel), a large-scale dataset for few-shot relation classification. It has 100 relations and 700 instances for each relation. You can find some baseline models from [here](https://github.com/thunlp/fewrel).

Due to the large size, we did not upload the glove file (pre-trained word embedding). Please download `glove.6B.50d.json` from [Tsinghua Cloud](https://cloud.tsinghua.edu.cn/f/b14bf0d3c9e04ead9c0a/?dl=1) or [Google Drive](https://drive.google.com/open?id=1UnncRYzDpezPkwIqhgkVW6BacIqz6EaB) and put it under `data/` folder.

## Usage

To run our code, use this command for training

```bash

python train.py {MODEL_NAME} {N} {K} {NOISE_RATE}

```

and use this command for testing

```bash

python test.py {MODEL_NAME} {N} {K} {NOISE_RATE}

```

where {MODEL_NAME} could be `proto` or `proto_hatt`, `{N}` is the num of classes, `{K}` is the num of instances for each class and `{NOISE_RATE}` is the probability that one instance is wrong-labeled.