This repository provides codes for paper "Dense Retrieval as Indirect Supervision for Large-space Decision Making" accepted by EMNLP 2023
- Download the dataset
NOTE: please refer to https://www.cs.utexas.edu/~eunsol/html_pages/open_entity.html for original UFET dataset
mkdir -p datasets && cd datasets
gdown https://drive.google.com/file/d/1HYqRFDbctj6VkqAZt1IVenBnUk7j65pd/view?usp=sharing && unzip UFET.zip
cd ..
- Prepare DPR data format
python explore_ufet.py --stage data_preparation
files are saved at ./datasets/UFET/crowd/dpr_processed/query_context_direct_256
3. Sample evaluate codes
python explore_ufet.py --stage evaluation
Download the dataset
mkdir -p repositories && cd repositories
git clone https://github.com/jianguoz/Few-Shot-Intent-Detection.git
cd ..
gdown https://drive.google.com/file/d/1iGtJqMXHLIXb8BU5fbUAZdjGpD_92Cdm/view?usp=sharing && unzip results.zip
python explore_banking77.py
# 5-shot train data: ./datasets/BANKING77/train_5/dpr_processed/query_context_direct_256_train_5_gtr-t5-xl_-1examples/trn.json
# 10-shot train data: ./datasets/BANKING77/train_10/dpr_processed/query_context_direct_256_train_10_gtr-t5-xl_-1examples/trn.json
python explore_clinc150.py
# 5-shot train data: ./datasets/CLINC150/train_5/dpr_processed/query_context_direct_256_train_5_gtr-t5-xl_-1examples/trn.json
# 10-shot train data: ./datasets/CLINC150/train_10/dpr_processed/query_context_direct_256_train_10_gtr-t5-xl_-1examples/trn.json
python explore_hwu64.py
# 5-shot train data: ./datasets/HWU64/train_5/dpr_processed/query_context_direct_256_train_5_gtr-t5-xl_-1examples/trn.json
# 10-shot train data: ./datasets/HWU64/train_10/dpr_processed/query_context_direct_256_train_10_gtr-t5-xl_-1examples/trn.json
This repo wouldn't be possible without the awesome UFET, Context-TE, Few-Shot-Intent-Detection
If you find DDR useful for your research and applications, please consider starring this repository and citing:
@article{xu2023dense,
title={Dense Retrieval as Indirect Supervision for Large-space Decision Making},
author={Xu, Nan and Wang, Fei and Dong, Mingtao and Chen, Muhao},
journal={arXiv preprint arXiv:2310.18619},
year={2023}
}