Augmenting Document Representations for Dense Retrieval with Interpolation and Perturbation

Official Code Repository for the paper "Augmenting Document Representations for Dense Retrieval with Interpolation and Perturbation" (ACL 2022)

Abstract

Dense retrieval models, which aim at retrieving the most relevant document for an input query on a dense representation space, have gained considerable attention for their remarkable success. Yet, dense models require a vast amount of labeled training data for notable performance, whereas it is often challenging to acquire query-document pairs annotated by humans. To tackle this problem, we propose a simple but effective Document Augmentation for dense Retrieval (DAR) framework, which augments the representations of documents with their interpolation and perturbation. We validate the performance of DAR on retrieval tasks with two benchmark datasets, showing that the proposed DAR significantly outperforms relevant baselines on the dense retrieval of both the labeled and unlabeled documents.

Installation

We refer to the repository of DPR as a skeleton code. Note that we mainly modified here and here for interpolation and perturbation, respectively.

conda create -n dar python=3.7
conda activate dar
conda install pytorch cudatoolkit=11.1 -c pytorch
cd DAR
pip install -r requirements.txt

If such an error shows up,

OSError: [E050] Can't find model 'en_core_web_sm'. It doesn't seem to be a Python package or a valid path to a data directory.

then execute the following command:

python -m spacy download en

Run

1. Train dense retriever

We use a single GeForce RTX 3090 GPU having 24GB memory to train DAR with a batch size of 32.

sh 0_train_retriever.sh

2. Generate dense embeddings

This step is for generating dense embeddings, which may require a parallelizable process.

sh 1_gen_emb.sh

3. Retrieve

sh 2_dense_retrieve.sh

4. Read (Optional)

This is an extractive QA setting without re-training a reader model.

sh 3_read.sh

Citation

@inproceedings{jeong2022augmenting,
   title={Augmenting Document Representations for Dense Retrieval with Interpolation and Perturbation},
   author={Soyeong Jeong and Jinheon Baek and Sukmin Cho and Sung Ju Hwang and Jong C. Park},
   booktitle={Association for Computational Linguistics (ACL)},
   year={2022}
}

starsuzi/DAR