ATTEMPT: Pre-training with Aspect-Content Text Mutual Prediction for Multi-Aspect Dense Retrieval

This repo contains the source code of our paper in CIKM 2023: Pre-training with Aspect-Content Text Mutual Prediction for Multi-Aspect Dense Retrieval

• This project builds upon Tevatron
• For the pre-training, run src/config_run_pretrain.sh
• For the fine-tuning, run src/pretrain_finetune1_c.sh

Paper released

Please refer to our paper for details: paper

Dataset released

Multi-Aspect Amazon ESCI Dataset (MA-Amazon)

Contacts

• Xiaojie Sun: sunxiaojie21s@ict.ac.cn or sunxj1999@126.com