Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/DaoD/COCA
CIKM 2021: Contrastive Learning of User Behavior Sequence for Context-Aware Document Ranking
https://github.com/DaoD/COCA
Last synced: 6 days ago
JSON representation
CIKM 2021: Contrastive Learning of User Behavior Sequence for Context-Aware Document Ranking
- Host: GitHub
- URL: https://github.com/DaoD/COCA
- Owner: DaoD
- Created: 2021-08-09T13:48:02.000Z (almost 3 years ago)
- Default Branch: main
- Last Pushed: 2022-09-28T09:06:04.000Z (over 1 year ago)
- Last Synced: 2024-02-29T05:37:41.362Z (4 months ago)
- Language: Python
- Homepage:
- Size: 62.5 KB
- Stars: 18
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: readme.md
Lists
- awesome-stars - DaoD/COCA - CIKM 2021: Contrastive Learning of User Behavior Sequence for Context-Aware Document Ranking (Python)
README
# Contrastive Learning of User Behavior Sequence for Context-Aware Document Ranking
[](#python)
This repository contains the source code and datasets for the CIKM 2021 paper [Contrastive Learning of User Behavior Sequence for Context-Aware Document Ranking](https://arxiv.org/pdf/2108.10510.pdf) by Zhu et al.
## Abstract
Context information in search sessions has proven to be useful for capturing user search intent. Existing studies explored user behavior sequences in sessions in different ways to enhance query suggestion or document ranking. However, a user behavior sequence has often been viewed as a definite and exact signal reflecting a user's behavior. In reality, it is highly variable: user's queries for the same intent can vary, and different documents can be clicked. To learn a more robust representation of the user behavior sequence, we propose a method based on contrastive learning, which takes into account the possible variations in user's behavior sequences. Specifically, we propose three data augmentation strategies to generate similar variants of user behavior sequences and contrast them with other sequences. In so doing, the model is forced to be more robust regarding the possible variations. The optimized sequence representation is incorporated into document ranking. Experiments on two real query log datasets show that our proposed model outperforms the state-of-the-art methods significantly, which demonstrates the effectiveness of our method for context-aware document ranking.
Authors: Yutao Zhu, Jian-Yun Nie, Zhicheng Dou, Zhengyi Ma, Xinyu Zhang, Pan Du, Xiaochen Zuo, and Hao Jiang
## Requirements
I test the code with the following packages. Other versions may also work, but I'm not sure.
- Python 3.8.5
- Pytorch 1.8.1 (with GPU support)
- Transformers 4.5.1
- [pytrec-eval](https://pypi.org/project/pytrec-eval/) 0.5
## Usage
- Obtain the data (some data samples are provided in the data directory)
- For AOL dataset, please contact the author of [CARS](https://arxiv.org/pdf/1906.02329.pdf)
- For Tiangong dataset, you can download it from the [link](http://www.thuir.cn/tiangong-st/)
- Prepare pretrained BERT
- [BertModel](https://huggingface.co/bert-base-uncased)
- [BertChinese](https://huggingface.co/bert-base-chinese)
### Contrastive Learning Stage
#### AOL Dataset
```
python runBertContras.py --task aol --bert_model_path ../BERT/BertModel/
```
#### Tiangong Dataset
```
python runBertContras.py --task tiangong --bert_model_path ../BERT/BertChinese/ --epochs 5 --temperature 0.05
```
### Ranking Stage
#### AOL Dataset
```
python runBert.py --task aol --bert_model_path ../BERT/BertModel/ --pretrain_model_path ../ContrastiveLearning/model/BertContrastive.aol.4.10.128.sent_deletion.term_deletion.qd_reorder
```
#### Tiangong Dataset
```
python runBert.py --task tiangong --bert_model_path ../BERT/BertChinese/ --pretrain_model_path ../ContrastiveLearning/model/BertContrastive.tiangong.5.5.128.sent_deletion.term_deletion.qd_reorder
```
### Trained Model
The trained model in both stages are available at the [link](https://drive.google.com/drive/folders/1_EgQg3lRVST07G5v9M9YpXvyF4nv2mgK?usp=sharing)
The diarectory structure is:
```
COCA
├── BERT
│ ├── BERTChinese
│ └── BERTModel
├── ContrastiveLearning
│ ├── BertContrasPretrain.py
│ ├── data
│ │ ├── aol
│ │ └── tiangong
│ │ ├── dev.pos.txt
│ │ ├── test.pos.txt
│ │ └── train.pos.txt
│ ├── file_preprocess_dataset.py
│ ├── log
│ ├── model
│ ├── output
│ │ ├── aol
│ │ └── tiangong
│ └── runBertContras.py
└── Ranking
├── BertSessionSearch.py
├── Trec_Metrics.py
├── data
│ ├── aol
│ └── tiangong
│ ├── dev.point.txt
│ ├── test.point.lastq.txt
│ ├── test.point.preq.txt
│ └── train.point.txt
├── file_dataset.py
├── log
├── model
├── output
│ ├── aol
│ └── tiangong
└── runBert.py
```
## Citations
If you use the code and datasets, please cite the following paper:
```
@inproceedings{ZhuNDMZDZJ21,
author = {Yutao Zhu and
Jian{-}Yun Nie and
Zhicheng Dou and
Zhengyi Ma and
Xinyu Zhang and
Pan Du and
Xiaochen Zuo and
Hao Jiang},
editor = {Gianluca Demartini and
Guido Zuccon and
J. Shane Culpepper and
Zi Huang and
Hanghang Tong},
title = {Contrastive Learning of User Behavior Sequence for Context-Aware Document
Ranking},
booktitle = {{CIKM} '21: The 30th {ACM} International Conference on Information
and Knowledge Management, Virtual Event, Queensland, Australia, November
1 - 5, 2021},
pages = {2780--2791},
publisher = {{ACM}},
year = {2021},
url = {https://doi.org/10.1145/3459637.3482243},
doi = {10.1145/3459637.3482243}
}
```