Reviewer-Rec

Task

Here are several examples of reviewer matching by using TD-IDF, LightGCN and GF-CF. Please refer to the following steps to run our codes.

Requirements

• Python version >= 3.6
• PyTorch version >= 1.6.0
• Network for initial run
• pip install -r requirements.txt (Note: sparsesvd can be installed from source)

Processed Data

We provide the processed data in data-4k and data-8k. We also provide the reviewer profiles linked to Open Academic Graph (OAG), where each reviewer is associated with respective published papers [Matchings Part-1 Download] [Matchings Part-2 Download].

Usage

TF-IDF

We use the embedding method of TF-IDF, the pretrained vectorizer and model is provided via Aliyun, and put this folder as reviewer_rec_TFIDF/get_paper_embedding/embedding_data. Also, download the data from Here, and put this folder as reviewer_rec_TFIDF/get_paper_embedding/data.

Moreover, download tfidf_model.pkl from Baidu Pan with password jey2 or Aliyun and svd_model.pkl from Baidu Pan with password qcng or Aliyun. Put these two models in into reviewer_rec_TFIDF/get_paper_embedding/.

Download paper_embedding.json from https://pan.baidu.com/s/1mvNnpRY6fWOM4mUE3WsZQQ?pwd=7suq or Aliyun and training_reviewer_embedding.json from https://pan.baidu.com/s/1ish6ofqTm5dPiz0PpDQ9Hg?pwd=8jy8 or Aliyun. Put these two embedding files into reviewer_rec_TFIDF/get_paper_embedding/embedding_data.

Please run:

cd reviewer_rec_TFIDF
python parse_paper_information.py

to get the results.

GF-CF and LightGCN

This code is heavliy bulit on the official implementation of GF-CF and LightGCN.

Download the data from Here, and put this folder as /reviewer_rec_LightGCN/data/reviewer_rec.

To run LightGCN, use the following command:

cd reviewer_rec_LightGCN
python main.py --dataset="reviewer_rec" --topks="[20,5,10,50]" --model "lgn" --gpu_id 0

To run GF-CF, use the following command:

cd reviewer_rec_LightGCN
python main.py --dataset="reviewer_rec" --topks="[20,5,10,50]" --simple_model "gf-cf" --gpu_id 0