Skip to content

Latest commit

 

History

History
49 lines (46 loc) · 2.7 KB

README.md

File metadata and controls

49 lines (46 loc) · 2.7 KB

How-to-run

(1)Data anlysis

The code under 'two levels' and 'three levels' is to calculate the MJS Divergence of two levels and three levels.
two level/2017-2019.py the file is to claculate the MJS of two levels of 2017-2019.
three levels/2017-2019.py the file is to claculate the MJS of two levels of 2020.
while 2020.py under 'two levels' and 'three levels' is to claculate the MJS of two/three levels of 2020.
when you want to run the code,please modify the actual path of the source dir to ensure you can run the code.

(2)cluster

This module clusters abstracts and keywords to find interesting information.Please load the following command:
cd ./cluster analysis/
python Clustering.py
Here is brief description of each file.

  • "decision.txt" Description of the reception, 1 for reception, 0 for rejection.
  • "stopwords.txt" Include common early stop words in the file
  • "userdict2.txt"Common key words and phrases in scientific papers

(3)sentiment analysis

first ,we should download pre-training model ELECTRA-Small. Click the link https://storage.googleapis.com/electra-data/electra_small.zip to download to ../../data/sentiment analysis data/

generate model

The realization of emotion analysis module. Please load the following command:
cd ./sentiment analysis/
python trainmodel.py
The script trainmodel.py first calls pretrain model to Initialization parameters.Then load the dataset training to fine tune.

sentiment predict

For emotional prediction, please use the following command:python predict.py
Here is brief description of each code.

  • "generate.py"Match the predicted results with the review.
  • "generatecount.py"Statistics of the average scores of different combinations of five angles.
  • "drawsentimentpic.py"Visualization of results

The format of training data is as follows:
text,label

(4)correlation analysis

This module analyzes the relationship between paper scores and citations.Please load the following command:
cd ./citeandscore anlysis/
python Citescatter17.py
The format of training data is as follows:
id title cite final_decision average_score

(5)arxiv analysis

This module analyzes the influence of publishing on arXiv.Please load the following command: cd ./citeandscore anlysis/
python ArxivGraph2017.py
If you like to crawl the raw datase,please use the following command:
python crawlarxiv.py
The script crawlarxiv.py,calls the arvix open source API port.

(6)rebuttal analysis

This model predicts the change of score after rebuttal.
python ./DBERTATTENTION.py