GenAI-in-Peer-Review

Repository of the paper titled "Can generative AI solve the peer review crisis?". The script performs the following analyses:

1. Citation Network:

   • Creates a directed graph of citations between studies.
   • Nodes represent studies, and edges represent citations.
   • Node size corresponds to the number of citations received.

2. Co-Citation Network:

   • Constructs an undirected graph highlighting pairs of articles frequently cited together.
   • Edge weights reflect co-citation frequency.

3. Keyword Co-Occurrence Network:

   • Builds an undirected graph showing relationships between keywords frequently appearing together in studies.
   • Node size reflects keyword importance (frequency).
   • Edge weights indicate co-occurrence strength.

4. Keyword Adjacency Matrix:

   • Outputs a CSV file containing the keyword co-occurrence matrix for further analysis.

Requirements

The script uses the following Python libraries:

• pandas
• networkx
• matplotlib
• seaborn

Install them with:

pip install pandas networkx matplotlib seaborn

Input

The script processes the MetadataFor132Studies.csv file. Key columns used in the analysis:

• EID: A unique identifier for each study.
• References: A semicolon-separated list of references for each study.
• Author Keywords: A semicolon-separated list of keywords associated with each study.

The CSV file also includes other metadata, such as Authors, Title, Abstract, and Cited by, which can be leveraged for extended analysis.

Outputs

The script generates the following files:

1. Citation Network Visualization: citation_network.png
2. Co-Citation Network Visualization: co_citation_network.png
3. Keyword Co-Occurrence Network Visualization: keyword_network.png
4. Keyword Co-Occurrence Matrix: keyword_cooccurrence_matrix.csv

Usage

Step 1: Prepare the Input File

Ensure the CSV file, MetadataFor132Studies.csv, is in the same directory as the script. Verify that columns EID, References, and Author Keywords are populated.

Step 2: Run the Script

Execute the script with:

python BibliometricAnalysis.py

Step 3: Analyze the Outputs

• View the PNG files for network visualizations:
  • Citation Network: Understand citation relationships between studies.
  • Co-Citation Network: Explore which studies are frequently cited together.
  • Keyword Co-Occurrence Network: Discover frequently co-occurring keywords in the dataset.
• Use the CSV file (keyword_cooccurrence_matrix.csv) for additional analysis or custom visualizations in Python, R, or Excel.

Example

Example of MetadataFor132Studies.csv:

EID,References,Author Keywords,Title
2-s2.0-85169501301,"Ref1;Ref2;Ref3","Keyword1;Keyword2;Keyword3","Study on Generative AI in Peer Review"
2-s2.0-85201631926,"Ref2;Ref4","Keyword1;Keyword4","Exploration of AI in Academia"
...

How It Works

1. Data Preprocessing:

   • Reads the input CSV file and cleans data.
   • Processes References and Author Keywords columns to extract meaningful information.

2. Network Construction:

   • Citation Network: Links studies based on citations.
   • Co-Citation Network: Links studies frequently cited together.
   • Keyword Network: Links keywords appearing together in studies.

3. Visualization:

   • Generates clear, publication-quality network visualizations using matplotlib and networkx.

4. Data Export:

   • Outputs the keyword co-occurrence matrix as a CSV file for extended analysis.

Notes

• The script assumes that the CSV file follows a standardized format (e.g., References and Author Keywords separated by semicolons).
• Adjust the visualization parameters (e.g., node size, edge weights) for larger datasets to improve clarity.