/Reproduce-DLmethods-Biodiv

This Repo contains the data and codes that were used to extract and analyse the reproducibility information of Deel Learning methods from publications in the Biodiversity domain.

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

Reproducibility of deep learning models in the biodiversity domain

This Repo contains the data and codes that were used to extract and analyse the reproducibility information of Deep Learning methods from publications in the Biodiversity domain

  • web_scraping.py ----> Python code that can scrape the publication information from Google Scholar with a specified query and time period (2015-2021)
  • Google_scholar_500_publications_run1.json ----> Json file containing information on 500 Biodiversity publications based on the query and time period as mentioned in the code
  • Variable_info_VK_v1.csv ----> Recorded variable level information as a binary response: available (y), not available (n) by annotator 1
  • Variable_info_WA_v1.csv ----> Recorded variable level information as a binary response: available (y), not available (n) by annotator 2
  • Inter_Annotator_Agreement.py ----> Python code for calculating the Inter-Annotator Agreement
  • Compare_annotations.py ----> Python code for checking the mismatching responses between two annotators
  • Data_after_inter_annotator_agreement.csv ----> Dataset after resolving the mismatches between two annotators responses
  • Category_and_levels.py ----> Python code for compiling categorical and publication reproducibility level information using variable level information
  • Final_data.csv ----> Final dataset containing variable, categorical, publication reproducibility level that was used for analysing the results