Wildfire-related publications (1980-2022)

The CSV file contains all 60,488 publications that are related with wildfire studies. 

54 columns are contained and can be summarized into 4 categories:

Columns (authors, ISBN, abstract, doi, type, title, journal, dates, affiliations) are accessed directly from the Web of Science (WOS) bibliographic database.

Columns that start with "token_" are the consumed time of LLM responses (in this paper, gpt-3.5-turbo that was updated on 13th Feb 2023)

Columns (study_area, study_period) are the interpreted or geoparsed results by LLM.

Columns that start with "main_" are the selected topics in Fig. 1. Those start with "sub_" are the sub-topics. The results are marked by 1 (true) or 0 (false).

Columns that start with "ignition_type_" are the interpreted results pertaining to different sources of ignitions. The results are marked by 1 (true) or 0 (false).

Columns that start with "paradigms_" are the interpreted results pertaining to different aspects of wildfire study paradigms. The results are marked by 1 (true) or 0 (false).

The code is written in Python and shows the process of interactions with gpt models.

The data for the main text figures are uploaded to this repository: Wildfire-related publications (1980-2022) associated main figure data (zenodo.org)