CareerStanding resume-sample study — reproducibility package

This package contains the analysis script and saved aggregate results behind
https://careerstanding.com/resources/resume-data-study-2483-resumes

Source datasets (download directly from their publishers):
1. https://huggingface.co/datasets/talanAI/resumesamples
   File: resumesamples.csv (save as data/talan.csv)
2. https://huggingface.co/datasets/Sachinkelenjaguri/Resume_dataset
   File: UpdatedResumeDataSet.csv (save as data/updated.csv)

To reproduce:
1. Install Node.js 20 or newer.
2. Save analyze.mjs in a directory and create a data/ subdirectory.
3. Download the two CSV files above to data/talan.csv and data/updated.csv.
4. Run: node analyze.mjs
5. Compare the generated results.json with the published results.json.

The primary dataset has 2,483 rows and 24 categories. The cross-check dataset
has 962 rows; 940 pass the script's minimum-length filter. The script uses
rule-based text detection. The resulting percentages describe these samples;
they do not measure interview probability or prove that changing a resume
causes more callbacks. The public datasets may contain identifying information
(including an Email field in the primary dataset). We do not redistribute raw
resume text or email addresses. No CareerStanding customer data was used.