Résumé Text Generator

Context

  • We developed an English language resume/cover letter text generator to guide students through the arduous process of applying for jobs. 

  • Our tool would generate a couple of lines after the user inputs a topic or a job title.

  • The output can then be adjusted and modified manually by the user. 

  • Our tool will provide the user with a template that they can personalize according to their own individual experiences.

Data

  • Our data was a collection of resume samples taken from livecareer.com

  • The dataset contains 2400+ resumes in both string and PDF format

  • The provided CSV file contains:

    • ID: Unique identifier and file name for the respective PDF

    • Resume_str : resume text only in string format

    • Resume_html : resume data in HTML format as present while web scraping

    • Category of the job resume : the categories included are: HR, Designer, Information-Technology, Teacher, Advocate, Business-Development, Healthcare, Fitness, Agriculture, BPO, Sales, Consultant, Digital-Media, Automobile, Chef, Finance, Apparel, Engineering, Accountant, Construction, Public-Relations, Banking, Arts, Aviation

Methodology

  • Preprocessing 

    • For each Experience section:

      • Get the job titles

      • Get the bullet points or the short summary of text that shows relevant skill

      • Remove all non-english characters, punctuation and numbers 

  • Algorithms used to solve the problem

    • GPT-Neo 125M 

    • GPT-Neo 355M

Preliminary Results

Input (a lot of data): Teacher: Developed lesson

Generated: Teacher: Developed lesson plans and administered test strategies to drive test result toward Common Core and Science Scope objectives

Input (some data): Project Coordinator: Collected

Generated: Project Coordinator: Collected customer feedback and made process changes to exceed customer satisfaction goals

Input (not a lot of data): Finance and Sales Consultant: Planned

Generated: Finance and Sales Consultant: Planned and organized corporate conference rooms

Ethical Considerations

  • The resumes that were used for fine-tuning purposes were only those given scores of 85+ (of 100) by LiveCareer1

    • Biases of recruiters could be reflected in the resumes selected

  • Potential misuse of our raw text generations

    • GPT-Neo is agnostic of client’s personal experiences, so it should NOT be relied on exclusively when writing one’s resumes/cover letters (yet!)

    • Our code merely assists with the writing style of the resume, it is up to clients to add personalization

Feel free to read our report included below or take a look at our code.