ML for SS: Applying GPT Models

Dr. Richard M. Crowley
rcrowley@smu.edu.sg

https://rmc.link/

Overview

Papers

Paper 1: de Kok (2024 MS Forth) “ChatGPT for Textual Analysis? How to use Generative LLMs in Accounting Research.”

Paper 2: Rooein et al. (2024 Working) “Beyond Flesch-Kincaid: Prompt-based Metrics Improve Difficulty Classification of Educational Texts.”

Paper 3: Kim, Muhn and Nikolaev (2024 Working) “Bloated disclosures: can ChatGPT help investors process information?”

Technical Discussion: Ensembling

We already covered implementing the OpenAI API last week

Conclusion

Wrap-up

LLMs can be used in many ways

  • Classification
  • Experimentation
  • Data validation
  • Data creation
  • Measurement (through methods applied to generative output)

Word to the wise

Be sure to implement the methods appropriately and with due caution.

Packages used for these slides

Python

  • openai
  • pandas

R

References

  • Kim, Alex, Maximilian Muhn, and Valeri V. Nikolaev. “Bloated disclosures: can ChatGPT help investors process information?.” Chicago Booth Research Paper 23-07 (2024): 2023-59.
  • de Kok, Ties. “ChatGPT for Textual Analysis? How to use Generative LLMs in Accounting Research.” How to use Generative LLMs in Accounting Research (March 1, 2024) (2024).
  • Rooein, Donya, Paul Rottger, Anastassia Shaitarova, and Dirk Hovy. “Beyond Flesch-Kincaid: Prompt-based Metrics Improve Difficulty Classification of Educational Texts.” arXiv preprint arXiv:2405.09482 (2024).