Earnings Call Signal Extraction (FDAP)

1-line summary

I built a Python-based system to scan earnings call transcripts and surface changes in management tone and messaging that matter for investment decisions.

Overview

This project focused on extracting useful signals from earnings calls where critical information is buried in unstructured text. I built a Python workflow that processes transcripts and highlights changes in tone, emphasis, and narrative over time.

Project Motivation

Earnings calls are subjective. Two analysts can read the same transcript and reach different conclusions. I wanted to make earnings analysis more consistent by identifying what changed quarter to quarter, rather than relying on intuition.

Technical Details

I cleaned and structured transcripts, split content into prepared remarks and Q&A, grouped commentary by themes, and tracked changes over time. I checked whether tone and emphasis shifts aligned with subsequent performance and market reactions.

Results and Impact

The framework made it faster to identify early warning signs and changes in management confidence, and helped guide where deeper fundamental work was needed.

Conclusion

Qualitative information can be handled with the same discipline as financial data. Data tools can sharpen judgment rather than replace it.