This project proposes a novel methodology to automatically learn financial lexicons that outperform the benchmark Loughran-McDonald lexicon in sentiment analysis tasks

A polarity based dictionary consisting of accounting vocabularies associated with categories of financial indicators.

From-scratch replication of Loughran and McDonald (2011) — SEC 10-K sentiment analysis with the LM Master Dictionary, and Fama-MacBeth regressions on filing-period excess returns.

EDGAR + Loughran-McDonald + FRED-driven borrower watchlist scoring 10-K disclosure tone delta against the prevailing credit-spread regime.

A peer-relative text-analysis pipeline that detects corporate distress from SEC 10-K Risk Factor disclosures. 79% recall + 61-69% extended-horizon precision across 24 failures in 15 sectors.

Jump detection via expanding-window Z-score on SPY and BRK.B (2020-2025) + logistic regression using Loughran-McDonald NLP sentiment from 48 SEC 8-K filings. McFadden R²=0.50. VCU FIRE 691.

SEC 10-K Filing Sentiment & Risk Analyzer

Replicate the Loughran and McDonald 2011 sentiment analysis study using 10-K filings and domain-specific financial dictionaries.