Financial NLP Engine

Financial documents mix narrative language with semi-structured metrics, making consistent extraction difficult across reports and news formats.

The project was a team effort to turn raw financial text into auditable structured outputs that downstream analysis could consume.

• Raw earnings reports and news land in data/raw/, then flow through parsing modules that extract clean text and structured fields into data/processed/.
• Dedicated sentiment and extraction stages pull EPS, revenue, and other metrics, writing final JSON/CSV outputs to data/output/.
• Integration modules connect parsed signals to market data so downstream analysis can correlate language with price movement.

• Separate ingestion and parsing modules normalize heterogeneous reports and news documents.
• Sentiment and metric-extraction stages produce structured EPS, revenue, and language signals.
• Integration modules join extracted outputs with market data and export JSON or CSV artifacts.

• Modular stages make extraction errors easier to trace than one end-to-end script.
• Financial NLP needs explicit provenance because a plausible number without its source is not useful.

• How should extraction confidence be calibrated across document formats?
• Which language signals remain useful after controlling for already-public market information?