v1.0.0 to v1.0.0

---

name: central-bank-communication-nlp-analysis

description: Quantitative NLP engine for extracting hawkish vs. dovish sentiment from

central bank statements, featuring n-gram negation handling and standardized scoring.

domain: Alpha Generation

subdomain: Alternative Data

tags:

- nlp

- sentiment-analysis

- central-bank

- fomc

- hawkish

- dovish

brokers_frameworks:

- Generic NLP

author: algo-trading-skills-contributors

license: Apache-2.0

---

## When to Use

Use this skill when processing central bank communications (FOMC statements, ECB press conferences) to generate quantitative trading signals. The engine translates unstructured text into a normalized Hawkish-Dovish index score. It acts as the feature-extraction layer for macro-economic trading models (e.g., trading FX, interest rate swaps, or equity indices based on policy surprises).

## Prerequisites

- Clean, parsed text of central bank communications.

- Basic NLP pre-processing capabilities (tokenization).

## Workflow

1. **Ingestion**: The system receives a raw text document (e.g., the latest FOMC statement).

2. **Pre-processing**: The engine tokenizes the text into sentences and words, converting to lowercase and stripping punctuation.

3. **Lexicon Matching**:

- The text is scored against specialized financial dictionaries (Hawkish vs. Dovish terms).

- **Crucial Step**: The engine checks for negation (e.g., "not tighten", "less accommodative") within a defined n-gram window to invert the score.

4. **Scoring**: A net sentiment score is calculated: $(Hawkish - Dovish) / Total\_Words$.

5. **Signal Generation**: The net score is compared to the historical baseline to detect "surprises" (hawkish or dovish shifts).

> Full procedure: see `references/workflows.md`.

> Standards reference: see `references/standards.md`.

> Printable pre-flight checklist: see `assets/checklist.md`.

## Common Pitfalls

- **Ignoring Negation**: Scoring "will not raise rates" as Hawkish because it contains the word "raise". Negation windows (looking 2-3 words back) are mandatory.

- **Using Generic Lexicons**: Using standard sentiment dictionaries (like VADER) which score "liability" or "cut" incorrectly in a central bank context. You must use specialized macro-finance lexicons.

- **Absolute vs Relative Scoring**: Trading directly on an absolute hawkish score. Markets price in expectations. The true signal is the *delta* between the current score and the previous statement's score.

## Verification

- Feed a mock FOMC statement into the engine. Include a negated phrase like "we do not plan to increase rates." Verify the engine correctly scores this as Dovish (or neutral) rather than Hawkish.

- Run `python scripts/test_central_bank_nlp_engine.py`.

## Related Skills

- `global-macro-economic-calendar-integration`

- `earnings-call-transcript-nlp-signal-research`