Sentence-level diff of current vs prior monetary policy statements for FED, ECB, BOE, BOJ. NLP applied to word-level deltas and weighted by section matrix. net_vector_shift quantifies the hawkish/dovish pivot using Section Multipliers (e.g., Forward Guidance = 3.0x).
Refreshed: Syncing latest... - Source: Official CB statement archives - DB cache 6h TTL
Parsing CB Statements...
Methodology & Interpretation
1. Delta-First NLP Engine
Instead of running sentiment analysis on the entire statement, the engine uses Python's difflib.ndiff to isolate word-level token differences between the current and prior statements. Sentiment NLP is applied exclusively to the changed text (additions, deletions, and modifications).
2. Net Vector Shift
The core metric of a policy pivot. It is calculated as the sentiment score of added/modified text minus 50% of the sentiment of removed text (e.g., removing a hawkish sentence creates a dovish shift). A positive shift indicates a Hawkish pivot, while a negative shift indicates a Dovish pivot.
3. Absolute Stance Index
A normalized 0-100 scale representing the absolute tone of the current statement. 0 is fully dovish (easing), 50 is neutral, and 100 is fully hawkish (tightening). This is calculated by scoring the entire text independently of the diff.
4. Storage & Processing
Prior statements are permanently stored in the CondorEdge database. New statements are checked automatically; when a change is detected, the current statement is promoted to "prior", and a new diff is permanently recorded.