Build a Macroeconomic Sentiment Extraction Pipeline with Google ADK
Create an automated sentiment and macro indicator extraction pipeline for wealth managers following global inflation announcements. Using Google ADK, your agent will process speeches and CPI data, extract interest rate signals, and output structured portfolio guidance.
What you are building
Develop a Google ADK agent to extract interest rate sentiment and economic indicators from central bank speeches and inflation releases.
How work is evaluated
Evaluate classification accuracy of monetary policy stance (Hawkish/Dovish/Neutral).
Datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
stance_classification_check
Stance matches expected hawkish classification
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
stance_score_mae
Mean absolute error of numerical stance score • target: 0.05 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Learning goals
Build a Google ADK agent for continuous text intelligence parsing
Extract hawkish/dovish sentiment polarity scores from official speeches
Correlate economic indicators (CPI) with monetary policy expectations
Format advisory outputs for wealth management relationship managers
Resources
Transcripts of central bank speeches and monetary authority press releases paired with analyst consensus scores.
Challenge setup
Evaluate classification accuracy of monetary policy stance (Hawkish/Dovish/Neutral).
Challenge input
JSON containing transcript_text and speaker
Google ADK
Google Agent Development Kit for Vertex AI integration.
Evaluated output
JSON with stance ('HAWKISH'|'DOVISH'|'NEUTRAL'), stance_score (-1.0 to +1.0), key_phrases
- Stance matches expected hawkish classification
- Mean absolute error of numerical stance score • target: 0.05 • range: 0-1
- Benchmark: FinNLP_Sentiment_Bench
- Stance Score Mae target: 0.05
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Shows the first task, up to three tools, and up to three evidence items.
View technical recipe
Configured tools
- Google ADK · Required
- Google · Optional
- Guidance · Optional
Evaluation contract
- stance_classification_check · Weight 1
- stance_score_mae · Weight 1
Metrics
- stance_score_mae · Target 0.05
Benchmarks
- FinNLP_Sentiment_Bench
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing transcript_text and speaker
Output: JSON with stance ('HAWKISH'|'DOVISH'|'NEUTRAL'), stance_score (-1.0 to +1.0), key_phrases
Execution harness: Python
Evaluation tools
- Google ADK · Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.