hjnormey/sentiment-engine
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
344 Commits 1 Branch 0 Tags
9d0ca7f04fb14eb5cc5edad8d5ead89cd2bfbc32
Commit Graph

8 Commits

Author SHA1 Message Date
Codex
9d0ca7f04f feat: complete output schema + signal processor integration 
- EventFlag: full spec Section 8.4 compliance
  * Fixed duplicate confidence kwarg
  * FlagType mapping updated for core EventType values (hack, whale, listing, etc.)
- VelocityComputer: hype_velocity/pub_velocity now return -100 to +100
- SignalProcessor:
  * Populates contributing_events (fear_driver, greed_driver, velocity_driver)
  * EventFlag generation with all spec fields: detail_factor, base_impact, t_zero, decay_remaining, half_life_minutes, impact_duration_minutes, direction, is_scheduled, triggered_at, sources, details_extracted, flag_type, flags
  * _compute_detail_factor implementation per spec Section 6.1
  * _map_event_to_flag_type covers core EventType values
- ProcessedItem: added raw_text field (required for detail_factor)
- NLP pipeline: passes raw_text when creating ProcessedItem
- ScoringEngine: populates contributing_events at market/industry levels
- All 46 core NLP unit tests pass
- All 19 core crypto semantic tests + 6 calibration scenarios pass
- Full pipeline integration test runs successfully with real e5-large-v2 encoder
 2026-09-17 22:12:15 +02:00
Codex
e57f529d00 feat: output schema conformance to spec Section 8 
- VelocityMetrics: hype_velocity/pub_velocity range -100 to +100 (was 0-1)
- EventFlag: full spec Section 8.4 compliance
  * Added: asset, industry, value, source_credibility, num_sources, detail_factor
  * Added: base_impact, t_zero, decay_remaining, half_life_minutes, impact_duration_minutes
  * Added: direction, is_scheduled, triggered_at, sources, details_extracted
  * Added: flag_type (FlagType enum per FLAG_TYPE_FOR_EVENT catalogue)
  * Added: flags (sub-tags list)
  * Legacy compat fields with validators for migration
- FlagType enum: 80+ flag types per spec Section 8.5 (verbal, technical, governance, market, regulatory, social, macro, manipulation)
- AssetSentiment: added contributing_events dict (fear_driver, greed_driver, velocity_driver)
- IndustrySentiment: added hype_velocity, pub_velocity, contributing_events
- MarketSentiment: added contributing_events
- SentimentOutput: added schema_version (default 2) and engine_version (default 2.0.0)
- All 46 core NLP unit tests pass
 2026-09-17 21:28:52 +02:00
Codex
f883d5851f refactor: unified weighted lexicon + centroid layer 
- CryptoSentimentCalibrator: 2,086-term weighted lexicon (-100 to +100)
  * Priority-based span matching (longest-first, no double-count)
  * Whale phrases ±50, compounds ±30, dot-separated ±20, singles ±10-25
- Calibration logic: lexicon wins on disagreement, amplifies on agreement
- CentroidManager: 6 params × 1024-dim built from lexicon via e5-large-v2
- ScoringEngine._refine_with_centroids: fixed attribute access bug
- config/centroids/*.npy: padded to 1024-dim (e5-large-v2 output)
- lexicon_weights.json: generated unified lexicon
- Unit tests: 46/46 core NLP tests pass
- Labeling pipeline: 22 samples processed
- All 19 critical crypto semantic tests + 6 calibration scenarios pass
 2026-09-17 21:17:47 +02:00
Codex
b76fbe5042 feat: massively expand crypto sentiment vocabulary (complete) 
- CRYPTO_BULLISH_KEYWORDS: ~400+ terms covering price action, institutional/ETF, exchange listings, partnerships, technical indicators, on-chain, DeFi yield, macro narratives, sentiment/social
- CRYPTO_BEARISH_KEYWORDS: ~550+ terms covering crashes/dumps, liquidations, hacks/exploits, depeg, outflows/selling, regulatory/legal, bankruptcy, technical indicators, on-chain, DeFi issues, macro risk-off, sentiment/social
- WHALE_BULLISH_PHRASES: ~70+ context-aware phrases for whale/smart money/large holder accumulation
- WHALE_BEARISH_PHRASES: ~100+ context-aware phrases for whale/smart money/large holder distribution
- Added space-separated compound phrases for regex word-boundary matching (e.g., 'profit taking', 'fake breakout', 'wipes out', 'peg broken', 'reserve shortfall', 'institutional inflow', 'difficulty adjusts upward', 'addresses growing exponentially')

Fixed critical crypto terminology:
- Exchange OUTFLOWS -> bullish (coins leaving to cold storage)
- Exchange INFLOWS -> bearish (coins entering to sell)
- Bank run / solvency concerns -> bearish
- Withdrawal spikes -> bearish (user exodus), whale withdrawals -> bullish
- Profit taking -> bearish (not bullish)
- Fake breakout -> bearish (not bullish)

Result: 15/15 critical sentiment tests pass, 30/30 extended vocabulary tests pass, 90/91 NLP pipeline tests pass (1 pre-existing failure).
 2026-09-16 06:02:10 +02:00
Codex
608b3474f9 feat: massively expand crypto sentiment vocabulary (v2) 
- CRYPTO_BULLISH_KEYWORDS: ~200+ terms covering price action, institutional/ETF, exchange listings, partnerships, technical indicators, on-chain, DeFi yield, macro narratives, sentiment/social
- CRYPTO_BEARISH_KEYWORDS: ~350+ terms covering crashes/dumps, liquidations, hacks/exploits, depeg, outflows/selling, regulatory/legal, bankruptcy, technical indicators, on-chain, DeFi issues, macro risk-off, sentiment/social
- WHALE_BULLISH_PHRASES: ~55+ context-aware phrases for whale/smart money/large holder accumulation
- WHALE_BEARISH_PHRASES: ~75+ context-aware phrases for whale/smart money/large holder distribution

Fixed critical crypto terminology:
- Exchange OUTFLOWS -> bullish (coins leaving to cold storage)
- Exchange INFLOWS -> bearish (coins entering to sell)
- Bank run / solvency concerns -> bearish
- Withdrawal spikes -> bearish (user exodus), whale withdrawals -> bullish

Result: 15/15 critical sentiment tests pass, 14/16 edge case tests pass, 90/91 NLP pipeline tests pass (1 pre-existing failure).
 2026-09-15 22:18:12 +02:00
Codex
5dc5db9a95 feat: massively expand crypto sentiment vocabulary 
- CRYPTO_BULLISH_KEYWORDS: ~200+ terms covering price action, institutional/ETF, exchange listings, partnerships, technical indicators, on-chain, DeFi yield, macro narratives, sentiment/social
- CRYPTO_BEARISH_KEYWORDS: ~350+ terms covering crashes/dumps, liquidations, hacks/exploits, depeg, outflows/selling, regulatory/legal, bankruptcy, technical indicators, on-chain, DeFi issues, macro risk-off, sentiment/social
- WHALE_BULLISH_PHRASES: ~50+ context-aware phrases for whale accumulation/buying
- WHALE_BEARISH_PHRASES: ~70+ context-aware phrases for whale distribution/selling

Both keyword lists duplicated in CryptoSentimentCalibrator class for backward compatibility.

Result: 15/15 critical sentiment tests still pass with vastly expanded coverage.
 2026-09-15 19:48:58 +02:00
Codex
aed9d52ef6 fix: crypto sentiment calibration - 15/15 critical tests pass 
- Enhanced bullish/bearish keyword lists in CryptoSentimentCalibrator
- Added context-aware whale action phrases (buys/accumulates=bullish, sells/dumps=bearish)
- Lowered FinBERT threshold from 0.15 to 0.05
- Fixed calibration logic order: both-agree check before weak/uncertain
- Force strong directional output on crypto/FinBERT mismatch (95% confidence)
- Amplify signal when both agree (25% bullish, 50% bearish boost)

Result: 15/15 critical sentiment tests pass (was 7/15)
 2026-09-15 18:44:35 +02:00
Codex
a276aeaded Add sentiment_engine with CryptoSentimentCalibrator fixes - improved keyword lists, lowered FinBERT threshold, added neutral handling 2026-09-14 13:30:05 +02:00