Use Cases
Practical applications of neuro-symbolic validation
The Logic Guard Layer validates AI-generated claims against authoritative external data sources. This page describes the two primary use cases implemented in this system, along with the APIs and claim types each supports.
Use Case A: Physical Measurement Data
Water Levels & Weather Data Validation
Validates claims about real-world physical measurements against official government data sources. This use case is critical for applications in hydrology, meteorology, environmental monitoring, and infrastructure management.
Data Sources:
PEGELONLINEBright Sky (DWD)Nominatim (optional)Supported Claim Types
| Predicate | Description | Data Source | Example |
|---|---|---|---|
hasWaterLevel | Water level at a gauging station | PEGELONLINE | 3.45 m at Cologne |
hasTemperature | Air temperature at a location | Bright Sky | 22.5 °C in Berlin |
hasWindSpeed | Wind speed measurement | Bright Sky | 15 km/h in Hamburg |
hasPrecipitation | Precipitation amount | Bright Sky | 2.3 mm in Munich |
exists | Entity existence verification | All adapters | Station "Köln" exists |
Example: Water Level Validation
"The water level at the Cologne gauging station was 3.45 meters on January 15, 2024 at 14:00."
Claim = ("Cologne", hasWaterLevel, 3.45, "m", "2024-01-15T14:00:00Z")
The validation process for this claim:
1
Extract
LLM parses text into structured claim
2
TBox Check
Verify: value >= 0, unit is meters
3
Resolve Station
Map "Cologne" to PEGELONLINE station ID
4
Query API
Fetch actual value from PEGELONLINE
5
Compare
Check if 3.45 m matches (within tolerance)
API Details: PEGELONLINE
PEGELONLINE is the official water level information system of the German Federal Waterways and Shipping Administration (WSV). It provides real-time and historical water level data for over 500 gauging stations across German federal waterways.
- Endpoint:
https://www.pegelonline.wsv.de/webservices/rest-api/v2 - Access: Public, no API key required
- Data: Water levels, station metadata, time series (15-minute intervals)
- World Assumption: Closed World (stations are exhaustively listed)
API Details: Bright Sky
Bright Sky is a free JSON API that provides easy access to weather data from the German Meteorological Service (Deutscher Wetterdienst / DWD). It simplifies access to DWD's open data offerings.
- Endpoint:
https://api.brightsky.dev - Access: Public, no API key required
- Data: Temperature, precipitation, wind, humidity, solar radiation
- World Assumption: Open World (not all locations may have data)
Use Case B: Procurement Document Data
EU Tender Notice Validation
Validates claims extracted from public procurement documents against the official EU procurement database. Essential for legal compliance, contract management, and procurement analytics.
Data Sources:
TED EuropaCPV Vocabulary (optional)Supported Claim Types
| Predicate | Description | Data Source | Example |
|---|---|---|---|
hasDeadline | Submission deadline for tenders | TED | 2024-03-15T12:00:00Z |
hasCPVCode | Common Procurement Vocabulary code | TED | 45233120-6 (Road construction) |
hasValue | Contract value/estimate | TED | 1,500,000 EUR |
hasPublicationDate | Official publication date | TED | 2024-01-20 |
exists | Notice existence verification | TED | Notice 2024/S 015-012345 exists |
Example: Tender Deadline Validation
"The submission deadline for tender notice 2024/S 015-012345 is March 15, 2024 at noon."
Claim = ("2024/S 015-012345", hasDeadline, "2024-03-15T12:00:00Z", null, null)
API Details: TED Europa
Tenders Electronic Daily (TED) is the official journal of record for European public procurement. It publishes approximately 750,000 procurement notices per year, representing contract opportunities worth hundreds of billions of euros.
- Endpoint:
https://api.ted.europa.eu - Access: Public for published notices (no API key for search/retrieve)
- Data: Notice metadata, deadlines, CPV codes, contract values, lot information
- World Assumption: Closed World (published notices are exhaustively indexed)
Claim-to-API Mapping
The following matrix shows which APIs validate which claim types:
| Claim Type | PEGELONLINE | Bright Sky | TED Europa | Local KB |
|---|---|---|---|---|
hasWaterLevel | YES | - | - | fallback |
hasTemperature | - | YES | - | fallback |
hasWindSpeed | - | YES | - | fallback |
hasPrecipitation | - | YES | - | fallback |
hasDeadline | - | - | YES | fallback |
hasCPVCode | - | - | YES | fallback |
hasValue | - | - | YES | fallback |
exists | YES | YES | YES | YES |
Error Modes by Use Case
Different use cases exhibit different failure patterns. Understanding these helps interpret validation results:
Use Case A: Measurement Data
| Error Type | Common Cause | Status |
|---|---|---|
| Station not found | Misspelled name, unofficial name, or station doesn't exist | NOT_FOUND_ABSENCE |
| Value mismatch | LLM hallucinated the measurement value | MISMATCH |
| Time out of range | Historical data not available for requested timestamp | NOT_FOUND_OUT_OF_SCOPE |
| Location ambiguous | Multiple stations match the location description | UNKNOWN |
Use Case B: Procurement Data
| Error Type | Common Cause | Status |
|---|---|---|
| Notice not found | Invalid notice ID, not yet published, or withdrawn | NOT_FOUND_ABSENCE |
| Deadline mismatch | LLM confused dates or used wrong timezone | MISMATCH |
| CPV code invalid | Malformed code or code doesn't exist in vocabulary | SCHEMA_VIOLATION |
| Value mismatch | Currency confusion, VAT inclusion/exclusion | MISMATCH |
Try It Yourself
Water Level Example
The water level at the Köln gauging station on the Rhine was
approximately 2.5 meters yesterday at noon.Weather Example
The temperature in Berlin reached 25 degrees Celsius on
July 15, 2024 at 2 PM.Procurement Example
The tender notice for road construction services in Bavaria
has a submission deadline of March 31, 2024 and uses CPV code
45233120-6.