System 03
Contextual AI Geocoding for Evidence Research
Put the evidence
on the map.
"How can contextual AI improve evidence geocoding in slavery, displacement, and war scenarios?"
Nominatim Pleiades OpenHistoricalMap Spatial Coherence Score 9 Parallel Backends
A five-stage pipeline running nine parallel geocoding backends with seven-component scoring, six structured failure modes, and a Spatial Coherence Score.
Five-Stage Pipeline
From raw evidence text to a resolved, scored geometry.
01
Entity Extraction
Identifying place-names amidst noisy evidence streams using contextual NLP.
02
Parallel Querying
Nine services hit simultaneously: Nominatim, Pleiades, WHG, OHM, Geonames, and more.
03
Result Aggregation
Candidate geometries pooled and deduplicated with a seven-component scoring model.
04
Disambiguation
Narrative context prunes the result tree to the most spatially coherent candidate.
05
Failure Classification
Six structured failure modes distinguish solvable ambiguity from genuine evidential gaps.
Pipeline in Action
Topodex geocoding pipeline being deployed against live evidence corpora.