Arabic NLP Destination Input - Translating Syrian Dialect Text and Voice into GPS Coordinates using NLP technologies

This presentation documents the design of an Arabic NLP feature built as part of a public transit routing application for Damascus, Syria, enabling users to express travel destinations in natural Syrian colloquial Arabic — including dialect text, local nicknames, spelling variants, and relative spatial descriptions — and receive structured GPS coordinates ready for a routing engine. The pipeline covers Arabic text normalization using CAMeL Tools, Named Entity Recognition extracting AREA, LANDMARK, RELATION, and MICRO_DIRECTION entities, hierarchical geocoding against a curated PostGIS Damascus place and alias database, and confidence-based disambiguation that auto-resolves above 0.85 and returns ranked Arabic candidates between 0.60–0.84. The system is deployed as a fully local Python FastAPI microservice with zero cloud dependency, integrated with a Node.js/Fastify backend and PostgreSQL/PostGIS spatial database. Target metrics justified against eight reviewed works include NER F1-score ≥ 78%, place resolution accuracy @1 ≥ 70%, place resolution accuracy @3 ≥ 88%, and API latency under 300ms p95.This work was conducted at Arab International University (AIU), Syria. The official website of the university is: https://www.aiu.edu.sy