1. Overview

This document outlines the complete integration plan for adding multilingual phonetic search to the existing WHG stack (Django + PostgreSQL/PostGIS + Elasticsearch). The goal is to support IPA-based matching, phonetic embeddings, and robust cross-lingual similarity without replacing existing infrastructure.

The system is split into two operational domains:

• Online stack (DigitalOcean)
  Django, PostgreSQL/PostGIS, Elasticsearch indices, query-time IPA conversion, and real-time search.

• Offline phonetic pipeline (Pitt CRC)
  Bulk IPA generation, embedding creation, Siamese model training, and ingestion of enriched data back into Elasticsearch.

Reference: Full technical background in WHG Place Discussion #81