Yapay Zeka Destekli Geocode Servisi (AI-Powered Geocode Service)
- Customer
- Istanbul Metropolitan Municipality
- Project manager on the customer side
-
- Year of project completion
- 2024
- Project timeline
- September, 2022 - October, 2024
- Project scope
- 16320 man-hours
- Goals
-
AI-Powered Geocode Service was developed to accelerate Istanbul’s spatial data-based management and to provide residents and institutions with more accurate and accessible address, coordinate, and parcel-based queries through an AI-supported approach. The primary objective of the project is to establish a standardized, reliable, and high-performance infrastructure for accessing spatial information across the city.
AI-Powered Geocode Service analyzes user inputs with intelligent algorithms, automatically correcting misspelled addresses and using AI-based fuzzy search methods to deliver accurate results when exact matches are not available.Generating responses in an average of just 100 milliseconds, the system ensures the most accurate interpretation of urban data. In line with Istanbul’s smart city vision, it provides highly accurate, location-based solutions for both individual and institutional users.
- Project Results
-
Across Istanbul, spatial data access times have been reduced to the millisecond level. The system successfully identifies correct addresses even in the presence of spelling errors, significantly improving accuracy.
In institutional services, address resolution processes have become substantially faster, enhancing workflow efficiency. Through popularity scoring derived from user interactions, the system has evolved into a self-optimizing structure over time.As a result, location-based decision-making processes in areas such as urban planning, transportation, and emergency management have been strengthened, enhancing overall operational efficiency.
The uniqueness of the project
AI-Powered Geocode Service is a comprehensive spatial search infrastructure developed entirely in-house by the Istanbul Metropolitan Municipality (IMM), utilizing its own human resources and technical capacity without external dependency. It integrates address, coordinate, and parcel-based searches within a single system.
With its powerful search mechanism, the project stands out in terms of both performance and accuracy. AI-supported correction and similarity algorithms interpret user inputs and dynamically generate the most relevant results. The system continuously learns from user interactions over time, using popularity scores to prioritize frequently searched addresses. This innovative approach holds unique value not only from a technical standpoint but also in terms of sustainable urban management. The system, which simplifies Istanbul’s complex spatial structure and makes it accessible to all, constitutes a cornerstone of the smart city infrastructure of the future.
- Used software
-
AI-Powered Geocode Service is built on a robust software ecosystem and modern technological infrastructure. Within the scope of the project, PostgreSQL and PostGIS were used for database management and spatial analysis, ensuring the accurate processing of large-scale geographic data. Through ElasticSearch integration, exceptional search performance has been achieved across address, coordinate, and parcel-based queries.
At the application layer, Python and FastAPI technologies were used together to create a scalable, reliable, and highly accessible service architecture. The AI-based algorithms at the core of the service detect and correct user input errors, significantly improving search accuracy. In spatial computations, the GIS infrastructure supports multiple coordinate transformation systems (EPSG 4326, 3857, 5254), ensuring consistency by converting data from various projections into a unified reference system. All these processes are executed through a server-based high-performance API, capable of responding to numerous queries simultaneously within milliseconds.Additionally, the system includes logging and score-based popularity analysis modules that analyze user behavior. By prioritizing frequently used addresses, the service continuously enhances the overall user experience.
- Difficulty of implementation
-
Normalizing millions of address records in a large metropolitan like Istanbul, managing projection discrepancies, and optimizing the system for real-time performance posed significant technical challenges.
The integration of multiple data sources, extensive performance testing, and ensuring high-accuracy operation of AI models required substantial engineering effort. Additionally, to design a system capable of correctly interpreting inaccurate user inputs, Natural Language Processing (NLP) techniques and similarity algorithms had to be specifically adapted for the project.
- Project Description
-
AI-Powered Geocode Service is an innovative solution developed entirely with IMM’s own resources to accelerate and improve the accuracy of spatial data access across Istanbul. The system enables users to retrieve precise results for address, coordinate, or parcel-based queries through a single search interface.
Equipped with advanced artificial intelligence algorithms, the service detects spelling errors, performs semantic-level corrections, and accurately identifies addresses even in cases of incomplete data. The system is based on a PostgreSQL infrastructure and utilizes ElasticSearch capabilities in certain scenarios. It ensures high performance by integrating full-text search with fuzzy matching techniques.For coordinate-based queries, the system automatically detects the input’s projection and converts it to EPSG 4326. It then identifies the nearest road, building entrance, or point of interest (POI). For address-based searches, results are ranked according to similarity ratio, popularity score, spatial hierarchy, and category values.By analyzing user behavior, the system continuously improves itself over time. Frequently searched addresses are prioritized, making it easier for residents to access the information they need. AI-Powered Geocode Service is designed with a scalable architecture that not only meets today’s needs but also supports the data-driven urban management of the future.
- Project geography
-
AI-Powered Geocode Service contains over 5 million records across Istanbul, including addresses, buildings, streets, neighborhoods, districts, and POI.
The system is scaled for use not only by individual users, but also by municipal departments, affiliated companies, and partner institutions. Its infrastructure is designed to be expandable, enabling potential implementation in other cities across Türkiye in the future.
- Additional presentations:
- AI-POWERED GEOCODE SERVICE_PRESENTATION.jpgAI-POWERED GEOCODE SERVICE_PROJECT_PRESENTATION.jpg
Events