ARCH — an educational media project focusing on lesser-known architectural structures and urban environment details. The project highlights interesting buildings, facades, doors, mosaics, and other elements that often go unnoticed.The main goal — to help the audience look at the city more closely, learn about architectural history, and independently explore their surroundings.Core formats: architectural reports, building histories, detail analyses, themed collections, and urban routes.

Five primary groups of sources are used to prepare the materials:

1) Cultural Heritage Object Registers. 2) State and City Archives. 3) Historical Maps and Photographs.

1) Books and articles on the history of architecture. 2) Research by architects and art historians. 3) Educational materials from universities.

1) Yandex Maps and 2GIS. 2) OpenStreetMap. 3) Historical city maps.

1) Archival photographs and postcards. 2) Museum and library collections. 3) Contemporary building photography.

1) Photographs and observations. 2) Proposals for research subjects. 3) Stories of city residents.

User-generated materials are used as a basis for investigation, but not as the sole confirmation of facts.

Core principle: the machine assists in collecting and organizing information; the human is responsible for its accuracy and semantic interpretation.

For each object, a separate entry is created with the following fields:

Tools: API, search engines, RSS, Web Scraping.Objective: To search for and gather information on architectural objects.Guidelines:Use official and open-source materials.Preserve links and data retrieval dates.Distinguish between primary and secondary sources.Do not assume that found information is automatically reliable.

Tools: OCR, Google Lens, Computer Vision.Objective: Extract text from archival documents, photographs, and inscriptions.Guidelines:Verify recognized names, dates, and addresses.Account for errors when working with older documents.Retain originals for cross-referencing.Do not use results without manual verification.

Tools: NLP, language models, LLM.Objective: Extracting key information and classifying entities.Guidelines:Extract names, addresses, dates, and people’s names.Group entities by category.Use AI for initial descriptions.Verify facts and terminology using specialized sources.Do not publish generated information without verification.