CODE: Red — media that explains which rare species live near you and how to behave if you encounter them.
Data Sources
Fundamental Sources:
Official Website of the Ministry of Natural Resources of Russia (Red Data Book section). What we use: Current list of species listed in the Russian Red Data Book (Order No. 162 dated 03/24/2020), status changes (e.g., the transition of the Sarus Crane to Category 1). Purpose: All sections. This is the sole legitimate source for the phrase «This species is protected by the state.»The Red Data Book of Russia (electronic version / reference portals). What we use: Species description, range, limiting factors, rarity category. Purpose: «Red Data Booker of the Day» (fact-finding), «Myth or Fact» (statement verification).Regional Red Data Books (administration websites / regional environmental ministries). What we use: Local lists of species not in the federal book but protected in a specific region. Purpose: «My Region» section (primary source).Scientific and Expert Sources (for fact-checking and in-depth posts)
Databases of Specially Protected Natural Areas (SPNAs) and Reserves. What we gather: Species lists found in specific territories, data from camera traps, and rehabilitation histories of particular animals. For what: «Return Heroes» (real-life cases of bison, Asiatic leopards), and «Encountered in Nature» (clarifying animal behavior in specific locales).
Scientific publications and range databases (e.g., WWF GIS systems / «Faunistics»). What we gather: Interactive maps of rare species habitats. For what: Visuals for videos («Here is a spot on the map where this animal lives»). It helps in the «My Region» segment to show the range.
Real-time and User-Generated Sources (for Engagement and Live Content)
iNaturalist (global citizen science platform). What we take: Geolocation data from sightings of rare species, validated by the community. This is «raw» data from people who have personally encountered an animal. For what: «Sighting in the Wild» (real scenarios: «here a person encountered a snake—what did they do?»), and «My Region» (crowdsourcing: «send in your observations»). This source requires verification as it is unofficial.
Reports from Rosprirodnadzor and rehabilitation centers. What we take: Statistics on the seizure of endangered animals and rescue stories. For what: «Heroes of Return» (statistics on animals rescued), and «What Everyone Can Do» (where to call if you find an injured animal).
Dividing responsibilities between machine and human in content production
The machine does the heavy lifting—it collects, sorts, and generates. The human does the smart work—it reviews, explains, and is responsible for content safety.
Phase 1. Data Collection and Aggregation (Machine Processing)
Step 1.1. Source Parsing. A parser script automatically gathers raw data from various databases (Ministry of Natural Resources, iNaturalist, rehabilitation center reports). The system tracks updates to species statuses and new user observations. Step 1.2. Sorting and Tagging. The machine automatically categorizes the collected data (for instance, an observation from iNaturalist is placed in the «My Region» folder), assigns tags to the species, and generates a basic record (name, status, region).Stage 2. Fact-Checking and Expertise (Human Review)
Step 2.1. Data Verification. The editor reviews raw machine data. They filter out false sightings from iNaturalist and cross-reference the species status with the current Red Book (federal or regional). Step 2.2. Safety Protocol Check. An expert (or the editor themself) reviews the guidelines for the «Wildlife Encounter» section, ensuring that advice regarding behavior when encountering wild animals (e.g., «take three steps back» or «do not touch the chick») is ecologically sound and safe for humans.
Phase 3. Content Generation and Preparation (Hybrid Processing)
Step 3.1. Drafting (Machine). The Neural Network (LLM) receives a verified fact and generates a draft post text according to a specified template. The image generator (Midjourney/DALL-E) creates background visuals based on the provided prompt. Step 3.2. Semantic Editing (Human). The editor rewrites the «robotic» text generated by the neural network, adding empathy and removing bureaucratic jargon. The main goal is to make the text lively while preserving factual accuracy.Stage 4. Platform Adaptation (Hybrid Processing)
Step 4.1. Automatic Cropping (Machine). Algorithms automatically segment the finished text into short titles for videos (Reels/Shorts) and layout the carousel for Telegram. Step 4.2. Visual Assembly and Review (Human). The designer/editor checks the overlaid text on the video (to ensure it doesn’t obscure important details in the frame), verifies font readability, and approves the final look of the post before publication.Phase 5. Post-Moderation and Feedback (Human Processing)
Step 5.1. Comment Moderation. A human responds to subscriber comments and collects new user photos for the «My Region» feature. Step 5.2. Database Update (Machine). Facts and cases approved by the human are returned to the project database to form the archive.Guidelines for Automated Services (Schematic)
