17 June 2026
Human-rights bureau · monitoring LGBTIQ+ rights
Bureau “We Are!”
Hate-speech monitoring and LGBTIQ+
rights-protection system in Ukraine
Evidence-based human-rights infrastructure: collecting, analysing and documenting discrimination and hate speech against LGBTIQ+ people during the full-scale war.
Project presentation for prospective donors
Director: Volodymyr Kosenko · EDRPOU 41591439 · lgbt.in.ua · monitor.lgbt.in.ua
Context
Mission and problem
The problem
During the full-scale war, hate speech and discrimination against LGBTIQ+ people are rising, yet scattered across social media, outlets and public actors’ statements. There is no systematic, methodologically sound record to rely on in courts, advocacy and reports.
Our response
We built an end-to-end monitoring system: automated collection of public materials, AI analysis, manual coding under a transparent methodology, and a public dashboard with verified data.
Monitoring
Daily documentation of hate-speech instances.
Analytics
Narratives, trends, actors, regional heatmaps.
Legal protection
ECtHR database, case law, legislation tracker.
Education
Methodology, publications, digital tools.
System status
The project in numbers
1,233
materials recorded in the system
1,371
ECtHR cases with SOGI annotations
87
verified and published records
1,042
materials analysed with NLP
1,135
materials processed with NER
232
corpus candidates pending
Methodological integrity. The gap between collected (1,233) and published (87) is deliberate quality control: no record goes public without manual review by a coder and a verifier. Accuracy over volume.
How it works
End-to-end monitoring pipeline
From automated collection of open sources to verified publication — every item passes six stages with quality control.
01
Collection
RSS, HTML, Telegram, YouTube, Google News, X/Reddit
02
Detection
Keywords, relevance filters, deduplication
03
AI analysis
NLP: sentiment, topic · NER: actors, entities
04
Coding
Narrative codes, severity 0–4, risk codes
05
Verification
Second-analyst review
06
Publication
Dashboard: aggregated safe data only
Telegram actorsYouTubeRSS / HTML outletsGoogle NewsX + RedditAll-Ukr. Council of ChurchesCommunity reportsArchiving (Wayback)
Trust in the data
Methodology and transparency
Narrative codebook
Standardised codes (war frame, cultural colonisation, gender ideology) classify rhetoric without reproducing hate speech.
Severity scale 0–4
Risk codes distinguish neutral mentions from direct calls to violence.
Neutral paraphrase
Instead of quoting offensive content — a neutral description; ethical and safe to reuse.
Dual control
Coder and verifier roles are separated; publication only after manual confirmation.
Privacy by design. Internal notes, personal data and beneficiaries’ identities never reach the public surface. Access to raw data is restricted by database-level security policies (RLS); anonymous users see only aggregated views.
What the world sees
Analytics products · monitor.lgbt.in.ua
Interactive dashboard
Daily dynamics, breakdowns by platform, narrative and region — with denominators and small-sample flags.
Regional heatmap
Geography of incidents across Ukraine’s regions.
ECtHR database (SOGI)
1,371 cases annotated by SOGI aspects, articles and principles — with HUDOC links.
Legislation tracker
Stages of key bills affecting LGBTIQ+ rights.
Case law
Rulings with SOGI markers, outcomes and links to ECtHR cases.
Export & citation
Aggregated data export (CSV/JSON) and methodology references.
WCAG 2.2 AA accessibility · table alternatives to charts · SRI protection of external scripts · automated quality checks in CI.
Engineering
Technology, infrastructure and AI
HTML / CSS / JSNetlifySupabase / PostgreSQL + RLSPython (NLP, NER)Hetzner VPSGitHub Actions CI/CDTwenty CRMBrevo
NLP pipeline
Sentiment, topic and demagoguery analysis; local models for privacy.
NER
Entity recognition: actors, organisations, locations for network analysis.
Hermes AI Agent
A secure Telegram advisor with 19 skills (monitoring, triage, digests, legal watch, donor packs). No access to code, files or DB mutations.
Security by design
JWT on server functions, restricted CORS, form protection, RLS on all tables, regular security audits.
Full list
Capabilities & features
- Daily auto-collection from 7+ source types
- Deduplication and filtering of irrelevant content
- Evidence archiving (Wayback) + link checks
- NLP: sentiment, topic, demagoguery
- NER: actors, organisations, locations
- Narrative codes and severity scale 0–4
- Dual verification before publication
- Public dashboard with live breakdowns
- Heatmap of Ukraine’s regions
- ECtHR database with SOGI annotations (1,371)
- Legislation tracker
- Ukrainian case law (SOGI)
- Community report form
- Legal-aid request form
- Export of CSV/JSON aggregates
- Weekly reports in Telegram
- Analytics integration (GA4, GSC)
- CRM for partners and donors
- Hermes AI Agent (19 skills)
- Accessibility WCAG 2.2 AA
- Bilingual (UK / EN)
- Video transcripts for analysis
Partnership
How support helps
The infrastructure is built and running. Donor support will scale reach, deepen analytics and strengthen legal work.
Expanding monitoring
More sources and actors; historical backfill to 2022 for multi-year trends.
ECtHR AI analytics
Automated annotation of ECtHR cases with modern models — requires API budget.
Analytics team
Coders and verifiers to process the corpus (over 950 in queue).
Advocacy & reports
Regular public reports, policy briefs and partner materials.
Why support this. We offer donors not a promise but a ready, transparent and ethical evidence system: reproducible methodology, privacy by design and a public dashboard with real-time results.
Get in touch
Let’s build evidence-based
rights protection together