Vertical · Multilingual newsrooms
One deployment, every official language your readers speak.
If your newsroom has to address readers in more than one language, the standard answer is to run a separate comment system per language and accept the fragmentation. There is a better one. RTS does it on Logora for the four Swiss national languages plus English, on a single platform. The same setup applies anywhere multilingualism is the rule, not the exception.
«dialogue» · same debate, every Swiss language
Single deployment, locale frontends, native moderation
/fr/
Faut-il maintenir le service public audiovisuel par redevance ?
/de/
Soll der öffentliche Rundfunk per Beitrag finanziert bleiben?
/it/
Mantenere il finanziamento del servizio pubblico tramite canone?
/rm/
Sustegnair il servetsch public via taxa?
One backend · one moderation pipeline · five reader experiences
- 5 languages served on a single deployment at RTS · dialogue.rts.ch French, German, Italian, Romansh, English · observable
- 1 backend one moderation pipeline, one reader database, one set of editorial rules Logora architecture
- Native moderation per language, no machine translation in the decision path AI + human moderation pipeline
Your reality today
What your day actually looks like.
-
One comment system per language
You ended up with three or four parallel deployments because no vendor supported your full language matrix. The moderation rules drift between them. The reader graph is fragmented. The editorial standards diverge. The compliance posture (GDPR, national data protection) has to be audited per stack.
-
Minority languages always lose
Welsh, Romansh, Frisian, Basque, Galician, Irish, Catalan, Luxembourgish, the smaller official language always gets a degraded reader experience. UI in the dominant language, moderation through translation, slower feature parity. The community notices, and it is a credibility cost for a public-service or national newsroom.
-
Cross-linguistic moderation is brittle
Toxic content in the smaller language often escapes detection because the AI was trained on the dominant one. When a moderator does intervene, it is through machine translation, which loses the cultural context that makes the call defensible.
What changes
With Logora, here is what shows up in your reports.
-
One backend, locale frontends per language
Single deployment, locale-specific URLs (/fr/, /de/, /it/, /rm/, /en/ on RTS), localized UI per language. The reader sees the platform in their language. The newsroom operates one tooling, not five.
-
Native moderation in every language you ship
Reviewers handle each language natively. The AI toxicity pipeline is calibrated per language, not routed through English. The decisions stay defensible in the language they were made in.
-
A consistent editorial standard across communities
Same debate format, same neutrality rules, same accessibility baseline. The German-speaking, French-speaking and Italian-speaking readers of a Swiss public broadcaster all get the same product, exactly the constraint a public-service mandate imposes.
The modules that matter to you
Where to look first.
-
Debates module
Structured for/against debates work across languages because the format does. Arguments rank by reader vote per side, not by editor selection, which is language-neutral.
Explore -
AI moderation
Toxicity classifier calibrated per language. Default thresholds tuned for public-service editorial standards, configurable per locale.
Explore -
SSO & gamification
One reader account across all language frontends. A Geneva-based bilingual reader keeps one identity whether they read the French or the German edition.
Explore -
RTS case study
The live reference: dialogue.rts.ch on a single Logora deployment, five locales, observable in production.
Explore
Verified, on production
dialogue.rts.ch, five locales.
Each URL renders in its own language with a properly localized tagline, not a translation. The Romansh frontend (/rm/) is particularly significant: Romansh is spoken by about 60,000 people in Graubünden, and very few digital products outside the SRG SSR perimeter ship UI in this language at all. If Logora handles Romansh, the moderation and the UX work for the languages you are likely to need.
«dialogue» · same debate, every Swiss language
Single deployment, locale frontends, native moderation
/fr/
Faut-il maintenir le service public audiovisuel par redevance ?
/de/
Soll der öffentliche Rundfunk per Beitrag finanziert bleiben?
/it/
Mantenere il finanziamento del servizio pubblico tramite canone?
/rm/
Sustegnair il servetsch public via taxa?
One backend · one moderation pipeline · five reader experiences
"If your moderation can handle Romansh, it can handle anything. The hard part of multilingual is not the dominant languages, it is the smaller ones, where the standard tools quietly degrade."
Quick answers
What we hear on every call.
-
We run a French and English newsroom in Canada. Does this actually apply?
Yes. Canada is a textbook fit: two official languages, federal-level public broadcaster obligation, two reader communities that should not be treated as a translation of each other. The RTS architecture (one backend, locale frontends, native moderation per language) maps directly to a Radio-Canada / CBC, La Presse / Toronto Star, or Le Devoir / The Globe and Mail setup.
-
How does the same platform handle four or five sets of moderation rules?
Moderation thresholds are configurable per locale. You decide the toxicity rules, the publishing rules, and the escalation rules in each language. The defaults match public-service editorial standards out of the box, and the publisher can tune per locale during onboarding. The reader-facing experience stays consistent because the format does not change; only the language and the threshold calibration do.
-
Can a single reader account move between language frontends?
Yes. SSO is identity-first. A bilingual reader in Brussels keeps one Logora account across the Dutch, French and German frontends of a Belgian newsroom. Reading history, badges and submitted arguments travel with the user. This is exactly the kind of cross-community feature that fragmented per-language deployments make impossible.
-
What about minority languages with limited AI training data?
For lower-resource languages (Romansh, Welsh, Frisian, Luxembourgish, Catalan), default toxicity thresholds are conservative, escalation to human moderation is faster, and we work with the publisher to calibrate the rules against actual community standards during the first weeks in production. The principle: never let the AI degrade quality just because a language has less training data.
-
Is the architecture observable somewhere?
Yes. dialogue.rts.ch, with locale paths /fr/, /de/, /it/, /rm/, /en/. Open all five in different tabs, each is the same product in a different language. We will walk you through how the deployment is wired in a 60-minute call.
A platform that respects every reader community.
A 60-minute call with Pierre or Henry. Bring your language matrix, your current moderation team structure, and your editorial neutrality constraints. We bring the RTS reference, the architecture diagram, and the moderation calibration plan for your specific language mix.