Telecommunications — FranceIA Data & Automatisation

A national telecom operator now qualifies every piece of content via AI, without losing control.

NLP and human-in-the-loop platform for the semantic qualification of editorial content of a national telecom operator. Tagger rules, continuous lexicon learning, quality dashboard, and monthly reporting.

Delivery model

Development center (CDR)

Via Access International

Typical duration

Renewable multi-year program

Team

AI architect, NLP developers, data engineers, QA, project manager

01Context

Significant volume of editorial content to qualify monthly
Complex business rules specific to the brand and regulation
Need for continuous lexicon learning (products, campaigns, events)
Requirement for full traceability of qualifications

02Intervention

Design of an NLP pipeline combining tagger rules and learning models
Human-in-the-Loop (HITL) loop for ambiguous case qualification
Real-time quality dashboard
Continuous lexicon learning mechanism from HITL decisions
Bias and confidentiality audits compliant with GDPR

03Deliverables

NLP platform in production with availability SLA
HITL console for editorial teams
Consolidated quality dashboard
AI governance documentation (quality, bias, confidentiality)
Monthly report on the qualified corpus

04Outcomes

Large-scale automated qualification with targeted human intervention
Continuous lexicon learning improving accuracy over time
GDPR compliance on content processing
Capacity to extend to new editorial use cases

Technologies used

Click a technology to explore our dedicated practice.

Python↗NLP↗PostgreSQL↗pgvector↗HITL↗Power BI↗Azure↗Claude↗

Related modernization paths

Path

GenAI in production (RAG) →

Path

AI agents via Copilot Studio →

Related expertise

IA Data & Automatisation →

Methodology

ATLAS methodology →

Does this context match yours?

We frame your program at Intake, with transparent budgeting. A short POC can be delivered in a few weeks before committing to a full program.