Our Data

The most comprehensive Moroccan & African media dataset.

Harch AI processes over 5 million articles per day from 30+ Moroccan and African media sources, in 3 languages, with 100M+ entities labeled daily. Here's what we track — and how we get it.

The numbers

A data operation at scale.

5M+

articles ingested per day

30+

Moroccan & African sources

100M+

entities labeled per day

120+

languages translated

Sources

Where the data comes from.

Moroccan Media (FR)

TelQuel

News + Business

42

Medias24

Business

38

Aujourd'hui Le Maroc

News

32

Le Site Info

News

24

Barlamane

Politics + Business

18

Al Ahdath

News

14

Infomediaire

Business

22

Moroccan Media (AR)

Hespress

News (AR)

28

Sabah Press

News (AR)

12

Al Maghrib Today

News (AR)

9

Morocco World News (EN)

News (EN)

15

Yabiladi

News + Community

7

African Business Media

Financial Afrik

Pan-African Business

24

Africa News

Pan-African News (EN)

16

Jeune Afrique

Pan-African Magazine

14

African Business

Business Magazine

11

The Africa Report

Business (EN/FR)

13

Aggregators + Regulatory

Google News Morocco

Aggregator (FR/AR/EN)

96

Bank Al-Maghrib

Regulator

4

AMMC

Markets Regulator

3

ANRT

Telecom Regulator

2

ONSSA

Food Safety

1

Data pipeline

From RSS to insight.

◆

Google News aggregation

Our primary source — aggregates from 50+ Moroccan media that block direct RSS. Alias-based queries capture every mention across all known company names.

▲

Direct RSS feeds

11 working RSS feeds from major Moroccan and African publications. Polled every 15 minutes. URL-based deduplication.

◉

Regulatory monitoring

Manual + automated tracking of BAM, AMMC, ANRT, ONSSA circulars and decisions. Pushed to platform within 1 hour of publication.

⌬

AI engine testing

Hourly queries to ChatGPT, Perplexity, Gemini, Claude, Copilot, Mistral, Grok with your industry's top prompts. Track citation, position, sentiment.

◇

Trilingual NLP

Articles handled natively in French, Arabic (MSA), English. NLP Darija pour les commentaires Hespress, forums, WhatsApp et TikTok — la couche UGC que aucun concurrent ne couvre. fastText + spaCy + HarchIQ.

◆

Knowledge graph

100M+ entities linked: companies, people, products, topics, events. Updated continuously as new articles arrive.

Data quality

How we keep it clean.

Deduplication

URL + title similarity + content hash. Same article syndicated across 3 outlets counts as 1 article, 3 mentions.

Relevance scoring

Every article gets a relevance score (0-100) based on entity density, position in article, and topic match.

Source authority

Each source has an authority score (1-10) based on reach, journalistic standards, and editorial independence.

Sentiment confidence

Every sentiment classification has a confidence score. Below 70% triggers human review.

Language verification

Articles flagged as 'AR' but containing >30% French are reclassified as code-switching for separate analysis.

Entity disambiguation

BMCE (bank) vs BMCE (building) — context-aware disambiguation using surrounding entity co-occurrence.

See the data live.

Explore our Risk Tracker and Harch 100 ranking — both powered by this dataset.

View Risk Tracker →