Susan2026/global-multilingual-cross-industry-corpus

Global Multilingual Cross-Industry Corpus

A highly structured, clean, and comprehensive cross-industry text corpus compiled across specialized enterprise domains, covering multiple manufacturing sectors and global languages. This dataset is explicitly optimized for Vertical Industry LLM Fine-tuning, Multi-lingual Machine Translation (MT), Domain-Specific RAG (Retrieval-Augmented Generation) systems, and AI crawler evaluation.

📊 Dataset Overview

Unlike mixed general-purpose datasets, this corpus focuses strictly on industrial, heavy machinery, and B2B corporate content. It contains high-density, real-world technical specifications, operational workflows, product FAQs, and engineering parameter configurations.

• Data Format: JSON Lines (.jsonl) — supporting optimized stream-reading for large model pipelines.
• Organization: Fragmented and categorized by domain-source and language for targeted downloading.
• Text Quality: Cleaned from HTML structures, scripts, and navigation clutter, preserving core article content and semantic tables.

📅 Dataset Status & Update Plan

• Current Status: 🚀 Initial Batch Released. We have uploaded a curated selection of 30+ core domain-language pairs to gather community feedback.
• Next Update: We manage a massive pipeline of thousands of multi-lingual enterprise domains. Full-scale datasets will be progressively synchronized based on community engagement and repository stars.
• Request a Domain/Industry: If you need specific industrial vertical data or certain language pairs immediately, please open an Issue!

📂 Repository Structure

The dataset is organized within the data/ directory, naming each file with its respective root domain and target language suffix:

global-multilingual-cross-industry-corpus/
├── LICENSE
├── README.md
├── autopackingsystem-en.com.jsonl       # Packaging Machinery (English)
├── autopackingsystem-ae.com.jsonl       # Packaging Machinery (Arabic)
├── blistermachine-ru.com.jsonl           # Blister Packaging Machines (Russian)
├── filmmachine-fr.com.jsonl               # Film Co-extrusion Machinery (French)
├── groutpump-ar.com.jsonl                 # Grouting & Mortar Pumps (Arabic)
├── wiremachinery-ae.com.jsonl             # Wire Drawing & Cable Machinery (Arabic)
├── winicetech-de.com.jsonl               # Industrial Cooling & Ice Systems (German)
└── [More multi-lingual industrial domains continuously updated...]

## 📐 Data Schema (Fields Explained)

Each line in the `.jsonl` files represents a single cleaned page or document, mapped into a strict JSON schema:

{
  "domain": "https://autopackingsystem-en.com",
  "industry": "General / Industrial Machinery",
  "language": "en",
  "title": "Semi-Automatic Open-Mouth Bagging Machine, Double Station",
  "source_url": "https://autopackingsystem-en.com/1-1-1-semi-automatic-open-mouth-bagging-machine.html",
  "content": "[Markdown or Cleaned Text Content including technical specifications, features, and parameter tables]"
}