MapleBridge A2A Trade Logic Dataset

MapleBridge 加中贸易智能匹配数据集

Bilingual Q&A dataset for Canada-China B2B procurement matching 加中跨境B2B采购匹配双语问答数据集

---

Overview / 概述

This dataset powers the semantic matching engine at MapleBridge.io — a live B2B platform connecting Canadian importers with Chinese manufacturers.

本数据集用于驱动 MapleBridge.io 的语义匹配引擎，该平台实时连接加拿大进口商与中国制造商。

The dataset contains structured bilingual procurement scenarios, supplier capability descriptions, and matching rationale — designed for:

• Fine-tuning LLMs for cross-border trade understanding
• Benchmarking AI matching accuracy in B2B contexts
• Training agents to understand buyer intent vs. supplier capability

---

Dataset Structure / 数据集结构

File: maplebridge_qa_dataset.jsonl

Each record follows this schema:

{
  "instruction": "Match this buyer need to the best supplier type",
  "input": "Looking for 10,000 units PCB, IPC Class 2, 4-layer, delivery within 6 weeks",
  "output": "Best match: mid-size PCB manufacturer with IPC Class 2 certification, export experience to North America, capacity >50K units/month",
  "language": "en",
  "category": "electronics"
}

Field	Type	Description
instruction	string	Task for the LLM
input	string	Buyer procurement need or supplier description
output	string	Expected matching rationale or result
language	en / zh	Language of the record
category	string	Product category

---

Product Categories / 品类覆盖

Category	品类	Example Query
electronics	电子元器件	PCB, sensors, LED modules
apparel	服装纺织	OEM clothing, fabric sourcing
machinery	机械设备	CNC parts, hydraulic components
furniture	家具家居	Custom furniture, flat-pack
medical	医疗器械	Disposables, diagnostics
food	食品农业	Organic products, bulk grains
building	建筑材料	Tiles, steel structures
packaging	包装材料	Custom boxes, bags

---

Intended Use / 使用场景

1. LLM Fine-tuning / 大模型微调

Fine-tune models to understand the nuance between:

• What a buyer asks for vs. what they actually need
• Chinese supplier self-descriptions vs. Western procurement terminology

2. RAG (Retrieval-Augmented Generation)

Use as a knowledge base for an AI procurement assistant that can:

• Translate buyer requirements into supplier search queries
• Score supplier-buyer compatibility

3. Benchmark Evaluation / 基准测试

Evaluate how well a model understands cross-border B2B matching across 8 product categories and 2 languages.

---

Live Platform / 在线平台

This dataset is used in production at maplebridge.io.

• Post a real procurement need: maplebridge.io/app
• See the matching engine in action — results delivered within minutes

---

Related / 相关资源

• 🔧 Protocol Spec: maplebridge-a2a-protocol-spec
• 🌐 Live Platform: maplebridge.io

---

Citation / 引用

@dataset{maplebridge2026,
  title     = {MapleBridge A2A Trade Logic Dataset},
  author    = {MapleBridge.io},
  year      = {2026},
  url       = {https://github.com/jinjihuang88-ui/maplebridge-a2a-trade-logic-dataset},
  note      = {Bilingual Canada-China B2B procurement matching dataset}
}

---

License

MIT © 2026 MapleBridge.io