DataAgent/TCNNet-Pretrain-NetCom-zhTW-3.7M

[TCNNet] Domain Knowledge Dataset for NetCom Field Continuous Pretraining (zh-TW)

A specialized domain knowledge dataset created for continuous pretraining of TCNNet-9B, a Chinese language model based on Yi-9B and specialized in networking and communications domains. The dataset contains articles from various networking, cybersecurity, and tech review sources written in Traditional Chinese.

Dataset Description

Dataset Summary

This dataset is curated specifically for continuous pretraining to enhance Yi-9B's domain knowledge in networking and communications fields. The content is collected from authoritative tech websites and blogs, focusing on:

• Networking equipment reviews and tutorials
• Cybersecurity educational content
• Technical product evaluations
• Network infrastructure discussions
• IT security news and analyses

Languages

Traditional Chinese (zh-TW)

Source Data

The articles were collected from various tech websites including:

• Networking and tech review portals (cool3c, techbang, kocpc, eprice)
• Cybersecurity educational sites
• Tech influencer blogs (techgirlz, john3c, onnidaily, ching3c)
• Hardware manufacturer sites (MSI, ASUS)
• Other tech media (myfeel, ahui3c, kaiching, tangpc, dnpz)

Data Processing

The dataset was created by:

1. Web crawling from authorized sources
2. Content extraction and cleaning
3. Format standardization
4. Quality filtering to ensure technical relevance
5. Deduplication to remove redundant content

Domain Coverage

The dataset focuses on:

• Network equipment and infrastructure
• Cybersecurity best practices
• Tech product reviews and comparisons
• Network protocols and standards
• IT security concepts and implementations

Dataset Characteristics

• Technical vocabulary rich in networking and security terms
• Professional-level content from industry experts
• Current technology trends and developments
• Practical applications and real-world scenarios
• Traditional Chinese technical writing style

Intended Uses

• Continuous pretraining of large language models
• Domain adaptation for networking and communications
• Enhancement of technical vocabulary and knowledge
• Improvement of Chinese technical writing capabilities

Additional Information

The dataset is specifically designed to augment Yi-9B's capabilities in:

• Understanding networking concepts
• Processing technical documentation
• Handling domain-specific terminology
• Analyzing tech reviews and comparisons

Dataset Statistics

Size: 3.7M text corpus data

Licensing Information

GPL-3.0

Data Availability

This dataset is licensed under GPL-3.0 (GNU General Public License v3.0) with the following considerations:

Usage Rights

• ✅ Academic Research: The dataset can be used for academic research purposes
• ✅ Non-commercial Use: Free for non-commercial applications
• ❌ Commercial Use: Commercial usage is not permitted

Data Sources

• All data is collected from publicly available articles and web pages
• Internal or sensitive data used in training TCNNet-9B has been excluded from this public dataset
• Sources include tech blogs, review sites, and educational platforms that are publicly accessible

Usage Guidelines

1. Attribution must be maintained according to GPL-3.0
2. Derivative works must be shared under the same license terms
3. Changes made to the dataset must be documented
4. The complete license terms must be included with any distribution

Access and Distribution

• The dataset is available for download through Hugging Face
• Users must agree to the GPL-3.0 license terms before accessing the data
• Any redistribution must maintain the same licensing and attribution requirements

Please note that while the dataset is derived from public sources, users are responsible for ensuring their usage complies with both the GPL-3.0 license and any applicable local regulations regarding web-scraped content.

---

Please cite our work if you use the data or model in your paper.

@misc{TCNNet,
  author = {Jiun-Yi, Yang},
  title = {TCNNet-9B: Localized Large Language Model for Taiwanese Networking and Cybersecurity},
  year = {2024},
  publisher = {Huggingface},
  journal = {Huggingface Datasets},
  howpublished = {\url{https://huggingface.co/datasets/DataAgent/TCNNet-Pretrain-NetCom-zhTW-3.7M}},
}