How to Use DocQA

Quick Start: DMflow Document QA

Want to let AI quickly learn your enterprise knowledge? Please follow the steps below, go to “Domain” and click on your newly created Document QA item.

After entering the interface, you will see supported upload types, mainly divided into three categories:

Advanced Configuration

Before starting upload, you can also set the following parameters:

• Expiration Time: Set the validity period of knowledge to avoid AI citing outdated information.
• Category: Help Large Language Models (LLM) more accurately cluster questions through category tags, improving retrieval accuracy.

File Upload Details (File)

• Word Count Limit: Basically unlimited.
• CSV: Supports selecting multiple fields as a group, and the system will process them together for chunking.
• JSON: Also supports multi-field groups and can use JSON Pointer to specify paths.
• HTML:
  • You can upload HTML files directly.
  • You can also provide URL links; the system supports XML format (such as sitemap, rss, atom).
  • Filter Function: For XML import, you can filter by link, title, or introduction. Note: This only filters items existing in the XML file, not the content after crawling the website.

Audio Upload Details (Audio)

• Supports audio files within 5MB (mp3, m4a, wav). The system will automatically transcribe to text before indexing.

Dialog Context

This is a secret weapon to improve Q&A quality. You can directly upload organized Q&A records (CSV format).

CSV Field Specifications:

• type: Fill in Q (Question) or A (Answer).
• sentence: Corresponding text content.
• session_id: Session ID.
  • Important Tip: Q and A with the same session_id will be saved in the same Chunk by the system as much as possible. This helps AI understand context relationships, which is very helpful for training continuous Q&A.