[AI] Extraction: Automatically Enrich Product Attributes

Summary

What is it?

The Extraction Module helps you enrich missing product attribute values by analyzing text (such as product descriptions or titles), images, and PDF files. It works after your products have been classified.

How does it work?

The AI analyzes available product information and fills in missing attributes, like dimensions or specifications. If an attribute has low confidence, it's flagged in orange for review. Images from the input file can also be displayed to help you confirm or edit attributes.

Extraction from public asset URLs pointing to PDFs or images is also supported.

 

 

Key Sections

  • Fields Section: On the left, you can see which sources are used for the AI.
  • Attributes Section: Here, you confirm and save the completed attributes which have automatically filled.

Attribute Types

  • Select Attributes: Predefined values you select from a list.
  • Textual Attributes: You can input text or numbers manually.

Attribute Statuses

  • Mandatory (Red Star): Must be filled before moving to the next step.
  • Important (Orange Dot): Alerts you if empty but doesn't block progress.
  • Optional: Non-blocking and not represented visually.

Filters

  • Origin Filter: Audit only AI-completed attributes, marked with a robot icon.
  • Importance Filter: View attributes based on their level of importance (mandatory, important, or optional).

Limitations

  • Maximum number of products (rows) per job: 100,000
  • Maximum number of attributes: 50
  • For PDF files, only the first PDF with the first 20 pages are processed as sources.
  • For images, only the 5 first images are processed as sources.

We advise to send fewer than 200 products (rows) with less than 20 fields.