Extract data from PDF (Scanned image) based file
Summary:
Use Case:
- The user uploads a file from the front end.
- Based on the user’s instructions, specific details are extracted from the document.
- The extracted details are stored in corresponding table columns, where they can be reviewed and processed further by the user.
Current Approach:
| Step | Description
|------|------------------------------------------------------------|
| 1 | **Start** |
| 2 | **Upload PDF file** |
| 3 | **Extract content and instructions** |
| 4 | **Convert to vector embeddings** |
| 5 | **Store in 23ai vector database** |
| 6 | **Perform vector search** |
| 7 | **Pass relevant vectors to LLM** |
| 8 | **LLM refines the result** |
| 9 | **Output final refined result** |
This Approach works fine when pdf is text based.
Obstacle:
We’ve encountered a significant limitation in processing the PDF documents due to the following reasons: