DS4SD / docling

Get your documents ready for gen AI
https://ds4sd.github.io/docling
MIT License
6.43k stars 311 forks source link

Hierarchical Topic Parsing #77

Closed thusithaC closed 1 month ago

thusithaC commented 1 month ago

First of all, amazing work and thanks for making this open source!

Issue

I'm interested in obtaining hierachical topics after converting PDF -> MD. My naive attempt resulted in a document well extracted, but with a somewhat flat structure.

image


## About This Insurance

## A brief overview

Motor Vehicle Insurance provides cover for your car, motorcycle, trailer and horse float, anywhere in Australia. These types of vehicles are referred to as 'your vehicle' in this booklet. Under this insurance you can choose from three different levels of cover, and three types of policies.

## Levels of cover you can choose from

You can choose from three levels of cover, depending on your needs and your type of vehicle.

## You can choose from:

· Comprehensive for cars, motorcycles, trailers or horse floats.

· Third Party Fire And Theft for cars or motorcycles, or

· Third Party Property Damage for cars or motorcycles.

I used the out of the box pipeline for this extraction DocumentConverter().

Is there a way I could obtain hierarchical topics (#, ##, ###). What components of the pipeline can I change?

Thanks!

dolfim-ibm commented 1 month ago

At the moment, the system is detecting only one level of section headers.

staar commented 1 month ago

New capabilities will come, but for now, the system only recognises titles and the first level of section headers.