Text simplification by visual elements

Hadas Bar is a start-up which is located in Israel and deals with some different areas around Artificial Intelligence.
Among our applications we developed one which makes easy access to ancient and complex texts. This system simplifies complex texts by using visual elements (diagrams, pictures, tables, trees and more) inline the text, so the reader can understand it better.
Our current system 
sysgad.com/home-en contains the entire Jewish Talmud 5400 pages (2700 two sides) with about 70,000 paragraphs. So far, we have handled about 10% of the text manually. For each paragraph or group of paragraphs we built a manual table and defined a category. The system uses some AI/ML methodologies for learning the manual categories and can automatically identify the categories in the rest of the text at a very high level. In addition, the system uses "Attention" from the area of Deep Learning (AI) to build a model from the manual tables. And using this model to build automated tables on the rest of the text.

I want to open to different discussin:

  • What could be the best AI methodolgies for this challenge?
  • Does this idea can simplify complex form the areas of:
    • Legal information
    • Regulation
    • Rules

Shalom

Oznake
deep learning Natural Language Processing

Comments

Profile picture for user dervijo
Poslal Jola DERVISHAJ Pon, 10/01/2022 - 11:10

Very interesting initiative,

Thanks for sharing and looking forward to the discussion!