Why is some text in pdf extracted as figure (image) and not text using adobe dc sdk extract

Extract uses AI to deconstruct the page so there are no settings and frankly because we just train the AI, we don't really know how it decides. We can just add more documents to the training so it can learn. Would you be comfortable sharing the PDF files with us? 

Hi there,

You can read about the different element types that are extracted in this demo (Look under Summary of Element Types): 
https://documentcloud.adobe.com/dc-visualizer-app/index.html

According to the document in the demo, "Figures" are "Non-reflowable constructs like graphs, images, flowcharts".

If there is some text in a graph or image, it would be extracted as a Figure, despite having what seems to be text. The following image is a good example in the afrorementioned Text Extract demo:

Hope that helps. Let us know if you have any questions.