AWS Comprehend is a great tool when you want to extract information from textual data. As a managed service it is really easy to setup and can be used with next to no prior knowledge of machine learning. But there is one minor thing that bugs me about Comprehend: The Output.

TL;TR output.tar.gz bad, flat json file good.
See python code below for transformation.


Author:Lukas Höfer
Tags:aws, comprehend
Categories:aws, cloud, machine-learning