AI Docker Application development. an Text Mining (specifically) NER application in Python.

This is an auto-tagging task. We receive quarterly financial statements from various teams. These documents are in plain html format. What we have is a Python application that auto-tags these documents based on: Taxonomies, and Previously 'filed' statements of the client Tagging has to be done in the document and convert it to xhtml based on tags extracted. Each tag will have labels and various dimensions from the taxonomy Right now our code is all in Jupyter notebook. Current task is make this into a full docker for deploying on AWS. We have taken a small sample to build this prototype. We also need to train the engine on larger corpus to achieve 99% accuracy in extracting the tags. This application will be learning engine. For prototype we have used Spacy, but for application development, I am open for suggestions as long as it is open-source. Especially keen if anyone is comfortable with Deep Learning architectures (Conditional Random Fields/ RNN etc) Additionally we have built this for one taxonomy (US taxonomy). It should be extendable to be able to learn and auto-tag on new taxonomies (say Europe). Deliverables Complete Docker Application for custom NER tagging. Documentation for using of the product (non-technical users) Documentation for developers Training and Testing on two taxonomies