NarrativeAnnotation: GNorm2 integration

[HermannKroll]

PubTator has updated its annotation tool. There is a successor for GNormPlus, called GNorm2. Before we replace GNormPlus, we should answer the following questions:

• How easy is the replacement? So what has to be changed in our pipeline?
• How fast ist GNorm2? It uses transformers (language models). We should test its performance on CPU and GPU (our mining server only has a CPU)

It seems that GNorm2 is written in Python. However, I would not directly integrate the code into our system because of requirement conflicts and a huge bunch of additional requirements. It could be a good idea to implement it by starting it via a subprocess and an additional Python environment (we did something similar for OpenIE 6 in our extraction toolbox).

Code:
https://github.com/ncbi/GNorm2