NLP / Named Entity Recognition + Linking

[Jiros]

We need additional expertise to assist with a number of issues that could benefit from NLP/Entity recognition. Our goal is to identify and create more meaningful relationships in currently unconnected sub-graphs within CovidGraph.

Our ethos is open and transparent so we would prefer open source solutions.

For example, within publication & patent text to identify:

• gene names
• drug names
• synonyms for people/institutions
• synonyms for locations

[mpreusse]

Here is a publication with some interesting ressources for our NLP tasks: https://www.nature.com/articles/s41597-020-0543-2

[Jiros]

I added an extract from the introdcution to the Nature article to #35, the researcher use case as it included a good description of what researchers might be looking for from a system like CovidGraph.

[Fohlen]

Hello there everyone! I came here because @yGuy made me aware of the project. BioBERT has been used very successfully on the COVID-19 papers (see https://covidask.korea.ac.kr/). However I think this is still an interesting issue to tackle, specifically it could be tested if the newest advances in RNN architectures (GPT3 by OpenAI) can achieve even better performance. I talked with my supervisor @coltekin (University of Tuebingen) and it seems like a suitable collaboration for my bachelors thesis in computational linguistics.

If this sounds interesting, it would be good to have a discrete list of entities that are really relevant to investigate.
All the best,

Lennard

[motey]

Hi Lennard,

If this sounds interesting

YES! 🚀

i was allready peering at gpt2/3 :)
would be great if you want to have a try on the NLP+Covid*Graph-thing.
The easiest way to bring you up on speed, regarding data model and relevant nodes is maybe if you join our chat via the matrix.org network see https://github.com/covidgraph/documentation/wiki#communication
You can add me via @tim.bleimehl:meet.dzd-ev.de and i can invite you to the relevant groups. from there we could also setup a call with some of us in the data-core group.

Also we are trying to get our hand on some more hardware (GPUs) to speed up computing if applicable in a later stage.