Getting Started¶
ner_for_protein_structures package is published alongside a scientific publication describing the development of a human-in-the-loop named entity recognition algorithm specific for protein structures.
Here we provide a number of command line tools to convert annotations found in BioC formatted XML files, as they have been exported from our annotation tool TeamTat (https://www.teamtat.org/), into other formats. The BioC format is the core data format for the publications as XML text. It was developed by the BioCreative Initiative and is documented here (https://bioc.sourceforge.net/). We also provide a tool to calculate performance statistics following the SemEval procedure (SemEval).
Installation¶
Clone the repository from the source code on Github:
git clone git@github.com:PDBeurope/ner_for_protein_structures.git
It is good practice to create a virtual environment for development:
python3 -m venv ner_venv
Now activate the venv.
source ner_venv/bin/activate
Next, install all the necessary dependencies using the provided requirements.txt
pip install -r requirements.txt
To be able to use some of the NLP tools install the scientific, English language model from Scispacy
pip install https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/releases/v0.5.3/en_core_sci_sm-0.5.3.tar.gz
Downloading and setting up models¶
The different available models, their performance stats and download links are given in section Available models. Huggingface supports git and all models can simply be downloaded through “git clone”. However, as the binary file for the models is too large for standard git, the large-file handler needs to be installed in the parent directory the model will be cloned into.
git lfs install
After the large-file handler was installed, the models can be cloned from Huggingface as in the example below.
git clone https://huggingface.co/PDBEurope/Bioformer8L-ProteinStructure-NER-v0.1
Alternatively, the models can be accessed through Huggingface’s inference API. This option does require a Huggingface account and an authentication token. The details on how to register and how to set up the token can be found on Huggingface
Annotation handbook and TeamTat user guide¶
The annotation handbook with details on how to annotate different entity types and the user guide on the annotation tool TeamTat have been made available and can be found here: