Biomedical Entity Relation Detection (BERD)

Authors: Jaehyun Lee, Wooheon Hong

Postech DI Lab project in 2021.01 ~ 2021.03

For security reasons, we do not disclose the code in the network and data folders.

It shows biomedical entity relation network

It also supports relation network for one or more queries.

Requirements

• python3
• see requirements.txt

conda create –n name python=3.7
conda activate name
pip install --upgrade pip
pip install -r requirements.txt

Datasets

We used PubMed thesis data by crwaling

Run

Commands

python app.py

Arguments

test_model.py provides more arguments than app.py provides.

python test_model.py 

• load_or_not

  • Select whether run model or use previously saved data

• query

  • ex) macrolactin, iridovirus

• n_papers

  • The number of pepers

• n_sentence(In the app.py, relation unit)

  • 0: document
  • 1: network consists of one sentence
  • 2: network consists of two consecutive sentence. The Swanson ABC model is available.

  The amount of entity pairs is large in order 0 > 2 > 1.

• is_swanson

  • Using Swanson ABC model
  • n_sentence must be 2

• type_scispacy

  • Select Biomedical Entity Linking model
  • Scispacy

• layout_algo

  • Node Coordinate Algorithm

• laout_dim

  • Select Plot Dimension 2, 3

• radius

  • Select hop except path

• n_nodes

  • The number of nodes

• cutoff

  • Select max path(link between query) length

• n_path

  • The number of paths

Security

The data.sqlite file is the database and stores the username, password, and email address in the Users table.

To protect user passwords, the password is hashed.