Connectivity-RE

Code/corpus repository for the Autonomic nervous system (ANS) connectivity extraction. For more details, see "Detecting Anatomical and Functional Connectivity Relations in Biomedical Literature via Language Representation Models" to appear at SDP2021@NAACL-HLT 2021.

export CONNECTIVITY_RE_HOME=/full/path/to/connectivity_re/repository

Download Bio-ELECTRA models

• The pre-trained Bio-ELECTRA mid and base sized models are available at Zenodo.
• The pre-trained Bio-ELECTRA and Bio-ELECTRA++ small ELECTRA models are available at Zenodo.

Building Java code for data preparation

• You need Java 1.8+ and Gradle build system for the Java portion of the code.

cd $CONNECTIVITY_RE_HOME
gradle clean install

ANS Connectivity Corpus

• The initial annotated corpus is in $CONNECTIVITY_RE_HOME/data/sparc/base/sparc_connectivity_nerve_ganglia_sampled_idx_joe_full.xml.
• The randomly sampled annotated corpus of 250 sentences for active learning control set is in $CONNECTIVITY_RE_HOME/data/sparc/base/active_learning/sparc_connectivity_nerve_ganglia_random_set_annotated_idx.xml.
• The annotated corpus for each of the ten iterations used for active learning tests are in $CONNECTIVITY_RE_HOME/data/sparc/base/active_learning/iterations/sparc_base_al_iter_<iter-no>_curated_idx.xml.

Viewing/Editing ANS Connectivity Corpus

cd $CONNECTIVITY_RE_HOME
java -jar relation-annotator-1.0.18.jar

GPU requirements

You need Tensorflow 1.15 and CUDA 10.0 for GPU.

Setup virtual environment

Ensure you have virtual environment support (e.g. for Ubuntu)

sudo apt-get install python3-venv

python3 -m venv --system-site-packages $CONNECTIVITY_RE_HOME/venv

source $CONNECTIVITY_RE_HOME/venv/bin/activate

pip install --upgrade pip
pip install tensorflow-gpu==1.15
pip install sklearn
pip install hyperopt

• The code for the ANS connectivity extraction classifiers are in $CONNECTIVITY_RE_HOME/scripts/bio_electra.
• The training/testing data for the binary and ternary classifiers are in $CONNECTIVITY_RE_HOME/scripts/bio_electra/data/finetuning_data/sparc and $CONNECTIVITY_RE_HOME/scripts/bio_electra/data/finetuning_data/sparc-multi, respectively.

For any questions, please contact [email protected].