NB: no need to clone repos by yourself, the script will checkout the most relevant branches

It assumes you have Java 11+ installed and the JAVA_HOME is at JAVA_HOME=/usr/lib/jvm/java-11-openjdk-amd64/. If your path is not like this, you can create a soft link. Or, you can modify the JAVA_HOME setting in the install_semehr.sh

1. download the two files into your computer

   # bash script for installation
   https://github.com/CogStack/CogStack-SemEHR/blob/safehaven_mini/installation/install_semehr.sh
   
   # semehr configuration template
   https://github.com/CogStack/CogStack-SemEHR/blob/safehaven_mini/installation/semehr_conf_template.json

2. run the downloaded bash script

   sh install_semehr.sh

   It will ask for a full path to install relevant software. All packages and repos will be installed there.

3. folder structure When it is installed successfully, the installation folder will contain the following folders.

   gcp - contains Java based NLP core software packages
   semehr - contains two Github repos of NLP and machine learning modules
   data - the working folder, which contains the following subfolders
        - input_docs: the free text documents to be analysed
        - output_docs: the NLP raw outputs
        - semehr_results: the semehr post processed results
        - phenome_results: the text phenotyping results
   

4. copy UMLS ontology into the system. (only needed if you would like to identify all UMLS concept mentions from free-text)

   • unzip preprocessed UMLS file (please get in touch if you have got your license of using UMLS - a preprocessed copy will be shared with you)
   • copy two subfolders in output/en/ into YOUR_INSTALLATION_FOLDER/gcp/bio-yodie-1-2-1/bio-yodie-resources/en

1. run nlp

   cd $install_path/semehr/CogStack-SemEHR
   python semehr_processor.py ../../data/semehr_settings.json

results will be saved to $install_path/data/semehr_results

2. [optional] run phenotype computing for stroke, for example

   cd $install_path/semehr/nlp2phenome
   python predict_helper.py ./pretrained_models/stroke_settings/prediction_task.json
   python doc_inference.py ./pretrained_models/stroke_settings/doc_infer.json

   reulsts will be saved to $install_path/data/phenome_results
3. [optional] use customised document level rules
   • goto cd $install_path/semehr/nlp2phenome
   • edit ./pretrained_models/stroke_settings/prediction_task.json. Change the rule_file to a customised rule file, for example using $install_path/semehr/nlp2phenome/settings/stroke-subtype-rules-full.json.