Doctor

The Doctor API returns NHS Choices pages from a search such as "What are the symptoms of dystonia?". The API can also be queried through the Swagger UI or a little embedded UI.

Heroku setup

This will provision the Doctor application and a Bonsai Elasticsearch cluster on Heroku. This has been done already for the ml-doctor application.

Note: due to free dynos and addons limitations, a subset of NHS Choices conditions is indexed on Heroku. For all the conditions, run Doctor locally with Docker as described below.

heroku create ml-doctor
heroku addons:create bonsai:sandbox-10
git push heroku master
open https://ml-doctor.herokuapp.com
open https://ml-doctor.herokuapp.com/swagger-ui.html

Local setup

This will run both the Doctor application and a standalone Elasticsearch instance locally. You may need to increase the amount of memory allocated to Docker to 4GB+ as Elasticsearch is quite demanding and JVMs usually default to ~25% of the total available memory offered.

gradle build
docker-compose up
open http://localhost:8080
open http://localhost:8080/swagger-ui.html

## Code Quality and Continuous Delivery

• Travis CI
• SonarQube with PMD and Checkstyle rules and Jacoco test coverage
• Automated deployment
• DockerHub integration

## Possible improvements

• Distributed crawl and real-time update/index (Spark, actors, queues...)
• Semantic search (knowledge graph, word2vec, deep learning approaches...)
• Boosted search fields (title...)
• Integration tests
• Proper UI
• Many things