ℹ️ In this repository we've contained all
- notebooks,
- results,
- datasets,
- utility files and configs,
- all other files that we've used in our blogposts.
👌 Everything is reproducible and contenerized so you don't need to worry about installing anything.
Build and run
docker-compose up --buildto build all containers and run the application.
Or run it in the background with
docker-compose up -d --build🌟 Now you should see that the container started and port with token can be found in the docker logs, app should be located at localhost:8888/"token".
Teardown
If, at any point, you would like to stop the containers and clean up any remaining data, run the following command:
docker-compose down --volumesThis will stop the containers and delete any stored data in Elasticsearch node.
Phonetic Analysis Plugin
In order to recreate the results of one of the experiments using phonetic analyzer, you have to install phonetic analysis plugin within the Elasticsearch container:
- Run bash within the container
docker exec -it elasticsearch_service /bin/bash- Install the plugin and exit the container
bin/elasticsearch-plugin install analysis-phonetic
exit- Restart the Elasticsearch container
docker restart elasticsearch_servicedata folder contains all the neccessary data for the tests.
That includes:
- SQUAD 2.0 dataset, and all our variations of it.
- SwiftUI dataset, from Stanford SwiftUI courses.
How we changed and used the dataset is described in the README file.
experiments folder contains all the experiments we ran.
All files are titled experiment_* and the second part is the name of the experiment.
All results are stored in the experiments/results folder and are reproducible via experiment.* notebooks.
We've also included some useful utilities for example for scoring the results or indexing Elasticsearch.
| Name | Link |
|---|---|
| Elasticsearch | https://www.elastic.co/elasticsearch/ |
| Project Jupyter | https://jupyter.org/ |
| Docker | https://www.docker.com/ |