Proposer Suggested Content

Get related content recommendations from Proposer, inside Composer

About

In the suggested-content directory is a Python file that will run a server at https:://localhost:5000 which will accept terms and phrases and find related terms.

The service is written in Python. This is because the gensim library is far more sophisticated than the equivalent in Scala or Java.

Dependencies

You will need nginx installed and running as well as dev-nginx to setup the correct domain for the tool.

All scripts are written using Python 3 so this is a hard dependency. You will also need pip installed in order to fetch dependencies.

It's highly recommended you use a virtual environment when using the service as this will save you a lot of pain. I'd also recommend using the virtualenvwrapper as this will make your life even easier.

Setup

The instructions below assume you are using both virtualenv and virtualenvwrapper.

1. Clone the repository.
2. cd into the cloned repo.
3. Setup the nginx mappings, using dev-nginx and the mapping.yaml defined in the nginx directory.
4. Create a virtual environment with mkvirtualenv <name-of-virtualenv> --python=<path-to-python3>
5. pip install -r requirements.txt
6. Now run the word-2-vec-service.py and the service will be at https://suggested-content.local.dev-gutools.co.uk
7. You will need to the pass the service a trained binary file which you can find in the Composer's S3 or make yourself using the word2vec service.