NLPeasy Workshop

Build NLP pipelines the easy way

Repository for the NLPeasy workshop.

For the workshop you have 2 possibilities to participate.

Mybinder.org

This is a pure online version so no installation needed on your laptop other than a browser. There might be issues to connect behind company firewalls though.

Downsides:

• your session will be closed after a period of inactivity of 15-60 minutes, e.g. when you loose your internet connection
• your work in the Jupyter-Notebooks will be lost then.

https://mybinder.org/v2/gh/d-one/NLPeasy-workshop/master

However, there is only 1-2 GB of RAM available, which is tough for our example. Also if you loose your connection or close your laptop for 10 minutes your session is lost. During the workshop we will provide you with bigger VMs in our cloud.

The same can also be done if you have docker installed (see instructions below) using:

JUPYTER_PORT=8888
docker run --rm -itp $JUPYTER_PORT:$JUPYTER_PORT doneai/nlpeasy-workshop jupyter lab --ip=\* --port=$JUPYTER_PORT --NotebookApp.token='' --NotebookApp.password=''

And then going to http://localhost:8888.

Own Laptop

You can work on your own laptop or server. For this you need:

• This Repository

• A Python Environment (>=3.6)

• Access to an Elasticsearch and Kibana Server: NLPeasy supports both connecting to an existing one or it can start one for you on docker. For this install one of the following:

• Docker https://www.docker.com/get-started, direct download links for macOS (DMG) and Windows (exe) (On Windows Home Edition make sure to have upgraded since May-2020).

• Elasticsearch and Kibana: https://www.elastic.co/downloads/ or https://www.elastic.co/downloads/elasticsearch-oss (pure Apache licensed version, does not include Java SDK on macOS!)

Setup

Get this repository:

git clone https://github.com/d-one/nlpeasy-workshop
cd nlpeasy-workshop

Then setup a virtual environment, install requirements and download a spaCy-model: Then on the terminal issue:

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python -m spacy download en_core_web_md

Also you might want to download some bigger files:

curl -LO https://github.com/rudeboybert/JSE_OkCupid/raw/master/profiles.csv.zip
curl -LO https://github.com/d-one/NLPeasy-workshop/releases/download/v0.2/okc_enriched_demo.pickle.tar.gz
curl -LO https://github.com/d-one/NLPeasy-workshop/releases/download/v0.2/elastic-data.tar.gz

This will download:

• profiles.csv.zip: our data for today
• okc_enriched_demo.pickle.tar.gz: the solution that should come out of NLPeasy (saving you a an hour of computation)
• elastic-data.tar.gz: if elasticsearch is pointed to the elastic-data/elastic-data as it's data folder, then you can see immediatly the indexed data and generated dashboard.

Start Jupyter Lab

Still in the activated virtual env venv you now can start jupyter lab

jupyter lab