Ground Truth Gold. Intelligent data labelling and annotation

23-Jan 2020 - 17:30

One day, your boss comes up to you and asks if you have any experience working with text data?
"Not rea…”
"Excellent! You know, we have all these descriptions about things, and it would be great if you could create some kind of service to suggest tags to make data entry a bit smoother. No rush, it's enough if you have something by the end of the week. Ok, cheers!"

The Holy Grail in data science and machine learning is the ability to obtain labelled datasets needed for training supervised models. Creating these labels by hand is often too slow and expensive.

During this Barrel AI hands-on gathering, we will start by a short theory session about NLP (Natural Language Processing) and continue with working with an exciting dataset from CityPolarna, Sweden's most popular activity site, containing information about ~10.000 public events arranged by the site's members.

From event descriptions and tags added by event hosts, we'll try to create a service that suggests tags for new events, armed only with some basic machine learning.

We'll take a look at some ways of transforming text into something suitable for classification (eg., bag of words, word embeddings), and try out some algorithms ourselves (eg., Naïve Bayes, Logistic Regression, and maybe even something fancy like CNN:s) and see which combinations of pre-processing and classifiers we think are best suited for the task. Time permitting, we might also look at how to bundle the tagger as a standalone service, and/or some tools that are useful when creating a dataset for training.

We'll mostly be working in Python, but it's possible an implementation in R will be added as well. More details coming in a future update. However, bring your own laptop since this is a practical session.

Event sponsoring NordAxon is welcoming you to one of their office spaces, Epic Spaces, and will assure some pizza and refreshment for the participating monkeys. There are only 24 spots available so don’t sign up if you don’t plan to come.After the event, thirsty and social hungry ones are welcome to join for a beer/wine at for example Café Alé.

See you,
/Barrel AI monkeys coordinators