Analyzing Twitter Location Data with Heron, Machine Learning, Google's NLP, and BigQuery

In this article, we will use Heron, the distributed stream processing and analytics engine from Twitter, together with Google's NLP toolkit, Nominatim and some Machine Learning as well as Google's BigTable, BigQuery, and Data Studio to plot Twitter user's assumed location across the US. We will show how much your Twitter profile actually tells someone about you, how it is possible to map your opinions and sentiments to parts of the country without having the location enabled on the Twitter app, and how Google's Cloud can help us achieve this. While it is safe to assume that most Twitter users do not enable the Location Services while using the Social network, we can also assume that a lot of people still willingly disclose their location – or at least something resembling a location – on their public Twitter profile. Furthermore, Twitter (for the most part) is a public network – and a user's opinion (subtle or bold) can be used for various Data Mining techniques, most of which do disclose more than meets the eye. Putting this together with the vast advances in publicly available, easy-to-use cloud-driven solutions for Natural Language Processing (NLP) and Machine Learning (ML) from the likes of Google, Amazon or Microsoft, any company or engineer with the wish to tap this data has more powerful tool sets and their disposal than ever before.