Add NLP capabilities for analyzing free text

[kec1510]

Quick summary

Wanted to throw this idea out there - I noticed that on the wireframe on the website we have a section where users can add comments about their health. I was thinking we can use NLP/word2vec to draw insights from user comments.

Motivations

The data that we are focused on drawing conclusions from so far is very concrete - i.e. symptoms described as "minimal", "moderate", or "severe". We can generate actionable insights from this data, but for a more robust overall picture of what COVID actually looks like in our society it is useful to analyze trends in how users describe the symptoms and overall stress they are experiencing in more detail. This analysis could also allow us to tailor the app better to the users' needs, for example in the education/resources section we can include how users can maintain good mental health in addition to their physical well-being if many users are experiencing a lot of anxiety.

Basic Methodology

@ngiangre mentioned that eventually we'll be querying user data from an API (with their consent) - so we would pull the comments/free text data and put it through a word2vec algorithm which can mathematically group similar words and concepts together so we can analyze trends. (Disclaimer that I am not an NLP expert - outsourcing this idea so anyone who can expand my idea of this methodology please go ahead!)