But like I said, I’m an algorithms guy. Those machine learning models I’ve tuned are in Python and R. And, I don’t want to spend all my time trying to glue them together with web code that I don't understand very well and am not terribly interested in.
Here is an interactive data application that was the keystone of a post about building topic models on the presidential State of the Union addresses. Try it out. You can choose the number of topics, as well as document frequency filters for the maximum and minimum number of documents a word can appear. More topics is like zooming in on the body of text. Fewer is like zooming out. A maximum document frequency of 0.75 means that if a word appears in more than 75% of the speeches it is not used since it is “too common”. In the same way, a minimum document frequency of 0.05 means that if a word appears in less than 5% of the speeches, it is “too rare” to contribute to an overall trend. (Note: the LDA model takes a few seconds to run.)
Here is a behind the curtain look at the Exaptive elements that make up this xap:
The xap is grouped into those same 3 functional areas: user input (HTML), algorithm (Python), and visualization(D3.js)... plus a fourth to show a progress bar when the app is processing a model and control, a universal color scheme across plots. Each section is made up of several components. Each component takes inputs on the left and outputs to the right. Which outputs are connected to which inputs is governed by a drag and drop wire. Inside each component, the best technology to accomplish that task is encapsulated. The platform takes care of communications between components.
Clicking on the component in the algorithmic layer shows the Python code used to build the Latent Dirichlet Allocation Topic Model.
This is the part I did not write. In fact, this is far better and more scalable D3 than I can write. All I have to do now is connect the output data of my topic model component to the inputs of the visualization components. Not only that, but the output of one visualization can be used as an input into another one to make the plots dynamic and interactive.
I’m an algorithms guy (please stop me if I start repeating myself), and what I want more than anything else is to be able to effectively communicate the models I’ve built and get them into the hands of other people to see what they can find. But somewhere out there is someone who is a great visualization person that would love to have good algorithmic components to wire into their visuals. They built these visualization components I am using. They make my work better, and I hope my algorithms can do the same for them. If I can effectively connect and collaborate with those interested in visualization, design, user experience, subject matter expertise, and even other types of algorithms, then we can amplify the quality of each other’s work.