Settling on my final idea from the previous blog post I thought I needed to look at what was needed to put my idea into a working for. Looking at what was involved in the idea I thought it was important to divide the problem in two. The larger more complex part of this project was going to be implementing a working machine learning solution. I am currently looking at TensorFlow a project by Google which aims to make the use of Machine Learning neural networks more accessible. (TensorFlow Blog Post).
Designing an interface to this application needed to be simple and easy to understand, I wanted it to be clear where the interaction with Twitter was happening and where the output would be seen and understood. I came up with three initial sketches which I then started to refine down into a single idea.
Top Left is the first idea that I came up with and was the final idea that I settled on. I will talk about this idea in more detail after I talk about idea two and three. Idea two was going to be a simple tabular view which showed the tweet on the left and the output on the right. I thought this was a simple idea but for me didn’t provide a good visual representation of locations in respect to each other. I also thought it boring and uninteresting. I will be keeping it in mind when designing as I could use this layout for a debugging tool to simply provide the data in a simple way. The Third idea that I had was a small map and a website which only looked at one tweet at a time. I liked the idea of viewing individual tweets, but because twitter allows you to access all the tweets a user has made I thought it would be more interesting to generate a visual representation of this.
This is the final paper layout and mock-up I came up with. Starting with the background I want the background of the application to be a full screen map taking this to the web I will be able to use Google Maps API to access a map and then to also plot the locations that are predicted by the Neural Network. I thought it was important to show that the accuracy of the neural network may not be 100% so I decided to have an area marker around a centre point which highlights the level of accuracy that could be gather from that tweet.
Clicking on the market would bring up the tweet that the neural network had associated with that location. I have a box top right on the screen where the user can read a “Mission Statement” this would provide a story and the reasoning for why someone would want to use the website and go onto talk about internet privacy and highlight what a user should post online and how comfortable they feel with people and machines being able to work out where they have tweeted from.
The next steps that I need to take will be creating a refined mock-up for the UI and look into the coding that will be required to create a neural network of this complexity. At this stage I am unsure if I will be able to implement a neural network which actually performs well enough to complete the task I have created for my project.
Commentaires