Enhancing Information Architecture with Machine Learning for Digital Media Platforms

My College Dissertation

My research explored the need for user experience design in machine learning functions, and the synergetic relationship between the two. It includes a proposed model for a dynamic topic selection system within the information architecture of social media platforms. This work holds massive potential for companies with digital media platforms, as well as the data science and product design community.

To read the full publication, click here.

What People Are Saying

“As a content creator, a feature like this would be a dream come true. I dread trying to come up with the right hashtags when posting something.”

— A Student in the Audience of the Research Symposium

“I like users can browse topics rather than search for specific ones. This could help people further explore content they already like, along with widening their social media exposure.”

— Steve Engel, Director of the Honors College

“I really love this. You should send this idea to the social media companies. It’s hard to believe social media sites do not do this already.”

— Chantel, a Graphic Design Student

The Problem

With 60% of the global population on social media with over 300 million uploads per day, long gone are the days of any static information architecture.

However, topic selection systems are typically static in online platforms.

The Solution

How Interests are Classified

Back-end Side

  • Computer vision

  • Deep neural networks

  • Natural language processing (NLP)

User-Facing Side

  • High-level topic selection

  • Keyword input

  • Topic selection (#)

Machine Learning helps bridge these two together!

The Proposed Model

Uploading a post

The user is able to select tags that are generated with artificial intelligence off of the image/video.

The tags are displayed in a hierarchical format. The first level being category, the second level being subcategory, and the third level being descriptives.

Scrolling through mass amount of options or manually inputting text minimizes to a single click.

The user journey

  • Step 1

  • Step 2

  • Step 3

  • Step 4

The Content Feed

A user is able to explore through topics that are relevant to their interests.

Once a user selects a topic, the options to the right regenerate to offer subtopics based off of what they’ve chosen.

In this way, a user can browse, instead of directly searching, for narrowing down their feed.

An Example

The Benefits

The Function

  • Creates less semantic ambiguity and synonyms

  • Can organize them into levels of abstraction

  • Decreases cognitive load

  • Assist in the search function

  • Texts and images together yields the highest accuracy for interest classification

The Design

  • Enhances principles of information architecture (Principle of progressive disclosure, Principle of focused navigation, Principle of attained growth)

  • Enhances the laws of UX (Hick’s law, Miller’s Law, Fitt’s Law, Law of Proximity, Law of Uniform Connectedness)

View More

  • Learn more about this proposed model.

    View

  • Explore the rest of the portfolio.

    Explore