GOOGLE NONSENSE LABORATORY

A Guided Process of Creating Logically-Generated Nonsense

ABOUT

A SET OF TOOLS THAT MANIPULATE LANGUAGE WITH MACHINE LEARNING

To provide a sense of what the tools in Nonsense Laboratory do, here is a brief description of each:

MIXER: Combines two or more existing words to create a nonsense word

MOUTHFEEL TUNER: A user can manipulate a sentence by emphasizing phonetic sound characteristics

RESPELLER: Respells sentences as if they were spoken without using certain sounds or letters

SEQUENCER: Invents new words by sequencing mouth movements

EXPLORER: A user can scroll a word map of related nonsense words, with an existing word as a base

PROJECT OR FILE

You can explore the tool on Google's Arts & culture website here.

The Result

Here are finished screens and a playthrough from the project.

THE CHALLENGES:

CREATING CLEAR AND ENGAGING TOOLS OUT OF COMPLEXITY

We needed to make the interfaces of the tools clear and engaging, as well as highlight the technology being used, without bogging users down by too much of the complexity of how it all works. Specifically:

Developing the UI of the individual tools themselves.
Deciding which tools and how many would be featured, because there were many machine learning technologies explored, and different features exhibiting those technologies could be clumped together into the same tool or separated into different tools.
Determining a site flow that would encourage users to jump between the different tools and try as many as possible.

PROBLEM STATEMENT

The average person needs more specific
and intuitive examples of Machine Learning in order to learn about its wide-spanning potential uses. Helping users become more engaged with this set of language tools can help increase their appreciation and understanding of machine learning as an area of research and creativity.

UNDERSTANDING THE USER

Two personas adequately describe the target demographic of this unique exploratory tool: The primary persona is the tech savvy web aficionado in their 30s, who is probably well familiar with other Google Arts & Culture experiments, and perhaps is working with art and technology. The secondary persona is the curious, lifelong learner, someone who is not deeply invested in technology, but who enjoys sharing entertaining interactive websites with friends, especially those that teach something new.

Bio

Hudson is a 32-year-old digital artist who creates interactive and anamotronic works. He receives grants from arts organizations and universities to complete his work. He attended Parsons School of Design in New York City.

Hudson Port-Davey:
Art & Tech Hipster

Stories & Scenarios

At his downtown Brooklyn warehouse-turned-studio, Hudson plots his next artistic endeavor. He searches for inspiration online, and will compile those inspirations and save them in a folder. He may use them to support grant applications as well.

Goals

He wants to discover unique uses of new technology so that he may study and use it in his artistic practice.

“I might consider using machine learning for my next big project, but I dont have a solid understanding of its applications.”

Behaviors

• He continually incorporates new technology into his works. In the past he incorporated generative art technology, augmented reality, and virtual reality.
• He has a shared Whats App thread with friends who constantly share new technology and inspiring works.
• He plays in a band and wonders how it might become a part of his practice.

Frustrations

Hudson wants a more immediate understanding of creative uses for new technologies. He feels as if he’s “seen it all” in his own cultural corner of Brooklyn. He wants to learn and be inspired, and see and experience technologies that are used in new ways, but are also understandable.

SOLUTIONS:

ORGANIZING FEATURES INTO TOOLS

We defined usability and designed the tool interfaces based on research findings:

We conducted research into linguistics and speech pathology by interviewing a speech pathologist to understand language teaching methods, and by reviewing existing research completed by the project lead on incorporating machine learning for language-based tools
Based on our research, we developed 5 tools within the lab that combined various features explored by the project lead, each one utilizing machine learning for a unique main purpose. For example, while the Respeller tool’s purpose is to respell sentences in terms of how they might be spoken out loud without certain sounds added or removed - it actually combines several separate features and technologies developed by the project lead into the single tool and interface.
Prioritized playability by arranging tools in a logical order, with the simplest first
Enabled seamless tool switching without returning to the main menu
Implemented a tooltip tutorial