73% of respondents of a survey conducted claimed 'practicing conversations' as the most difficult part of language learning. 


Chattie, a voice based application, that helps language learners practice real world conversations and in turn improve vocabulary.


Voice UX, Writing Sample Dialog, Speed Dating, Wizard of Oz Testing, Decision Trees, Mixed Method User Research



Team of 4, VUX Designer

Poster presented at UXPA, 2018

Blog in UX Matters




Language learners who are not fluent in speaking feel embarrassed in social situations

It is predicted, that in the year 2020, the number of people learning English as a Second or Foreign Language alone will double to nearly 2 billion people. While there are many resources that help language learners read and write, people face difficulty in finding partners and resources to practice speaking. And so, language learners who are not fluent in speaking end up feeling embarrassed and uncomfortable to talk in social situations. 




Conversation Design Process 


What difficulties do language learners face?

To understand what motivates people to learn a new language and what problems they face while language learning, we sent out  a survey to students at the University of Michigan. We received 148 responses, out of which 73% said that speaking is the most difficult part of language learning.


What do existing solutions offer?

We did a comparative analysis of existing language learning tools to uncover the gaps in existing solutions. These included DuoLingo, Memrise, Speechace, Mondly (digital), language learning classes (physical) and Spotify Podcasts, Google Translate (indirect). We found that most tools were geared towards written language learning, not conversation practice nor feedback for pronunciation. 

Comparative Matrix.png


Voice User Interface as a Potential Solution

At this point, it was clear to us that we need to help language learners learn how to speak the language. Also, this had to be a two-way street with back and forth. A voice application was the perfect modal for this, as it allows the user to speak the new language and could also be designed for conversations in practical real world scenarios.


What do language learners who own voice assistants say?

We conducted four in-depth interviews with people who own own voice assistants and have tried to learn a language (beginner to intermediate level). We spoke to them about how they learnt to speak, what techniques they found useful, what worked for them and what didn't. We also asked them about their perceptions of voice assistants - what features they liked or didn’t like,  and what they expected from their assistants. We created an affinity wall to synthesize our research and generated key findings.


Say Hello to Chattie!

Our mixed methods user research affirmed our solution of a voice based companion that would help language learners practice conversations eventually making them more confident while speaking in real world scenarios. They would also improve their vocabulary along the way.


Intermediate Language Learners as Target Audience  

Based on our key findings, we decided to target intermediate level language learners, that is learners who know basics of the language and want to improve their speaking proficiency and vocabulary.


Chattie as Voice Based Companion

We created a persona for our voice based companion ‘Chattie’ to have a clear picture of who the target audience is communicating with, and what is the tone and personality of the VUI.



Writing Sample Dialog  

To design a sample dialog, it was important for us to understand how people spoke in real world conversations. We chose a topic, say “Food”, and started a conversation by asking people “What food do you like to eat?” and observed their responses. This helped us create an initial conversation flow.  We iterated over this by conducting multiple small user tests. This helped us make our responses natural and human-like, helped us identify unscripted responses, and shaped our overall information architecture. Throughout this exercise, we constantly referenced Cathy Pearl’s ‘Designing for Voice User Interfaces’ and Google’s Conversation Design guidelines.

Initial conversation flow

Final Sample Dialogs



Wizard of Oz Testing 

Once we had designed our features, we tested them with users through 2 rounds of Wizard of Oz tests. We used Google text-to-speech software to record and play Chattie’s part of the conversation in response to the user. Our user tests helped our refine our conversation and made Chattie sound more friendlier.



Winning the Student Design Competition at UXPA International '18 

I had the wonderful opportunity of presenting our project as a poster at UXPA International conference in June 2018 as part of the Student Design Competition. We had the chance to interact with professionals & academics in UX Design & Research and received constructive feedback about our project. The conference is a great place to network, and I learnt about cutting edge work in the industry. 

San Francisco Bay Area