BYJU's :
BYJU's :
Language learning voice assistant
Language learning voice assistant
Language learning voice assistant
Voice User Experience
Voice User Experience
UX research
UX research
Strategy
Strategy
Prototyping
Prototyping
User testing
User testing
Project planning
Project planning
Conversation Design
Conversation Design
THE PROBLEM
THE PROBLEM
Overcoming language learning barriers for kids 6-13 years
Overcoming language learning barriers for kids 6-13 years
In India, English is widely recognized as a crucial language for educational and professional success. However, Indian children in resource-limited areas face significant obstacles in learning and practicing English.
In India, English is widely recognized as a crucial language for educational and professional success. However, Indian children in resource-limited areas face significant obstacles in learning and practicing English.
→ Limited Access to Resources: Scarcity of English language learning materials and resources in their local environment.
→ Lack of Exposure: Minimal opportunities to hear and practice English in daily life.
→ Fear of Judgment: Anxiety and embarrassment when speaking English in front of peers.
→ Increased reluctance to participate in English-speaking activities.
→ Limited Access to Resources: Scarcity of English language learning materials and resources in their local environment.
→ Lack of Exposure: Minimal opportunities to hear and practice English in daily life.
→ Fear of Judgment: Anxiety and embarrassment when speaking English in front of peers.
→ Increased reluctance to participate in English-speaking activities.
Role and responsibility
Role and responsibility
Role and responsibility
Lead product designer
Lead product designer
Deliverables
Deliverables
Deliverables
User interviews
Conversation Research
Dialogue Scripts
Wireframes
High-Fidelity Designs
User interviews
Conversation Research
Dialogue Scripts
Wireframes
High-Fidelity Designs
Team collaboration
Team collaboration
Team collaboration
Data analysts
Product manager
User researcher
GATHERING DATA
GATHERING DATA
Fake door testing
Fake door testing
Launching an assistant on the existing app to collect data
Launching an assistant on the existing app to collect data
Total users' data collected
Total users' data collected
mins
mins
mins
Engagement Duration per session
Engagement Duration per session
recorded dialogues
recorded dialogues
ANALYSING DATA
ANALYSING DATA
Recorded dialogues
Recorded dialogues
Recorded dialogues
My analysis of interactions and conversations from kids using the beta version of the voice assistant focused on three key criteria:
My analysis of interactions and conversations from kids using the beta version of the voice assistant focused on three key criteria:
Intent Broader objective of users command
Utterance How the users were phrasing their request
Variations Optional inputs/ Variants of dialogues
Intent Broader objective of users command
Utterance How the users were phrasing their request
Variations Optional inputs/ Variants of dialogues
Intent sorting
Intent sorting
Intent sorting
The conversations between the kids and the voice assistant were analysed deeply to understand how the kids were interacting with and expecting from the assistant. To effectively categorise and respond to user commands, we identified two types of intents:
The conversations between the kids and the voice assistant were analysed deeply to understand how the kids were interacting with and expecting from the assistant. To effectively categorise and respond to user commands, we identified two types of intents:
Global Intent: Complex requests requiring comprehensive responses.
- Controls & Help
- Conversational
- Hint and help
- Error states
Local Intent: Simple yes/no responses.
Global Intent: Complex requests requiring comprehensive responses.
- Controls & Help
- Conversational
- Hint and help
- Error states
Local Intent: Simple yes/no responses.
DESIGNING PERSONALITIES
Buddy, your language learning partner
Buddy, your language learning partner
Giving the assistant a personality
Giving the assistant a personality
KEY USE CASES
KEY USE CASES
Coversation Design Canvas
Coversation Design Canvas
SAMPLE SCRIPTS
SAMPLE SCRIPTS
Writing Dialogue Flows
Writing Dialogue Flows
The assistant is designed to address various intents within the app. Below, you'll find how I detailed interaction flows illustrating how the system responds to a child's request to learn a new word or engage in other educational activities. These dialogue flows demonstrate my approach to maintaining kid's curiosity while structuring intuitive conversations.
The assistant is designed to address various intents within the app. Below, you'll find how I detailed interaction flows illustrating how the system responds to a child's request to learn a new word or engage in other educational activities. These dialogue flows demonstrate my approach to maintaining kid's curiosity while structuring intuitive conversations.
Triggering the assistant
Triggering the assistant
Interaction flow - Learning a new word
Interaction flow - Learning a new word
Dialogue Flow Mapping
Dialogue Flow Mapping
SCRIPTS
SCRIPTS
600+ managed intents
600+ managed intents
Other Intents, Utterances, & Prompts
Other Intents, Utterances, & Prompts
ACCESSIBILITY
ACCESSIBILITY
Inclusive Persona Spectrum
Inclusive Persona Spectrum
Ensuring Accessibility and Inclusivity in Language Learning for Children
Ensuring Accessibility and Inclusivity in Language Learning for Children
USABILITY TESTING
USABILITY TESTING
Wizard of Oz Testing
Wizard of Oz Testing
Results, insights and actions taken
Results, insights and actions taken
Seconds
Seconds
Seconds
Optimised average dialogue length
Optimised average dialogue length
%
reduction
reduction
reduction
in task failure and interaction duration
in task failure and interaction duration
%
reduction
reduction
reduction
In abandoning a teaching unit
In abandoning a teaching unit
%
improvement
improvement
improvement
in task completion rate
in task completion rate
Lessons learnt
Age appropriate content
Create clean design Shorter is better. Offer fewer options and direct paths.
Anticipate kid specific errors for e.g we observed pronunciation errors -
Age appropriate content
Create clean design Shorter is better. Offer fewer options and direct paths.
Anticipate kid specific errors for e.g we observed pronunciation errors -
Child
Child
Astronaut =
Winter =
Librarian =
Assistant
Assistant
Ask me not
Physicist
Life brain
Design decisions
Simplified and shortened responses to match children’s attention spans.
Broke down complex tasks into more manageable steps.
Streamlined dialogues for quicker turn-taking and reduced cognitive load.
Simplified and shortened responses to match children’s attention spans.
Broke down complex tasks into more manageable steps.
Streamlined dialogues for quicker turn-taking and reduced cognitive load.
Impact
Reduced interaction durations, improving task completion rates.
Lowered cognitive load, making it easier for children to understand and follow instructions.
Enhanced dialogue efficiency, leading to reduced drop-offs
Reduced interaction durations, improving task completion rates.
Lowered cognitive load, making it easier for children to understand and follow instructions.
Enhanced dialogue efficiency, leading to reduced drop-offs
