From September 17 to 19, 2024, 11 OXD team members joined seven other teams for the “Cryptid Coders in the Wilderness of BC” Hackathon, organized by the BC Ministry of Water, Land, and Resource Stewardship (WLRS). The challenge was to solve a real-world problem or identify new opportunities around accessibility challenges for BC Parks.
Fun fact
To align with the “Cryptids of BC” theme, we used AI to generate the team name “Gastown Gumberoos”—a reference to a mythical, leathery-skinned bear from lumberjack folklore.
Illustration of The Gumberoo by Richard Svensson, 2008.
OXD has been working with BC Parks on improving the bcparks.ca website over the past few years. Because of our long-standing relationship, we wanted to use the Hackathon to test cutting-edge AI technologies and crowdsourcing approaches to enable BC Parks to provide accurate, up-to-date information about accessible facilities in the parks.
The problem
BC Parks is dedicated to preserving British Columbia's natural spaces and ensuring recreation in those spaces is accessible for everyone. However, their current accessibility microsite only provides information for a narrow set of accessibility concerns and lacks updated information for a wider audience with diverse needs.
Our Hackathon challenge
Our team identified valuable but outdated information on accessibility.bcparks.ca. When the site was initially built, park students collected photos of accessibility features, but BC Parks staff didn’t have the resources to regularly update the site. We needed a solution to keep it up-to-date.
Our solution: Using AI tools and crowdsourcing.
Our goal was to create a dynamic, personalized resource for park visitors with diverse accessibility needs. Using AI, we envisioned a platform where users could find up-to-date information. Using crowdsourcing, we were able to solve the lack of up-to-date accessibility feedback.
Example Use Cases:
- A visitor with a broken leg could check if paths are crutch-friendly.
- A visitor with limited vision might want screen-reader-friendly park descriptions.
- A neurodivergent visitor could determine if a park is too busy for comfort.
With AI-powered semantic search, users could search terms such as “no stairs,” and the system would return related results, including “senior-friendly” or “stroller-accessible,” even without exact matches.
Our approach
OXD assembled a team of 11 experts, including six staff members from our ongoing BC Parks project and five additional enthusiasts. Using the existing data from accessibility.bcparks.ca, we developed a mobile-first web app to:
- Allow users to upload park accessibility feedback and images.
- Use AI to tag accessibility features from visitor feedback
- Provide intelligent, semantic search functionality by finding tags that matched a user’s search query.
- Use relational database queries to return parks with matching tags.
We then built wireframes for two user flows: visitors looking for accessible parks and visitors adding new information about accessible features.
Our technology stack included:
- OpenAI API for natural language processing (NLP)
- Django + Django REST Framework + PostgreSQL for database management
- Vue.js + Quasar for frontend development
- Docker for local development containerization
- GitHub with a CI/CD pipeline that automatically built the app and deployed it to the Amazon cloud infrastructure (as mentioned below) as the code was changed or updated
- Hosted on AWS (Elastic Container Service, S3) for cloud infrastructure
Our solution
In two-and-a-half days, we developed a prototype that demonstrated the power of combining AI with crowdsourced data to enhance park accessibility information. Using semantic search, we were able to take unstructured information (information that either doesn’t have a predefined data model or isn’t organized in a pre-defined manner) from existing visitor feedback text and output the data to support a broader range of search needs.
Key features included:
- User-friendly accessibility review submission
- AI-powered tagging for a standard list of accessibility features; when a park visitor submits feedback, the model will most relevant tags
- Semantic search to match user queries with relevant park features
- Integration of existing data from the website for enhanced accuracy
The possibilities this opens up for accessibility at our beautiful BC parks are awesome! Well done, OXD Gastown Gumberoos.”
Hackathon organizer
Our team identified several additional areas for improving park accessibility through AI tools and crowdsourcing that couldn’t be fully explored within the hackathon’s time constraints. These include collecting more granular feedback on specific park areas, refining AI models for better tagging accuracy, and using AI to assess visitor feedback sentiment. We also saw potential in AI-supported image moderation, analysis, and data extraction to streamline crowdsourced information, as well as gamification strategies to encourage real-time user contributions.
[...] OXD took this opportunity to explore some potentially cool ideas. It’s amazing how much creativity can come out when folks have a couple of days to focus on something. It’s great inspiration for the team as we explore a permanent and modern solution to this work."
Hackathon organizer
Our results and their future impact
Within the two-day hackathon, our prototype showcased the power of combining AI and crowdsourced data to enhance park accessibility. Our app allowed users to submit accessibility feedback, which AI tagged with standard features and processed through semantic search to match users’ specific needs. Features included user-friendly submission forms, intelligent tagging, and integration of existing data for accuracy. The BC Parks team was enthusiastic about the potential: “The possibilities this opens up for accessibility in our parks are awesome!”
At OXD, we believe technology can make public spaces more accessible for everyone. Want to explore how AI and crowdsourcing can enhance your services? Contact us today to bring innovative, user-centred solutions to life.