Team Name:
+852
Team Members:
+852
Accessing government services, such as the Australian Taxation Office (ATO), presents challenges for individuals who are deaf or have hearing and speech impairments. Current solutions like the National Relay Service (NRS) involve human intervention through chatbot or video relay for sign language, which can be slow and labour-intensive. Our proposal aims to enhance the NRS using AI to provide more accessible, secure and private communication options.
1. Deaf, Hearing, and Speech-Impaired users
2. Call centre operators at the National Relay Service (NRS)
Bonus. Non-English speak user
We propose augmenting the NRS with AI to assist in chatbot and video relay services for sign language. This includes an application that enables users to interact via chatbot or sign language when calling government services.
A chat-based interface: AI converts text to voice and converts voice to text. (Can be extend for language translation)
Sign language video recognition by AI and converts to voice. Voice can be converted by AI to sign language.
Personal information, such as tax numbers, can be securely stored in the app and easily retrieved with a simple approval click, eliminating the need for manual entry or signatures.
Continuous verification via mobile camera can enhance trust, similar to a phone call.
The chatbot requires user approval before sending any personal information
We propose integrating any text<->speech (e.g Google cloud text to speech and Google cloud speech to text ) and Sign Language Recognition to handle phone calls on behalf of users, providing sign language recognition and speech-to-text functionalities.(research.google, n.d.;Paperswithcode.com, 2019)
Data is encrypted with two keys—one held by the user and the other by the government. Data can only be accessed with consent from both parties, enhancing privacy.
We will actively engage users by organizing co-design workshops and focus groups, including individuals from the deaf, hearing-impaired, and speech-impaired communities. These sessions will gather critical feedback during the design and development stages, ensuring that the solution aligns with their needs and expectations.
By involving advocacy groups and government agencies early on, we can foster a collaborative environment, ensuring the app meets accessibility standards and stakeholder requirements.
To build public trust, we will employ a two-key encryption system. This system ensures that all data shared in dialogues between users and government services remains encrypted.
- User’s Key: The user controls one encryption key.
- Government’s Key: The government holds the other encryption key.
Data can only be accessed when both the user and the government explicitly agree to its use. This ensures data privacy while maintaining government compliance.
To enhance user experience, the app will store certain pieces of personal information, such as taxation numbers, securely. This will allow users to quickly approve the use of stored information (e.g., answering identity verification questions) without needing to type or use sign language.
By simplifying the process, the app enhances accessibility while maintaining privacy through user-controlled data sharing.
While modern mobile devices have sufficient processing power to perform both sign language recognition and text-to-speech conversions locally, doing so would limit auditing and verification capabilities in case of any disputes during the call. To maintain a high level of trust similar to traditional phone call recordings, all translation and processing are performed on the backend. This decision ensures full audibility and accountability for the system.
Reduced Human Intervention: The proposed solution reduces the need for human involvement in NRS, speeding up the communication process and lowering the operational burden.
Increased Trust: Our use of encrypted data, controlled by both parties, alongside transparency in AI decision-making, helps build and maintain public trust.
Avoiding Misuse: Since the app is government-provided and restricted to official interactions, misuse can be prevented.
N/A
Go to Challenge | 29 teams have entered this challenge.
Go to Challenge | 35 teams have entered this challenge.
