Team Name:
Hacks
Team Members:
Hacks
This challenge focuses on operationalising and expanding the application of the Style Manual through AI tools. We explore innovative ways to improve the accuracy, efficiency and inclusivity of government content. By leveraging AI, we can focus the government communication around the needs of people, adhere to Australian style, spelling and conventions, and make content accessible to diverse audiences.
Government documents often use complex language that is not easily understood by the general public. This complexity can hinder people’s ability to make informed decisions and comply with their obligations. The Australian Government Style Manual provides comprehensive guidelines to ensure clarity and consistency in government writing, but applying these guidelines manually can be challenging and time-consuming.
Most large language models are trained on global internet data, which predominantly uses US English conventions. Our project aims to tackle this issue by ensuring that AI tools used in government accurately recognize and apply Australian English spelling, grammar, and style guidelines. This adaptation will make AI-generated content more relevant and aligned with local standards.
We address the issue by leveraging AI to transform bureaucratic jargon into plain Australian English, ensuring government content is clear, accurate, and inclusive.
To prevent errors and misinformation in AI-generated content, we explore the use of retrieval-augmented generation (RAG). This integration helps in avoiding "hallucinations" and maintains the integrity of government communications. Furthermore, our AI tool is calibrated to apply the Australian Government Style Manual guidelines to generate the content, paraphrasing, and feedback.
We plan to implement the solution according to the proposed Initiative for Accessibility, Trust, and Inclusion.
Description of Use The Australian Bureau of Statistics (ABS) Education dataset provides extensive information on school enrollments, higher education, vocational training, and qualifications in Australia. It can be utilized to train a large language model (LLM) by feeding it structured data on educational outcomes, trends, and demographics. The model can be enhanced with retrieval-augmented generation (RAG) to provide real-time, data-driven outputs, such as answering questions about Australian education statistics, making predictions, or generating insights based on the latest trends in the dataset. This integration ensures accurate, updated content generation for research and policy-making purposes.
Description of Use The Australian Election Study (AES) dataset contains valuable data on voter behavior, political attitudes, and election results in Australia. It can be used to train a large language model (LLM) with retrieval-augmented generation (RAG) by integrating historical voter preferences, political trends, and public opinion data. This enables the model to generate accurate, contextualized responses on topics like Australian elections, voting trends, and political sentiments, providing real-time, data-driven insights for research, analysis, and reporting.
Description of Use The Australian Tax Gaps dataset provides estimates of tax gaps across different segments of the tax system. This dataset includes information on the differences between the amount of tax collected and the amount that should theoretically be collected, offering insights into compliance, revenue loss, and the efficacy of tax administration. By analyzing these gaps, the dataset helps assess the health of Australia's tax system and supports policy-making to enhance compliance and reduce tax evasion.
Description of Use This dataset is used to be trained by the LLM. The LLM will then generate outputs based on this dataset depending on the user's input.
Description of Use The dataset for this project is derived from the Australian Government Style Manual, which is a crucial resource designed to ensure consistency and clarity in written content across the Australian Government. It provides comprehensive guidance about writing for the government, including grammar, punctuation, tone and inclusive language. The Style Manual helps improve the quality of government writing, making it more accessible and user-friendly. The dataset includes various types of government documents that adhere to the Style Manual guidelines for Policy Documents, Reports, Forms, Briefs, and Digital Content. The dataset also encompasses the rules and guidelines outlined in the Style Manual, such as Grammar and Punctuation, Tone and Style, Inclusive Language, Spelling and Conventions, and Accessibility and Inclusivity.
#AI in Governance
Eligibility: Open to everyone.
Go to Challenge | 35 teams have entered this challenge.
#Civic Participation
Eligibility: Open to everyone. Submissions should use at least one government data source.
Go to Challenge | 27 teams have entered this challenge.
#Digitally Fair
Eligibility: Open to everyone. Use at least one ATO data set
Go to Challenge | 14 teams have entered this challenge.
#Improving Service Accessibility
Eligibility: Australia only.
Go to Challenge | 29 teams have entered this challenge.
#PLaiN LANGUAGE
Eligibility: All Australian competitors.
Go to Challenge | 23 teams have entered this challenge.
