Team SMRT
“Planting the seeds for better public-government engagement and services.”
“Our parliament’s perspectives on the national priorities, its disparities with the citizen’s needs and expectations and how better understanding of the public's needs can create a better future.”
“We are living in a digitised world!”, this is old news in 2021. Digital technologies have transformed the way we go about our everyday business - our banking is fully digitised, we make personal and professional connections on digital platforms and we order food at home with a few taps on our smartphone.
Could you name one key pillar of our democratic society that has completely fallen behind the digital curve? Parliament springs to mind - both federal and the state. The public’s engagement with parliamentary discussion remains mostly a one-way street that is hard for citizens to navigate. The information that does trickle down is diced up, packaged, and spun by layers of third-party news organisations and commentators.
When was the last time you were aware of the parliament discussion on a topic of interest to you? Did you make time to listen to the whole parliament session? Did you find the information that you were hoping to find? Did you want to fact-check the information but do not know where to start and then other priorities crept in?
SMRT Public Indexes of Engagement (SMRT - PIE) provides a digital platform for matching a person’s interests to the national discourse. It solves the issue of disparity between the citizen’s needs and public service design by targeting the information gap between citizens and elected officials.
The platform is designed to facilitate a better exchange of ideas, interests, and needs of Australians with all levels of government to inform better policy and service design. SMRT-PIE is a platform of two-way exchange, bringing government to the people and informing government about what matters most to the community. Citizens follow issues that they care about, learn which elected officials are most vocal on these issues, understand how their own elected official has voted on these issues, and discover open government datasets related to these issues. Elected officials can easily see which issues matter most to their constituents, and shape their actions accordingly. In this way SMRT-PIE establishes a feedback loop mechanism between citizens and their elected officials that updates with the changing attitudes, preferences and service needs of the citizens.
The platform works by breaking down live parliamentary video streams into “information chunks”, finding the related public datasets accessible via ABS API, and picking out key data points such as the speaker and topic. The process can be broken into three applications as below.
In this step, the SMRT-PIE Engine - a series of machine learning (ML) algorithms and associated functions and storage - processes the parliament video stream into “information chunks” (60-second video snippets). The key steps include:
For more information, Please watch the video where we explain this process.
In this step, the SMRT-PIE Engine matches the citizen’s topic of interest against the ‘metadata’ for public APIs such as ABS API (beta), data.gov.au, state government APIs, or even international public data available via Statistical Data and Metadata Exchange (SDMX). The steps include:
For more information, please watch the video where we explain this process.
The SMRT-PIE WebApp is the front-end for citizens to find the information they need or register for an email updates about topics of interest. The steps include:
Visualising key data:
* The home page of the WebApp provides a visualisation of the day’s key data.
* On parliamentary sitting days, a stacked bar chart shows topics of discussion identified on a minute-by-minute basis by the SMRT-PIE Engine. A word cloud depicts the most common keywords used in parliament that day, and a horizontal bar chart shows which elected members spoke and for how long.
* All data is updated live every 60 seconds without needing to refresh the page.
Granular topical data:
Email Subscription (not yet built)
* We anticipate that some citizens would rather receive a point-in-time visualisation, for example at the end of each parliamentary sitting day.
* The WebApp will allow citizens to subscribe to receive email notifications on their topics of interest at a timeframe that suits them.
* We also anticipate that some citizens would like to be notified when the SMRT-PIE Engine identifies a similar topic during the processing of a “live” parliament stream, or if a topic of interest is currently under discussion in parliament. In these cases, an email notification will give the citizen a link to the parliament stream, the information-chunk, and option to receive the complete summary at the end of day.
For more information, please watch the video where we explain this process.
*The pandemic’s side effects comes in many forms. *
As the Covid-19 pandemic wears on, its impact is obviously felt in every region, every home, and indeed in every aspect of life.
Di’s Story...
Di lives in Fremantle, Western Australia. Life was usually great with nice weather and proximity to the ocean. Di runs a local business with customers and supply chains on the East Coast and overseas. The business was severely impacted by the disruptions caused by Covid-19 as well as ongoing trade tensions with overseas countries. Besides keeping the business afloat, Di has to constantly keep up-to-date with small business related discussions and announcements in WA state and Federal parliaments, and understand what it means for her business. The financial hardships and uncertainties are severely affecting her mental wellbeing.
Di wants to know about the government’s efforts to help her business enter new markets and supply chains, changes to bankruptcy laws, financial and business assistance, as well as support for mental health. She isn’t sure which level of government is responsible for which issues or policies, and doesn’t particularly care “as long as they get things done”.
Di stumbles upon the SMRT-PIE platform. She provides her topics of interest and immediately views easy-to-understand charts and graphs visualising when her topics were discussed, and by whom. It takes Di approximately 10 minutes to get across every topic at a high level, and learn which datasets are driving political and policy decisions. The SMRT-PIE WebApp’s data visualisations and list of related public datasets help Di to delve deeper for further insights. Di decides to subscribe for email notifications so she is aware whenever the topics are discussed in parliament, and so she can watch the stream live or just choose to receive the video and summary chunks at the end of the day.
Dan’s Story....
Dan is the state MP for the Fremantle electorate. Dan is young, recently elected, and eager to serve his constituents. Dan makes every effort to connect with his constituents so he can voice their concerns in the parliament and discuss emerging issues with his colleagues. Dan often worries that he just does not have enough time to connect with all of his constituents. He feels he has a good grasp of the key issues, but lacks an understanding of how his constituents feel about new, emerging, or niche issues.
One day Dan receives an email from Di, a local business owner, where she expressed her concerns for the lack of discussion in the parliament on the issues that matters most to her and many locals like her in Fremantle. In addition, Di has provided a whole list of recent parliamentary sessions and some analysis she had done using ABS and other publicly held datasets. Di also recommends that elected officials use the SMRT-PIE WebApp to better understand what matters most to the citizens.
Dan is astounded by the level of details provided by Di and immediately visits the SMRT-PIE WebApp. He selects his electorate, Fremantle, and sees the top topics of interest registered by locals. In addition, he can also find the topics related information in the form of 60 second video and text summaries from other parliaments across the country. Dan immediately reaches out to colleagues across the state government and asks them to consider supporting new services for supply chains, mental health, and business support programs. In addition, he asks his staff to use the related ABS and other public datasets to provide a rich analysis on all the issues so he can make a persuasive case in parliament that best reflects the interests of his electorate. Dan tells all of his colleagues about the SMRT-PIE platform and now he checks the platform everyday for trending topics in his electorate, and feels that he is better able to serve his constituents.
SMRT-PIE brings the government to the people and establishes a feedback loop between the citizens and elected officials for better policy and services design. The tool breaks down the parliament streams into information chunks based on topics of interest to citizens. SMRT-PIE also pulls the relevant public datasets from ABS and other APIs to enable fact checking by citizens. The portal allows citizens to nominate their topics of interest and produces local profiles that could inform the elected officials on most pressing needs of a local community and promptly design the services tailored to the community’s needs and expectations.
Better engagement and service design between the citizens and the levels of government increases public trust and thus ensures a timely, trusted, and collaborative response to address future public challenges and uncertainties.
In this step, the SMRT-PIE Engine - a series of machine learning (ML) algorithms and associated functions and storage - processes the parliament video stream into “information chunks” (60-second video snippets). The key steps include:
Real-time processing of parliamentary video streams.
ML algorithms identify the speaker.
Voice-to-text transcription of the video stream.
ML trained model matches the text transcriptions against topics of interest suggested by citizens (for example “environment”, “homelessness”, “covid”).
Creating “information chunks”: 60-second video and text summaries with data-points such as the topics of interest, a timestamp of the results, and details of the speaker. This is pushed to a database for storage.
The data generated in the process includes: Speaker Name, Topic of Interest, Video Stream link (start and end time), ML-generated summary of the discussion.
For more information, Please watch the video where we explain this process:
Link:
In this step, the SMRT-PIE Engine matches the citizen’s topic of interest against the ‘metadata’ for public APIs such as ABS API (beta), data.gov.au, state government APIs, or even international public data available via Statistical Data and Metadata Exchange (SDMX). The steps include:
ML model creates a layer of ‘meta-data’ for datasets accessible via public APIs.
ML model finds the closest match against topics of interest.
ML model saves the results in a database for later retrieval.
Data Engine GitHub: https://github.com/dronovroman/gh2021/
User Portal GitHub: https://github.com/stoicalpirate/govhack-2021
Google Drive: https://drive.google.com/drive/folders/17Eiuxmu1yZmQDF0KC3w-b2OsQBDep7ny
Fine-tuning ML for ASR: https://youtu.be/4wwLlR7farw
Scraping video streams: https://youtu.be/XUbHuHuMFEs
ASR implementation with fine-tuned model: https://youtu.be/pOAJVqa_tCI
Speaker recognition and topic modelling: https://youtu.be/E-k0n8d0-wI
Project SMRT Pie Portal overview: https://youtu.be/xk0cj4MMH7k
Project SMRT Pie Portal backend walkthrough: https://youtu.be/ftTQObQAIIs
Project SMRT Pie Portal backend data handling workflow: https://youtu.be/6OA0mujKLY8
Description of Use The SMRT-PIE Engine - a series of machine learning (ML) algorithms and associated functions and storage - processes the parliament video stream into “information chunks” (60-second video snippets). The key steps include: Real-time processing of parliamentary video streams. ML algorithms identify the speaker. Voice-to-text transcription of the video stream. ML trained model matches the text transcriptions against topics of interest suggested by citizens (for example “environment”, “homelessness”, “covid”). Creating “information chunks”: 60-second video and text summaries with data-points such as the topics of interest, a timestamp of the results, and details of the speaker. This is pushed to a database for storage. The data generated in the process includes: Speaker Name, Topic of Interest, Video Stream link (start and end time), ML-generated summary of the discussion.
Description of Use The SMRT-PIE Engine - a series of machine learning (ML) algorithms and associated functions and storage - processes the parliament video stream into “information chunks” (60-second video snippets). The key steps include: Real-time processing of parliamentary video streams. ML algorithms identify the speaker. Voice-to-text transcription of the video stream. ML trained model matches the text transcriptions against topics of interest suggested by citizens (for example “environment”, “homelessness”, “covid”). Creating “information chunks”: 60-second video and text summaries with data-points such as the topics of interest, a timestamp of the results, and details of the speaker. This is pushed to a database for storage. The data generated in the process includes: Speaker Name, Topic of Interest, Video Stream link (start and end time), ML-generated summary of the discussion.
Description of Use We scraped the key questions from the survey and matched them against the topics generated by the ML algorithms for the parliament streams. SMRT PIE tool will provide the matched datasets and API links to the user along with the data description and meta-data.
Description of Use We scraped the key questions from the survey and matched them against the topics generated by the ML algorithms for the parliament streams. SMRT PIE tool will provide the matched datasets and API links to the user along with the data description and meta-data. The results are shown in the visual layer provided on the visual layer.
Go to Challenge | 20 teams have entered this challenge.
Go to Challenge | 3 teams have entered this challenge.
Go to Challenge | 4 teams have entered this challenge.
Go to Challenge | 26 teams have entered this challenge.