Innovation for Democracy

How Does Taiwan's Civil Society Leverage Technology and Citizen Activism to Enhance the Transparency of Parliament?

In 2025, many citizens in Taiwan were dissatisfied with how parliament reviewed and cut the national budget. Civil society organizations and advocates criticized the budget review process for turning budget cuts into a tool of political retaliation between parties, with essential funding for national defense, critical infrastructure, and social welfare being significantly cut or frozen without clear justification.

Rather than waiting for solutions, these groups took action. By leveraging technology, data, and citizen participation, they developed a crowdsourcing social media bot and a data visualization website to monitor how legislators decided to cut or freeze government budgets. This initiative not only increased transparency in the budget review process but also addressed long-standing issues in parliamentary open data, which for decades had been released in fragmented, unstructured, and non–machine-readable formats.

 In this article, data journalist Zoe Lee from READr, who led the project, explains how Taiwan’s civic tech community once again mobilized to advance parliamentary transparency.

What Are the Bot and Website About?

The website publishes core budget proposal documents and tracks progress in real time. It also visualizes the data, clearly showing which budgets have been cut or frozen and which legislators made those decisions, providing a comprehensive overview of budget allocations across different ministries and agencies. As a result, citizens no longer need to search through scattered budget amendment documents or calculate funding cuts themselves, and they have clear, reliable evidence to hold legislators accountable.

Building on the original version, the new website incorporates interactive features and crowdsourced collaboration. This is where the bot comes in. Built on LINE—Taiwan’s most popular messaging app—to maximize user engagement. It guides users through a structured question-and-answer process, inviting citizens to help the project team convert image-based budget amendment proposal sheets submitted by legislators into machine-readable text.

A backend data management system has also been established. In the future, the website will not only support ongoing updates to the annual budget review process but also allow historical budget proposal data to be imported, making it easier for citizens to reflect on and compare patterns across different legislative terms.

While the website is not an entirely new project, efforts by Citizen Congress Watch (CCW) and FNF’s Global Innovation Hub in 2025 have made it more participatory and accessible. Through these improvements, it has evolved from a one-off news resource into a platform that can be continuously updated by citizens. This development also marks a new milestone in the Taiwanese civic tech community and civil society’s 12-year journey.

Why Is It So Difficult to Oversee the Parliamentary Budget Review Process?

The 2025 central government budget review in Taiwan’s parliament faced a series of unprecedented setbacks, as discussed earlier (see A Comprehensive Analysis of Why the Budget Proposal Got Stuck). Yet even before these events unfolded, I had been discussing a long-standing question with Ronny Wang—a veteran member of Taiwan’s largest civic tech community, g0v, and founder of OpenFun Ltd.: could parliamentary budget review data be released in a standardized, accessible format?

Ronny and I have collaborated for many years on open government data projects, including efforts to make political donation records and legislative voting data publicly accessible. This was not our first attempt to tackle parliamentary budget review data. In fact, despite years of collaboration, budget review records remain among the least transparent datasets maintained by the Legislative Yuan.

Some may argue that the data is already transparent because it is publicly available online. Technically, that is true. However, anyone who has attempted to follow the full review process quickly encounters significant barriers: inconsistent document formats, fragmented sources, and a lack of structure that makes meaningful analysis extremely difficult.

For example, to determine which legislators cut or froze the Ministry of Culture’s proposed budget before it was finalized, one must first locate the meeting minutes of the Education and Culture Committee to identify who proposed amendments and for what reasons.

But the process does not end there. After committee review, budgets go through party negotiations as well as second and third readings in plenary sessions. During these stages, legislators may introduce new amendments, modify existing ones, or withdraw proposals entirely. This requires tracking additional meeting records and repeatedly locating references to the Ministry of Culture’s budget. And this is just one ministry. Each year, Taiwan’s central government budget involves hundreds of agencies and thousands of proposals. While most of this information exists online, it remains effectively inaccessible.

At the time, I took over the raw data Ronny had extracted from parliamentary meeting minutes and tried to organize and archive it in various ways, while still struggling with how to present such a complex dataset to readers. Before we could arrive at a solution, however, the 2025 budget review controversy erupted, triggering intense debate across social media. That controversy became the catalyst that pushed us to act.

How Budget Review Controversies in 2025 Changed the Way Information Was Presented

As public debate intensified, some legislators began posting budget amendment proposals online to accuse opposing parties of “recklessly cutting budgets.” In response, those accused posted proposals they claimed demonstrated irresponsible behavior by their opponents. While these actions did increase disclosure, they presented only fragments of the full picture.

But these hectic debates on social media made it clear to us that there was strong public demand for comprehensive, non-selective data. Rather than aggregated statistics, citizens wanted access to the full content of budget amendment proposals. After internal discussion, our team worked intensively for four days—thanks to colleagues who sacrificed their weekends—to launch the first version of the oversight website.

The platform displayed every budget amendment proposal, including who submitted it, the reasons for cuts or freezes, and the amounts involved. Readers could react using “like” or “angry” emojis, and READr journalists tracked each proposal through to its final outcome in the third reading.

The response was immediate. The platform received widespread attention and a flood of suggestions for improvement. As an open-source project, engineers contributed UI and UX enhancements, while engaged citizens helped parse the data and identify additional categories essential for monitoring the budget review process. Even as our small team was stretched thin—READr is fundamentally a newsroom, and we published five in-depth reports during the review—the project continued to grow.

We also received valuable feedback from scholars and civic organizations. Several academics told us this was a project they had long wanted to pursue but were discouraged by the high barriers to processing the data. Citizen Congress Watch noted that although budget review performance is part of their annual legislator evaluations, the complexity of the data requires substantial manpower.

Why Open Data and Data Journalism Matter

Under an open-source framework, READr has previously developed several platform-based projects. One example is the Policy Tracker created for the 2022 and 2024 elections, which allowed readers to contribute data on candidates’ policy pledges and collectively monitor their implementation. READr journalists compiled policy records for mayors in Taiwan’s six major cities, tracked campaign pledges, and produced related reports.

Publishing the underlying data transformed these projects. Rather than merely presenting campaign promises in news articles, we opened the data, shared our methodology, and made it reusable. Others could replicate the process, expand the dataset, or apply it to new contexts.

A similar success came through our collaboration with Tainan Sprout, a local civic oversight organization. Among all cities and counties in Taiwan, the policy datasets for the 2022 Tainan mayoral election and the 2024 legislative election were the most comprehensive. Tainan Sprout had already been collecting the data; READr helped define the structure and aggregate the results.

However, newsrooms inevitably need to move on to the next story. It can’t always spend so much time on collecting data for one topic.

The Challenge of Data Maintenance

Ideally, civic organizations maintain and expand these datasets. Environmental groups can track environmental policies, and parliamentary watchdogs can monitor legislative behavior.

But, in practice, integrating these tools into NGO workflows requires more time, planning, and manpower than they can afford. As a result, many collaborations remain one-off projects tied to elections or major events.

READr faces similar limitations. We are a news organization, not a permanent parliamentary watchdog. If future budget reviews generate less controversy, sustaining the same level of resources will be difficult.

This time, however, Citizen Congress Watch and the Global Innovation Hub of the Friedrich Naumann Foundation for Freedom (FNF) recognized the project’s potential and offered support to tackle that problem.

Building a Sustainable Process

Because Citizen Congress Watch relies on budget review data for its daily oversight work, we explored how the original platform could be transformed into a long-term tool. With support from CCW and FNF, we re-engineered the website and redesigned the data-building process so that even users without advanced technical skills could operate it.

Previously, the platform relied on a single spreadsheet containing over 8,000 budget amendment entries, making management inflexible. We developed a backend system that allows administrators to edit and search entries individually, while establishing relationships between different columns such as review meetings, co-signing legislators, and budget categories.

We also minimized manual work. Parliamentary meeting minutes can now be converted into structured budget proposal data with a single click, and extracted content from government documents can be uploaded just as easily. This significantly reduces CCW’s reliance on manual transcription.

Still, some obstacles remain. Many amendment proposals exist only as image-based PDFs, and formatting varies by committee and legislator, making automated processing difficult.

Let Citizens Help

These limitations do not affect how final review outcomes are displayed. Our primary source for conclusions is the official proceedings record released by the Legislative Yuan, usually within a week of each session and available in text format. These records list all proposals that passed or were reversed and can be processed programmatically.

However, some amendment proposals are withdrawn or merged during negotiations. The original documents appear only in the Legislative Gazette released a month later—and only as images. Without these materials, it is impossible to fully understand the negotiations behind the final outcomes.

To address this, we introduced a new citizen collaboration feature: a LINE bot. By adding the bot on LINE—Taiwan’s most widely used messaging app—users can help convert image-based proposals into structured, machine-readable data in just a few minutes.

The bot presents users with proposal images and guides them through identifying key information: proposed cuts or freezes, amounts, sponsoring legislators, and the type of motion involved. Through this process, participants not only contribute data but also gain a deeper understanding of how budget reviews work.

This approach builds on our experience digitizing political donation records, which were once only accessible in person at the Control Yuan. At the time, we scanned and segmented the records so volunteers could transcribe them CAPTCHA-style. Over time, READr and civic volunteers built comprehensive datasets spanning multiple legislative terms. Eventually, the law was amended to require online publication, eliminating the need for manual processing.

Looking back, we often ask whether we would be where we are now if we had waited. When the political donations project launched in 2014, officials advised us to “wait for the government.” It took four more years for open data legislation to pass.

Our conclusion is clear: progress comes from action, and citizen participation is what makes long-term change possible.

The True Meaning of Open Data

We are deeply grateful to Citizen Congress Watch and the Friedrich Naumann Foundation for supporting this project as an open-data, open-source initiative. All contributed data will be publicly available, enabling reuse across sectors. After eight years in data journalism, I am convinced that barriers to data access severely limit both academic research and public understanding. A solid, open data foundation is the starting point for any meaningful inquiry.

We will continue developing this project within the g0v Open Parliament Hackathon community. If you are interested in joining the discussion or participating in future events, you are welcome to join the g0v Slack channel and connect with us. And you can find the source code of this project here on github and fork it!

*This article is written by Zoe Lee, Feature Producer and Data Journalist, READr