Our goal is to raise awareness of government spending practices to prevent government corruption. According to a study by the World Bank, $2,500,000,000,000 (you read that right, $2.5 Trillion!) is lost annually in the construction industry alone due to corrupt political and business practices.

Our inspiration for tackling this problem was a highly visible recent instance of this corruption. Hurricane Maria ravaged Puerto Rico in September 2017, killing thousands, upending tens of thousands, and completely destroying the infrastructure of the island. To make this terrible situation even worse, the government-run utility company of Puerto Rico, PREPA, issued an absurd contract to rebuild the island's electrical infrastructure the same day as the hurricane hit. The $300M contract was won by Whitefish Energy Holdings LLC, a Montana company with only 2 employees, which had been formed just 6 months prior. It was glaringly apparent that Whitefish was incapable of handling this massively important contract and information came out in the months after confirming that there were shady dealings behind this contract award. Not only was this a money sink, but this added further duress to the citizens of Puerto Rico, some of whom waited nearly a year (328 days) to get power back due to the ineptness of Whitefish and ensuing re-compete of the contract.

For more information about government corruption and the Whitefish incident see:

What it does

We built a platform capable of detecting corruption using publicly available government procurement data from, as well as a system to visualize and quantify the extent of this corruption. On the landing page, a user enters the name of a company that contracts with the federal government. Upon entering a company, our novel corruption index (COIN) score will be displayed, giving an indication on a scale from 1 (extremely corrupt) to 100 (not corrupt) . We intend the COIN score to be a sort of FICO credit rating for government contractors, and we hope to raise public awareness of non-honest entities.

We also used natural language processing and sentiment analysis from the Google Cloud API to determine relationships between key players in a scandal using news articles written about the topic. Our program iterated through the most recent publications on a scandal, pulling out key names and places, and developing a connection network between them. This network of connection is then displayed graphically, so that the user can understand how the key players in a scandal are connected.

How I built it

We used natural language processing and sentiment analysis from the Google Cloud API to build the model for the connections display. The graph of the connections was made using R Shiny. The COIN score was determined by importing information from in to R and extracting the metrics we deemed important in calculation our corruption index score.

Challenges I ran into

Although we originally thought sentiment analysis would have strong predictive power to determine who the shady characters are in a given corruption case, we found that we couldn't get strong information using just this metric. Using the Whitefish scandal as an example, we developed our own algorithm which looked at the proximity and frequency of names and companies deemed to be key players in an event.

Accomplishments that I'm proud of

The output connections graph in R Shiny is able to identify connections between key players in other high-profile corruption cases like Theranos and the Harvey Weinstein trial.

What I learned

We learned about how to use Google Cloud's natural language processing and sentiment analysis, how to analyze data for social good, and how to make beautiful and user-friendly displays in R Shiny.

What's next for bidwatch

Awareness is one of the most powerful tools to promote social justice, ensure equality, and upkeep the quality of a democracy. On the contrary, lack of access to information reinforces old habits, old feedback loops and vicious networks, both in individual and societal levels.

Corruption in top-management-level networks and in procurement are widely considered to be the most harmful kind of misbehavior, since it usually leads to enormous amounts of undesired expenditure of public funds.

Bidwatch was developed with the vision and intention to fight corruption with the help of AI based on natural language understanding of news pieces and other publicly available data. This web, integrated with public governmental data on procurement contracts, provides insights to everyone with a very user friendly interface.

We created a COIN Score, with which we also aims to simplify the interpretation of information based on data provided by the US government. In the near future, machine learning capabilities will allow Bidwatch not only to detect past frauds and misappropriations, but also to indicate potential future fraudsters.

Join us in transforming the world, to ensure a brighter tomorrow!

Share this project: