OpenAI Streamlines Data Collection for Residential Fire Deaths

FEMA’s educational mission for residential fires 

One of the purposes of the Federal Emergency Management Agency’s (FEMA) Fire Administration is to manage disasters by providing risk reduction education. Each day via internet search, FEMA culls the previous day’s information related to residential fires with fatalities and whether a functional smoke detector was present. This daily data gathering helps FEMA raise awareness regarding the danger and frequency of residential fire deaths through its Home Fire Fatalities in the News landing page. The process was manual, requiring several hours to an entire day of Google searching by a single employee. 

Transitioning to an automated process 

Recognizing the time savings a modernized workflow could offer, FEMA enlisted Tepa LLC to develop an automated process. Tepa utilized Argis Solutions, Inc. as a subcontractor to achieve this goal. 

Working together, Tepa and Argis developed an automation of the data collection process utilizing Python scripting and web scraping. A Google custom search engine scans for any news articles related to residential fire incidents using common search terms and develops a list of reported incidents. Text from each article is scraped to collect further information such as the date, location, cause, and number of fatalities using the OpenAI API. Once the information is parsed out, the script creates an Excel report. The report sorts articles by incident date and location so duplicate stories can be easily viewed and the most accurate article chosen from the selection of duplicates. At the end of each nightly run, the program emails the report to any email addresses provided for FEMA employees to review in the morning. 

Data collection in seconds 

With the new automated process, this information takes only six to eight seconds to collect, offering considerable time savings. As OpenAI continues to learn and improve its searching ability, the results will become even more refined, continuing to improve the process. 

“Working with Argis is a real pleasure,” said Nathan Wysocki, GISP and technical project manager for Tepa, LLC. “The team we formed worked very collaboratively and seamlessly to solve a very complex and uncommon request from our client. The real challenge of this project was determining the best method for automating the daily scouring of the internet for house fire fatalities reported in news articles. Argis suggested and delivered an AI solution that significantly reduced the number of articles needed for review by a staff member, lowering the number of daily human hours required to search for and catalog results.”