I began my library career in the 1960s at a large public library. I can remember the basement full of government documents ranging from tiny pamphlets to oversized books. Each week, a new supply of them would arrive, and I would head down to interfile the new entries with the old, using a totally inscrutable filing system. These items were there for everyone to use, but I can’t recall ever retrieving one for a patron. Since I had the contents to myself, I did occasionally discover a gem in those stacks. Most memorably, I found publications from NASA containing hundreds of full-page photographs of Earth taken by Gemini astronauts.I knew that discovery would improve for these works once they were digitized. By the mid-1990s, basements full of documents morphed into smaller spaces full of terminals reading CD-ROMs. This was much better for easy access, but there was still a problem: Most of the materials were written in a language best known as “Governmentese.” The writing was accurate, but was in a style that people found unpalatable.
BEYOND CD-ROMS
After the mid-1990s, there was a governmental gold rush to place all of these materials where they could be accessed by citizens 24/7. The Federal Register (a compilation of all regulations in force or proposed for the future) was mounted on the web in 1994 and continues today. The Code of Federal Regulations is online as the e-CFR and is updated every day. It is a listing of all regulations in effect, divided into 50 broad topics such as health and transportation. In the 1990s, Congress tried to make the Code of Federal Regulations entirely electronic, but it lost to a groundswell of objections from people who did not want to go entirely paperless.
In 2003, Regulations.gov launched. This site helps users learn about new and proposed rules while encouraging them to give feedback—allowing them to have a stake in the actions of their government. It seemed like this was as far as the government could go in using automation to open doors to government information. It wasn’t until the 2020s that AI became a force in the world of human-computer interactions.
AI TO THE RESCUE
While it was clear that strides were being made in accessibility, more work was needed. To that end, a New Jersey-based company named Gadget Software developed a new program, TopicLake Insights, which adds AI-based search-and-retrieval software to the entire corpus of federal regulations. It is an API—a set of rules or protocols that enables software applications to communicate with each other to exchange data, features, and functionality. The product addresses search limitations by expanding queries to include a word cloud. This helps the searcher to expand the possibilities of search to find new connections with similar concepts.
Searches in TopicLake Insights will retrieve AI-generated summaries of regulations and then move on to the official regulations themselves. Because it is, for now, a free program, my sign-up process was quick and seamless. Once enrolled, you are able to choose up to five agencies to track. I wasted no time in choosing NASA and the Library of Congress. Initially, you can see what regulations are in transition at the current time. After that, you will get daily briefings for any changes in the agencies that you have selected.
A MAN WHO HELPED TO MAKE IT HAPPEN
I was very happy for a chance to speak on the phone with Max Riggsbee Jr., co-founder and chief product officer at Gadget Software. (Our conversation occurred before the Nov. 5 elections.) It turned out to be a spirited discussion that delved into the big questions about the direction of information science. Riggsbee said that government information is a perfect testing zone for AI enhancement. This is because there is a massive universe of data that is entirely in the public domain and a consistency of writing style that covers the entire corpus. Before moving ahead with introducing TopicLake Insights, Riggsbee noted that he and his team spent considerable time in Washington, D.C., talking to the main agencies.
Riggsbee told me that the free API training portal launched in March 2024 and that the free Policy Alerts product launched in July 2024. TopicLake Insights’ website attracts more than 5,000 weekly visitors. Law schools, government workers, and individual lawyers in corporate practices are the typical users. There is much searching being done, and the data TopicLake Insights creates helps to determine the future of the program. Riggsbee emphasized that this data collection shows only the types of data being used the most and is not connected with any actual searchers.
We talked about the nature of AI-generated metadata in searching and visual discovery. As a former library cataloger, I am aware that metadata creation evolved from traditional systems such as library card catalogs. This means that searching is essentially linear. When you search for, say, asteroids, you get results that mention the term in certain documents. By adding a word cloud, the AI-generated metadata unlocks access to topics that relate to the keyword(s) you selected, such as meteor, black hole, and orbit. The combination of keyword search and visual discovery leads to a more satisfying outcome. I was impressed that Riggsbee is interested in the librarian’s perspective on search effectiveness. That has been a driving force in my scholarship since the day in the mid-1990s when I heard OCLC founder Fred Kilgore speak on finding the perfect balance between effective search results and an impractical flood of data.
TIP ON A NEW PRODUCT
Early in our talk, Riggsbee asked if I had tried a product from Google called NotebookLM. I admitted that I had not, and before the day was out, he sent me the link. Because I have a Google account, I was signed in as soon as I entered the program. The way it works is very simple. You round up text from a variety of sources, place it in one file, and copy and paste it into NotebookLM. Seconds later, it creates a well-balanced summary of the ideas you gave it, including outlines and test questions. I fed it copies of everything I’d ever written about AI and got a very impressive piece in return.
CONCLUSION
Government information access has made massive shifts throughout the duration of my library career. Far from the days of dusty library basements, it is now available to anyone at any time. Riggsbee and I agreed that the main issue of discoverability is found in all forms of scholarly communication. The solution is to make information available to the user when they search using their own language and not rigid rules of classification. Then give them abstracts that are also in language that the average educated user can digest. It sounds simple, but the devil is in the details. TopicLake Insights is a worthy step forward.