Optimal Solutions Group, LLC (Optimal) is presenting its customized chatbot, OptiBot.AI, at the annual Federal Committee on Statistical Methodology (FCSM) conference in Hyattsville, Maryland, from October 24–26, 2023. OptiBot.AI is a chatbot capable of responding to user queries pertaining to the federal data that it is trained with. Optimal has developed an OptiBot.AI to help federal statistical agencies make their data tables more accessible. OptiBot.AI allows users to query federal data tables by asking any question the table(s) is able to answer.
Federal statistical agencies present a vast majority of the data they collect in a tabular format, which means that inquiries into these data largely depend on the end user’s ability to understand and access these tables.
This provides several challenges. For instance, many tables are large and complex and incorporate features like indents and nested tables, which can make the tables difficult to read. In addition, these tables are often embedded in a front end that includes some built-in constraints (e.g., it can only access one table at a time). Factors such as these can make the experience of working with tabular data taxing and time-consuming for the average user. OptiBot.AI effectively addresses these challenges by interacting with just one application. By removing the burden of manually navigating multiple tables, OptiBot.AI can expedite information retrieval and data analysis.
OptiBot.AI is powered by ChatGPT using its API on Python, and its underlying Large Language Model (LLM) is being trained on various federal statistical tables such that the corpus is limited to these tables and specified so that it only provides precise answers. Additionally, OptiBot.AI would allow end users to ask questions about the data in plain language and receive accurate answers in plain language, summarize the findings, identify highlights, and draft content on the results for a wide range of audiences. OptiBot.AI marks a considerable step forward in making federal data more accessible to a variety of users, including researchers, policy analysts, and the general public. Researchers often find it challenging to summarize the results presented in a table to inform subsequent analyses and research. Similarly, policy analysts may find it difficult to effectively summarize the results provided in a table to inform policy or assess a policy’s impact. The general public, on the other hand, may be unfamiliar with the data collected by a federal agency, and they may require training; or they may have difficulty interpreting the results in the table. OptiBot.AI provides a cost-effective solution to these challenges by substantially reducing barriers such as users’ lack of familiarity and training involving federal data.
OptiBot.AI is part of Optimal’s Real-Time Framework, which enables timely public policy evidence-based decision-making. In addition, OptiBot.AI will help agencies become more Algorithm Optimization (AO) friendly, reflecting a newer trend over Search Engine Optimization (SEO). OptiBot.AI will undergo pilot testing next, followed by determining the latent user base and, finally, deployment. OptiBot.AI will undergo continuous testing to decrease barriers to access, which will allow more highly skilled users, as well as the general public, to utilize federal statistical data.