Introduction
The Department of Labor (DOL) is embracing AI technologies to improve the efficiency, accuracy, and accessibility of its services. AI is being applied across a wide range of functions, from automating data extraction in complex forms to enhancing user experience with multilingual support and chatbots. These innovations are not only streamlining internal processes but also ensuring that the department can better serve the public by making information more accessible and accurate. The following use cases illustrate how AI is transforming key aspects of the DOL’s operations, enhancing both internal workflows and public-facing services.
Use Cases
-
1. Form Recognizer for Benefits Forms
This project involves a custom machine learning model that extracts data from complex forms, tagging data entries to field headers. The input is a document or scanned image, and the output is a JSON response with key/value pairs extracted by the model. This automation improves the efficiency and accuracy of data extraction from benefits forms.
-
2. Language Translation
This project uses natural language processing (NLP) models to translate published documents and website content into different languages. This helps make information more accessible to non-English speakers, ensuring that all users can understand and benefit from the Department of Labor’s resources.
-
3. Audio Transcription
This project involves using NLP models to transcribe spoken language into text. This transcription is used for record-keeping purposes, making it easier to store, search, and retrieve spoken information in a text format.
-
4. Text to Speech Conversion
This project uses advanced NLP models to convert written text into speech that sounds more natural and human-like. This technology can be used in various applications, such as virtual assistants and accessibility tools, to provide a more engaging and understandable auditory experience.
-
5. Claims Document Processing
This project trains custom NLP models to analyze physician’s notes and identify causal language. This helps in determining whether a physician’s note provides a cause-and-effect relationship, which is crucial for processing claims accurately and efficiently.
-
6. Website Chatbot Assistant
This project involves a chatbot that assists users by providing basic information about programs, contact details, and petition case statuses. The chatbot enhances user experience by offering quick and easy access to information, reducing the need for human intervention.
-
7. Data Ingestion of Payroll Forms
This project uses a custom machine learning model to extract and tag data from complex payroll forms. The input is a document or scanned image, and the output is a JSON response with key/value pairs. This automation improves the efficiency and accuracy of data extraction from payroll forms.
-
8. Hololens
This project uses AI integrated with Hololens technology to allow inspectors to visually inspect high and unsafe areas from a safe location. This enhances safety by reducing the need for inspectors to physically access dangerous areas.
-
9. DOL Intranet Website Chatbot Assistant
This project involves a conversational chatbot on the Department of Labor’s intranet websites. The chatbot helps answer common procurement and contract-related questions, improving internal communication and efficiency.
-
10. Official Document Validation
This project uses AI to detect mismatched addresses and garbled text in official letters sent to benefits recipients. This ensures that the documents are accurate and clear, reducing errors and improving communication with recipients.
-
11. Electronic Records Management
This project uses AI to identify data within documents and NLP to classify and summarize them, ensuring that electronic records meet NARA metadata standards for permanent federal documents. This improves the organization and accessibility of federal records.
-
12. Call Recording Analysis
This project involves the automatic analysis of recorded calls made to Benefits Advisors in the Department of Labor’s Interactive Voice Response (IVR) center. The goal is to improve the efficiency and accuracy of call handling by analyzing call content automatically.
-
13. Automatic Document Processing
This project automates the processing of continuation of benefits forms by extracting data from pre-defined selection boxes. This streamlines the handling of these forms, making the process faster and more accurate.
-
14. Automatic Data Processing Workflow with Form Recognizer
This project uses an automated system to process complex workflows and extract necessary data. This improves the efficiency and accuracy of data handling in complex processes.
-
15. Case Recording Summarization
This project uses an open-source large language model to summarize publicly available case recording documents that do not contain personal identifiable information (PII) or other sensitive information. The summaries are reviewed by human note takers to ensure accuracy. This helps in making case information more accessible and easier to understand.
-
16. OEWS Occupation Autocoder
This project involves an autocoder that processes state-submitted response files containing occupation titles and job descriptions. It assigns up to two 6-digit Standard Occupational Classification (SOC) codes with probabilities as recommendations for human coders. Codes above a certain threshold are added to the response file and sent back to states to help with SOC code assignment, improving the accuracy and efficiency of occupational classification.
-
17. Scanner Data Product Classification
This project uses machine learning to classify bulk data received from corporations about the cost of goods and services. The machine learning model uses word frequency counts from item descriptions and logistic regression to estimate the probability of each item being classified into Entry Level Item (ELI) codes. The highest probability category is selected, and classifications below a certain threshold are flagged for human review. This process helps in efficiently and accurately categorizing data for the Consumer Price Index (CPI).
-
18. Expenditure Classification Autocoder
This project involves a custom machine learning model that assigns reported expense descriptions from Consumer Expenditure Diary Survey respondents to specific expense classification categories, known as item codes. This automation improves the accuracy and efficiency of expense classification.