Artificial Intelligence (AI) for Past Performance Evaluation
Introduction
The President’s Management Agenda calls for using automation software to improve efficiency of government services. A key focus area is past performance evaluations. Past performance information is relevant information, for future source selection purposes, regarding a contractor’s actions under previously awarded contracts or orders. The Contract Performance Assessment Reporting System (CPARS) is the official source for past performance information. Government-wide acquisition modernization efforts, led by the DHS Office of the Chief Procurement Officer through the Procurement Innovation Lab, seek to determine the extent to which artificial intelligence (AI) can help contracting officers make more efficient and effective use of CPARS data by rapidly identifying relevant records.
CORMAC is engaged with DHS Procurement Innovation Lab via a multi-phase contract to use AI/NLP solutions for past performance evaluations using CORMAC’s Envisioning and Prediction Enhancing System (CREPES) product. CREPES serves as a next-generation decision support tool, applying the powers of machine learning (ML) with natural language processing (NLP) to streamline the time-consuming process of past performance evaluation for the federal acquisition workforce. CREPES, a SaaS product in AWS, uses best practices such as Agile/Kanban, HCD, TDD, and DevSecOps.
The CREPES product applies ML algorithms, developed based on a human-scored baseline, to data derived from text analytics. It does so using NLP, which compares an active solicitation to offerors’ CPARS records. A meta-analysis is done to combine the statistical results from both factorized data and free-form narrative. Together, this generates a score describing how relevant past projects are to the current requirement, as the government considers success in those contracts to correlate strongly with positive future performance.
Innovations to Bring Business Value
Augmented Analytics: Combining the powers of Machine Learning, NLP, HCD and Data Visualization.
The CREPES product capitalizes on the advancements in ML and NLP to access a whole new world of possibilities. We use various NLP and ML libraries for text analytics and sentimental analysis to automate labor-intensive past performance evaluation. CREPES blends information engineering with textual data using neural networks to summarize and predict a vendor who is likely to succeed when awarded a contract. It utilizes the capabilities of In-memory computing on AWS to scan through thousands of performance records and recommend a vendor based on different criteria, explaining the logic behind its reasoning. CREPES provides an unbiased list of contractors ranked by performance, rather than relying on human intuition to estimate which contractor is most likely to perform well. Agile scrum methodology brings cadence and predictability to the development process. Human-centered design (HCD) techniques are used to create the optimal end-user experience. Additionally, CORMAC uses data visualization techniques to display results more intuitively via vector graphics and reports.