Employee Retention Credit Text Clustering

Overview

Development stage

DeployedSource: c) Deployed – The use case is being actively authorized or utilized to support the functions or mission of an agency.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

The output is not presumed to be high-impact and is not used as the principal basis for significant decisions/actions

Start date

2024-01-31 (day precision)Source: 1/31/2024

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Administrative Functions

Operational functions

No function tag matched Derived by keyword rules; see methodology

Problem the AI is intended to solve

Application of Line-Item Consolidation algorithm to identify highly similar text and apply additional features to similar clusters to enable identification of suspicious texts.

Expected benefits

This enables the agents to increase their capacity and reduce time researching and identifying potential suspicious trends of topics used in claims.

System outputs

Detailed explanations from taxpayers as to why they are claiming the Employee Retention Credit (ERC) are grouped into semantically similar topics to create clusters. The ERC is a refundable tax credit for certain eligible businesses and tax-exempt organizations that had employees and were affected during the COVID-19 pandemic. The model produces a table of cleaned tax disclosures and their assigned cluster (numerical label).

Technology

AI classification

Natural language processingSource: Natural Language Processing: AI that processes, interprets, and shares information in human language.

System name(s)

Not reported

Custom-developed code

Yes

Public source code

Not reported

Sourcing

How it was built

Contract and in-houseSource: c) Developed with both contracting and in-house resources

Vendor (as reported)

Not reported

Vendors (standardized)

None on the standard list

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

No

Privacy Impact Assessment

Not reported

Authorization to Operate

Yes

Demographic features

k) None of the above

Training and evaluation data

We the effectiveness in the algorithm in supporting the identification of potentially risky clusters. We did this in two ways, the first was to identify clusters that had a high proportion of additional risk flags, since similar text indicates that text is potentially coming from a single source group, the rest of the cluster is also a target for review. The other method was to identify groups with a large number of Preparer EINs. This indicates that the preparer ID is missing, but they may have a shared source. One critical aspect in the evaluation process involved setting a minimum similarity threshold for the clustering process. We also conducted spot checks on a random sample of groups to ensure consistency, where we found that the text within each group was predominantly consistent, highlighting the algorithms’ ability to group semantically similar descriptions.

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported