ELIS Evidence Classifier Service

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

Not reported

Start date

2020-09-01 (day precision)Source: 2020-09-01T00:00:00

Withheld from public reporting

Not reported

Mission

Topic area

Service Delivery

Operational functions

Document processing and extraction Derived by keyword rules; see methodology

Problem the AI is intended to solve

The adjudicators and contractors spends too much time sifting through digital evidence documents for relevant information.

Expected benefits

To enable end users to navigate directly to the page(s) containing evidence documents of interest instead of sifting through large PDF documents. Evidence tagging intends to accelerate case processing by identifying specific types of documents (e.g., I-589, passport photo spread, marriage certificate) and applying a metadata tag to that document object in ELIS. This way, when a user opens a case with potentially hundreds of pages of evidence documents, rather than scrolling through them one at a time to find a specific document of interest, they have clickable "bookmarks" in the UI generated from these tags that will jump directly to the corresponding page.

System outputs

Tagged evidence. The system inputs an image (scanned document from Lockbox) and outputs either a specific label, such as "Border Crossing Card - Front," or no label if that document is not recognized as one of the classes.

Technology

AI classification

Classical / predictive MLSource: Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relatio…

System name(s)

ELIS

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)

SAIC and DV United

Vendors (standardized)

SAIC

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

Data and privacy

Involves PII

Yes

Privacy Impact Assessment

www.dhs.gov ↗

Authorization to Operate

Yes

Demographic features

Not reported

Training and evaluation data

The system consists of a single vision-based object recognition model, and many text-based binary classifiers. The text models were trained and evaluated on separate class-specific sets of production data sampled from evidence documents, and each data point is the linearized OCR text obtained from a single scanned page image and AWS Textract. These training and testing sets are then annotated by data scientists on our team.

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