leeroy-jankins/Federal-Trust-Fund-Accounting-Guide

# 📚 Federal Trust Fund Accounting Guide

• Maintainer: Terry Eppler
• Ownership: US Federal Government

---

📋 Overview

Dataset Summary

The Federal Trust Fund Accounting Guide Dataset is a structured natural-language dataset derived from the Federal Trust Fund Accounting Guide, dated June 2001, with selected revised material dated January 2003.

The source guide explains accounting and reporting concepts for Federal trust funds, excluding trust revolving funds. It describes how Federal trust funds are established, how trust fund receipt and expenditure accounts are structured, how trust fund receipts are designated as available or unavailable, and how trust fund activity is reported using Treasury Appropriation Fund Symbols, the United States Standard General Ledger, the SF 133 Report on Budget Execution and Budgetary Resources, the USSGL 2108 Year-End Closing Statement, Program and Financing Schedules, and agency financial statements.

The guide is organized into eight scenarios:

• Scenario I: Basic Trust Fund Accounting
• Scenario II: Trust Fund Balances Precluded from Obligation
• Scenario III: Trust Fund Receipts Available for Investment but not Obligation
• Scenario IV: Trust Fund Investments
• Scenario V: Treasury-Managed Trust Fund Allocation Accounts
• Scenario VI: Trust Fund Appropriation Transfers for Specific Treasury Appropriation Fund Symbols
• Scenario VII: Trust Funds with Contract Authority, Part I: Appropriations to Liquidate Contract Authority Funded by Nonexpenditure Transfers
• Scenario VIII: Trust Funds with Contract Authority, Part II: Transfers of Contract Authority

The dataset is intended to support retrieval-augmented generation, federal accounting question answering, budget-execution classification, transaction-pattern extraction, scenario-based accounting analysis, citation-aware response generation, and domain adaptation for language models working with Federal trust fund accounting and reporting.

Supported Tasks

Retrieval-Augmented Generation

The dataset may be used to build vector indexes, lexical indexes, or hybrid retrieval systems that retrieve relevant passages from the guide in response to questions about Federal trust fund accounting, USSGL accounts, TAFS structure, investment accounting, budgetary reporting, proprietary reporting, and nonexpenditure transfer treatment.

Question Answering

The dataset supports extractive, closed-domain, and open-domain question answering where answers should be grounded in the source guide text.

Example questions include:

• What fund types make up the Federal Government account structure?
• What is the difference between trust funds and Federal funds?
• What does the Federal Trust Fund Accounting Guide cover?
• How are trust fund receipt accounts and expenditure accounts structured?
• What is the difference between receipts available for investment and receipts available for obligation?
• Why do trust funds generally not use USSGL accounts 3100 and 5700?
• What USSGL accounts are used for trust fund transfers?
• How are trust fund investments in Treasury securities recorded?
• How are Treasury-managed trust fund allocation accounts reported?
• How are appropriations to liquidate contract authority handled for trust funds?

Summarization

The dataset can be used to summarize scenarios, transaction narratives, reporting requirements, USSGL account treatments, Federal trust fund structures, and budgetary or proprietary accounting examples.

Text Classification

The dataset can support classification by scenario, fund activity, USSGL account, transaction code, report type, budgetary treatment, proprietary treatment, transfer type, investment treatment, or contract-authority accounting pattern.

Named Entity Recognition and Information Extraction

The dataset may be used to extract entities such as:

• USSGL account numbers and titles;
• Treasury Appropriation Fund Symbols;
• transaction codes;
• Treasury forms;
• OMB forms and reports;
• budgetary reports;
• proprietary financial statements;
• trust fund scenarios;
• Federal agencies;
• accounting references;
• fund types;
• transfer types;
• investment terms.

Dataset Structure

The recommended dataset structure is one record per logical unit, such as scenario, subscenario, transaction, reporting schedule, trial balance, financial-statement section, paragraph, table, or semantic chunk.

For retrieval and training workflows, transaction-level and scenario-level chunks are especially useful because the guide is organized around accounting scenarios and journal entry examples.

Recommended Fields

Field	Type	Description
id	string	Stable unique identifier for the record.
source_document	string	Source document name.
document_date	string	Source document date or revision date.
scenario_number	string	Scenario number, such as Scenario I or Scenario VIII.
scenario_title	string	Scenario title.
section	string	Section heading or subsection heading.
transaction_number	string	Transaction number or closing-entry identifier, if applicable.
transaction_code	string	Transaction code, such as TC A186, TC B107, or TC F204.
accounting_basis	string	Budgetary, proprietary, or both.
ussgl_accounts	list[string]	USSGL account numbers and titles appearing in the record.
forms_and_reports	list[string]	Forms or reports referenced in the record.
fund_type	string	Trust fund, trust fund receipt account, trust fund expenditure account, or related fund type.
tafs	list[string]	Treasury Appropriation Fund Symbols or illustrative symbols.
page	string	Source page reference.
title	string	Best available heading for the record.
text	string	Cleaned source text for the record.
summary	string	Optional short summary of the record.
keywords	list[string]	Extracted or curated keywords.
citations	list[string]	Source citation metadata, such as scenario, page, and transaction reference.
chunk_index	integer	Sequential chunk index within the scenario or section.
token_count	integer	Approximate token count for the record.

Example Record

{
  "id": "federal_trust_fund_accounting_guide_scenario_i_txn_001",
  "source_document": "Federal Trust Fund Accounting Guide",
  "document_date": "June 2001",
  "scenario_number": "Scenario I",
  "scenario_title": "Basic Trust Fund Accounting",
  "section": "Basic Trust Fund Accounting",
  "transaction_number": "1",
  "transaction_code": "TC A186",
  "accounting_basis": "Budgetary and proprietary",
  "ussgl_accounts": [
    "4114 Appropriated Trust or Special Fund Receipts",
    "4450 Unapportioned Authority",
    "1010 Fund Balance With Treasury",
    "5800 Tax Revenue Collected"
  ],
  "forms_and_reports": [
    "Fiscal Service Optional Form 1017-G",
    "SF 133",
    "USSGL 2108"
  ],
  "fund_type": "Federal trust fund",
  "tafs": [
    "20X8000.001",
    "20X8000"
  ],
  "page": "Scenario I, page I-3",
  "title": "Record receipts deposited into a trust fund receipt account",
  "text": "To record receipts deposited into a trust fund receipt account, which are automatically credited to the corresponding trust fund expenditure account...",
  "summary": "Records trust fund receipts as appropriated trust or special fund receipts and tax revenue collected, without using an appropriation warrant.",
  "keywords": [
    "trust fund receipts",
    "appropriated trust fund receipts",
    "fund balance with treasury",
    "tax revenue collected",
    "TAFS",
    "USSGL"
  ],
  "citations": [
    "Federal Trust Fund Accounting Guide, Scenario I, page I-3, transaction 1"
  ],
  "chunk_index": 1,
  "token_count": 145
}

Dataset Creation

Source Data

The dataset is derived from:

Federal Trust Fund Accounting Guide, June 2001, including selected revised material dated January 2003.

The source guide addresses Federal trust fund accounting and reporting using scenario-based examples. It discusses trust fund receipt accounts, trust fund expenditure accounts, available and unavailable receipts, investment activity, transfers, Treasury-managed trust funds, appropriations to and from specific Treasury-managed trust fund TAFS, and trust funds with contract authority.

Processing Pipeline

A recommended processing workflow includes:

1. Extract text from the source PDF.
2. Preserve scenario titles, page labels, transaction numbers, transaction codes, and report headings.
3. Preserve USSGL account numbers, account titles, debit and credit structures, and budgetary/proprietary labels.
4. Preserve references to SF 133, USSGL 2108, Program and Financing Schedules, OMB Form and Content Statements, SF 1151, SF 132, SF 224, OPAC, TAFS, and related Treasury or OMB reporting mechanisms.
5. Normalize page headers, page footers, line breaks, hyphenation, spacing artifacts, and duplicated scenario headings.
6. Split the guide into scenario-level, transaction-level, and report-level records.
7. Attach metadata for scenario number, scenario title, transaction code, USSGL accounts, report type, page reference, and source.
8. Validate that each record can be traced back to the source guide.
9. Optionally generate summaries, keywords, question-answer pairs, accounting-topic labels, or instruction-tuning examples from the extracted records.

Recommended Chunking Strategy

For retrieval use cases, chunk by accounting-document hierarchy where possible:

• scenario;
• year or subscenario;
• transaction;
• budgetary entry;
• proprietary entry;
• closing entry;
• trial balance;
• SF 133 presentation;
• USSGL 2108 presentation;
• Program and Financing Schedule presentation;
• OMB Form and Content Statement presentation.

When sections are long, split into chunks of approximately 500 to 1,000 tokens with 50 to 150 tokens of overlap. Every chunk should retain the scenario number, scenario title, page reference, transaction code, and relevant USSGL account metadata.

For accounting and budget-execution retrieval, preserving transaction boundaries is more important than achieving uniform chunk size. Avoid splitting in the middle of debit and credit entries, report crosswalks, or paired budgetary and proprietary examples.

Data Splits

The dataset may be distributed as a single corpus or divided by scenario or task.

Suggested Split Design

Split	Purpose
corpus	Full guide-derived corpus for retrieval and indexing.
scenarios	Scenario narratives and explanatory accounting guidance.
transactions	Transaction-level records with USSGL entries and transaction codes.
reports	SF 133, USSGL 2108, Program and Financing Schedule, and financial-statement records.
accounts	USSGL account references and account-use examples.
qa	Validated question-answer records derived from the guide, if included by the dataset maintainer.
train	Records used for model training, instruction generation, or embedding index creation.
validation	Records reserved for prompt, retrieval, or model-selection validation.
test	Records reserved for held-out evaluation.

For retrieval-augmented generation, a single corpus split may be sufficient.

splits:
  - name: corpus
    num_examples: 1
  - name: scenarios
    num_examples: 8
  - name: transactions
    num_examples: 0
  - name: reports
    num_examples: 0
  - name: accounts
    num_examples: 0
  - name: qa
    num_examples: 0
  - name: train
    num_examples: 0
  - name: validation
    num_examples: 0
  - name: test
    num_examples: 0

Intended Uses

Primary Intended Uses

This dataset is intended for:

• building retrieval-augmented generation systems for Federal trust fund accounting;
• developing Federal budget-execution and USSGL question-answering systems;
• evaluating citation-grounded answers against source accounting guidance;
• training or fine-tuning models on Federal trust fund accounting terminology;
• extracting structured references to USSGL accounts, TAFS, transaction codes, reporting forms, transfers, investments, and contract-authority concepts;
• supporting research in accounting-domain, budget-domain, and government-document natural-language processing.

Example Use Cases

• A Federal accountant asks how to record receipts deposited into an available trust fund receipt account.
• A budget analyst searches for the SF 133 treatment of receipts available for investment but not obligation.
• A financial manager retrieves guidance on trust fund investments in Treasury securities.
• An analyst compares expenditure transfers and nonexpenditure transfers for trust funds.
• A data scientist builds embeddings for scenario-level and transaction-level retrieval over Federal trust fund accounting guidance.
• A model is evaluated on its ability to classify entries by USSGL account, scenario, transaction type, report type, or budgetary/proprietary accounting treatment.

Out-of-Scope Uses

The dataset should not be used as the sole source for:

• professional accounting advice;
• official financial-statement preparation decisions;
• audit opinions;
• determinations of compliance with current Treasury, OMB, or FASAB requirements;
• replacement of current Treasury Financial Manual guidance;
• replacement of current OMB Circular A-11, OMB Circular A-136, or USSGL guidance;
• agency-specific trust fund accounting determinations without review of the establishing legislation and current appropriation language;
• decisions involving classified, controlled unclassified, proprietary, procurement-sensitive, or personally identifiable information.

The dataset reflects text derived from historical Federal accounting guidance. Users remain responsible for confirming currency, authoritative status, applicability, and later changes before operational use.

Licensing and Use Restrictions

The source document appears to be Federal Government accounting guidance. The dataset maintainer should verify the authoritative source, current publication status, and applicable reuse conditions before redistribution.

Recommended dataset license field:

license: mit

Suggested license note:

This dataset is derived from a Federal Government accounting guide. Dataset maintainers should verify whether all extracted, transformed, enriched, or generated fields are distributable and should document any restrictions associated with derived annotations, summaries, question-answer pairs, or third-party processing.

Data Quality

Strengths

• High-value domain text for Federal trust fund accounting and budget-execution NLP.
• Strong scenario-based structure with explicit transaction examples.
• Rich USSGL account references, transaction codes, report crosswalks, and financial statement presentations.
• Useful for citation-grounded retrieval and accounting guidance question answering.
• Strong fit for extracting structured accounting patterns, TAFS references, transfer treatments, investment accounting, and contract-authority accounting examples.

Known Limitations

• The guide is dated June 2001, with selected material revised in January 2003.
• Treasury, OMB, USSGL, and FASAB guidance may have changed after publication.
• PDF extraction may introduce page-header, footer, table, line-break, hyphenation, or spacing artifacts.
• Some tabular debit and credit entries may require manual or table-aware validation.
• Scenario examples are illustrative and not an exhaustive list of all Federal trust fund transactions.
• Some accounting determinations depend on agency-specific establishing legislation, appropriation language, OMB guidance, Treasury guidance, and current USSGL crosswalks.
• Generated summaries or question-answer pairs, if included, should be validated against the source guide.

Bias, Risks, and Limitations

This dataset represents Federal trust fund accounting guidance and reflects the Treasury, OMB, USSGL, and FASAB reporting environment at the time of the guide.

Potential risks include:

• using historical guidance when later Treasury, OMB, USSGL, or FASAB updates apply;
• treating illustrative scenarios as exhaustive accounting rules;
• applying trust fund guidance to special funds, general funds, revolving funds, deposit funds, or trust revolving funds without verifying applicability;
• treating model output as professional accounting advice or audit evidence;
• failing to preserve debit and credit structure during chunking;
• missing distinctions between budgetary and proprietary accounting entries;
• missing distinctions between availability for investment and availability for obligation.

Models trained or evaluated on this dataset should be designed to cite source passages, identify uncertainty, preserve transaction context, distinguish budgetary from proprietary accounting, and defer to current authoritative sources and qualified professionals for final accounting determinations.

Personally Identifiable Information

The source guide is a public accounting guidance document and is not expected to contain personal records, private individual-level data, or personally identifiable information. Dataset maintainers should still review extracted text and metadata before publication to confirm that no unintended sensitive information was introduced during processing.

Security Considerations

The dataset is derived from Federal accounting guidance. It should not be combined with classified, controlled unclassified, procurement-sensitive, proprietary, or personally identifiable data unless the resulting dataset is governed under appropriate controls.

Maintenance

Dataset Maintainer

The maintainer is not specified in the source document.

Update Frequency

The dataset should be refreshed when current authoritative Treasury, OMB, USSGL, or FASAB guidance supersedes or materially changes the accounting treatments described in the guide.

Because individual trust fund accounting treatments may depend on current law, current USSGL accounts, current OMB reporting requirements, and current Treasury Financial Manual guidance, maintainers should track currency at the scenario, account, transaction, and report-crosswalk level.

Versioning Recommendation

Use semantic versioning tied to both dataset-processing changes and source-guidance updates.

Example version history:

v1.0.0 - Initial dataset release based on the Federal Trust Fund Accounting Guide.
v1.1.0 - Added scenario-level summaries, keywords, and report-type metadata.
v1.2.0 - Added transaction-level records and USSGL account metadata.
v1.3.0 - Added validated question-answer pairs and accounting-topic labels.
v2.0.0 - Refreshed source text after authoritative guidance update.

Citation

When using this dataset, cite the source guide and the dataset release.

Dataset citation:

@misc{federal_trust_fund_accounting_guide_dataset,
  title        = {Federal Trust Fund Accounting Guide Dataset},
  author       = {{Dataset Maintainer Not Specified}},
  year         = {2026},
  howpublished = {Hugging Face Dataset},
  note         = {Derived from the Federal Trust Fund Accounting Guide}
}

Recommended source citation:

@misc{federal_trust_fund_accounting_guide,
  title        = {Federal Trust Fund Accounting Guide},
  author       = {{United States Government}},
  year         = {2001},
  month        = {June},
  note         = {Includes selected revised material dated January 2003}
}

Dataset Card Authors

The dataset card was prepared for a document-derived Hugging Face dataset based on the Federal Trust Fund Accounting Guide.

Acknowledgements

This dataset is based on the Federal Trust Fund Accounting Guide, which provides scenario-based Federal trust fund accounting and reporting guidance using USSGL, TAFS, SF 133, USSGL 2108, Program and Financing Schedule, and agency financial-statement examples.