Data Offerings · Court & Judicial Records

U.S. Court Opinion Datasets: 8.5 Million Full-Text Judicial Records

Nearly four centuries of American case law, cleaned and ready to import. Every record pairs the complete opinion text with its case name, filing date, and citations. Sold by Macon Apps on AWS Data Exchange as a single file or as 38 decade files.

8,507,968court opinions

1658 – 2026years covered

~86 Bcharacters of opinion text

CSVUTF-8, RFC-4180

Two ways to buy the same court records corpus

Both products contain the identical 8.5 million judicial opinion records with the same five-column schema. Choose the packaging that fits your pipeline: one large file for a single bulk load, or decade-sized files for incremental and parallel processing.

Single file · bulk load

U.S. Court Opinions Full-Text Dataset (8.5M+ Records)

8.5 million full-text U.S. court opinions from 1658 to 2026 Q1 in one ready-to-import CSV. Case names, filing dates, citations, and roughly 86 billion characters of cleaned opinion text, one record per row.

Spans 38 decades of American jurisprudence, earliest opinion filed in 1658.
Opinion texts average about 10,000 characters (roughly 1,600 words) and range up to 3.4 million characters.
Empty and whitespace-only records excluded; every row has non-empty opinion text.
Embedded newlines properly quoted, nulls distinguishable from empty strings, header row included.

Delivery: AWS Data Exchange, S3 object
Format: 1 CSV, ~81.5 GB, UTF-8
Revisions: All historical and future revisions
Personal data: None (per listing)

$250per month, 1-month contract

$2,500per year, 12-month contract save 17%

View on AWS Marketplace Listing prodview-bi7eqpz3xm7zq

38 decade files · incremental

U.S. Court Opinions by Decade, 8.5M Records in 38 Files (1658 to 2026 Q1)

The same 8.5 million full-text opinions split into 38 decade-by-decade CSV files, from the 1650s onward. Load the 1990s for a focused study, sample across eras, or stream all 38 files for the complete corpus.

Each file holds the records whose filing date falls within that decade (for example cases_1990s_no_empty_cases.csv).
Identical schema in every file, so one import routine handles the entire collection.
Decade-sized files are far easier to load, sample, and parallelize than a single 81 GB download.
Empty and whitespace-only records excluded; UTF-8, RFC-4180, header rows.

Delivery: AWS Data Exchange, S3 objects
Format: 38 CSV files, UTF-8
Revisions: All historical and future revisions
Personal data: None (per listing)

$250per month, 1-month contract

$2,500per year, 12-month contract save 17%

View on AWS Marketplace Listing prodview-tya72fxmn44jo

What's in every record

Five columns, the same in both products and in all 38 decade files. Exported directly from PostgreSQL for reliable import into any database, data warehouse, or ML pipeline.

Column	Description	Typical use
`case_name`	The name of the case	Display, search, linking to other legal datasets
`date_filed`	Filing date of the opinion	Time-series analysis, era sampling, decade partitioning
`citations`	The case's citations as reported	Reporter lookup, citation network construction
`citations_norm`	Normalized citation string	Matching, joining, and deduplication across sources
`cleaned_text`	Full narrative opinion text, cleaned and whitespace-normalized, original paragraph structure preserved	Legal NLP, embeddings, language model training, full-text search

Which product should I choose?

	Full-Text Dataset (single file)	By Decade (38 files)
Records	8,507,968	8,507,968 (same records)
Files	1 CSV, ~81.5 GB	38 CSVs, one per decade
Best for	One-time bulk load into a warehouse or database	Studying a single era, sampling by period, parallel or incremental ingestion
Schema	Identical: `case_name`, `cleaned_text`, `citations`, `date_filed`, `citations_norm`
Price	$250 per 1-month contract, or $2,500 per 12-month contract
Listing	AWS Marketplace	AWS Marketplace

Built for legal research, legal NLP, and AI

A complete, cleaned corpus of judicial opinions is the foundation for anything that needs to read, search, or learn from American case law.

Language Model Training

Roughly 86 billion characters of cleaned legal English for pre-training, fine-tuning, or building domain-specific legal LLMs and embedding models.

Citation Network Analysis

Reported and normalized citation strings on every record make it practical to build citation graphs, trace precedent, and measure influence across centuries.

Legal Research & Search Tools

Power full-text and semantic search, retrieval-augmented generation, and case-finding products with a corpus that already includes the full opinion text.

Empirical Legal Studies

Filing dates on every opinion and decade-partitioned files make era-by-era analysis, historical trend studies, and academic sampling straightforward.

Legal NLP & Classification

Train and evaluate models for issue spotting, outcome prediction, summarization, named-entity recognition, and other legal text tasks.

Data Products & Warehousing

Clean, consistent CSV straight from PostgreSQL loads into Redshift, Snowflake, BigQuery, Postgres, or DuckDB with one import routine.

How delivery works on AWS Data Exchange

AWS Data Exchange delivers the data straight into your AWS account. There is nothing to download over the public internet and no separate vendor account to create.

Subscribe on AWS Marketplace

Choose a 1-month or 12-month contract. The charge appears on your regular AWS bill.

Accept the entitlement

The data set appears in AWS Data Exchange in your account, with all historical and future revisions included.

Export to your S3 bucket

Export the CSV files to any bucket you own, in the same region, with a click or an API call.

Load and build

Import with your usual tooling. Standard AWS infrastructure charges for storage and transfer apply.

Frequently asked questions

What courts and years do these judicial records cover?

Both datasets contain 8,507,968 full-text U.S. court opinions. The earliest opinion was filed in 1658 and coverage runs through the first quarter of 2026, spanning 38 decades of American case law.

What is the difference between the single-file and by-decade datasets?

They contain the same records with the same five-column schema. The Full-Text Dataset ships as one CSV of about 81.5 GB. The By-Decade Dataset splits the same records into 38 CSV files, one per decade, so you can load a single era or process the corpus file by file in parallel.

What format is the court opinion data delivered in?

UTF-8, RFC-4180 CSV with a header row, exported directly from PostgreSQL. Embedded newlines are properly quoted, null values are distinguishable from empty strings, and every record has non-empty opinion text.

Does the dataset contain personal data?

The AWS Data Exchange listing classifies the data sets as containing no personal data. The records are published court opinions with case names, filing dates, citations, and opinion text.

How much does the court opinions dataset cost?

Each dataset is $250 for a 1-month contract or $2,500 for a 12-month contract, a saving of about 17%. Purchases are made through AWS Marketplace and billed to your AWS account. Additional AWS infrastructure costs, such as S3 storage, may apply.

What is the refund policy?

Because the product is downloadable data that cannot be returned once accessed, approved refunds are limited to 50% of the purchase price. Requests must be sent to [email protected] within 14 days of purchase and include your AWS account ID and reason. If the dataset is materially defective (corrupted, inaccessible, or materially different from the listing), we will remedy it or issue a full refund.

Are data dictionaries or samples included?

The listings do not include a separate data dictionary or sample file. The schema is fully described on this page, and you can email us with questions about the data before you subscribe.

Questions about the data?

Ask about coverage, schema, licensing, or how the corpus fits your use case before you subscribe. Reach us at [email protected] or 937-217-7327.

Full-Text Dataset By-Decade Dataset

Use of each dataset is subject to the End User License Agreement shown on its AWS Marketplace listing. AWS, AWS Marketplace, and AWS Data Exchange are trademarks of Amazon.com, Inc. or its affiliates.