Added fraud subset

Added subset of data for filings during periods of fraud

README.md

@@ -42,6 +42,7 @@ The data is sourced from the Electronic Data Gathering and Retrieval (EDGAR) sys
 
 - `clean`: 159B tokens of cleaned text
 - `default`: 111B tokens of cleaned and deduplicated text (referred to as "final" in the paper)
+- `fraud`: 0.3B tokens of text filed during periods of fraud based on SEC [Accounting and Auditing Enforcement Releases](https://www.sec.gov/enforcement-litigation/accounting-auditing-enforcement-releases) and [Litigation Releases](https://www.sec.gov/enforcement-litigation/litigation-releases) (NOT deduplicated)
 - `sample`: 1.1B tokens randomly sampled from `default` stratified by year
 
 ## How can I use this?