language:
- en
size_categories:
- 1K<n<10K
SaulLM-7B: Pioneering the first Legal Large Language Model
Perplexity Analysis
This dataset presents the data used in the paper "SaulLM-7B: Pioneering the first Legal Large Language Model" in "6.3 Perplexity Analysis" section. The dataset contains the perplexity scores of SaulLM-7B, Llama2-7B and Mistral-7B across a corpora of recent text.
Cleaning
We proceeded to standardize the data by removing any special characters using unicodedata normalization. We also filtered for text that have proper punctuation and a minimum length of 20 tokens. Additionally, we also filtered the text by using a Naive Bayes classifier that aims to classifiy if a given text is a consice piece of text or not.
Data
The corpora consists of four distinct categories, each serving different purposes:
Party Submissions: This category includes documents submitted by parties involved in legal proceedings. These submissions were sourced exclusively from Texas briefs filed after October 1, 2023.
Legal Decisions: Documents in this category are legal judgments or rulings. They were obtained from the International Centre for Settlement of Investment Disputes (ICSID), specifically filtering for decisions made after October 2023.
Contracts: This category comprises contractual agreements. The contracts were sourced from the Electronic Data Gathering, Analysis, and Retrieval (EDGAR) database. Specifically, data from 2024 Q1 was focused on to ensure the exclusion of any data previously seen by the model.
Legislation: This category encompasses bills introduced in the United States Congress. It includes bills submitted to either the U.S. House of Representatives or the U.S. Senate after October 1, 2023.