Tax Law NLP Resources
收藏资源简介:
Resources helpful those doing research using NLP and tax law. This includes: A version of BERT-base-cased trained on 900M tokens from the case.law corpus. A curated corpus of IRS private letter rulings and Tax Court memorandum and summary opinions scraped from the IRS and Tax Court websites and OCR’ed and corrected as necessary. A 500-dimension word2vec vector representation of tax-law terms (including multi-term collocations) trained on this corpus plus tax-specific cases from the case.law corpus.
本数据集专为从事自然语言处理(NLP)与税法领域研究的人员提供研究支撑,具体包含以下内容:1. 基于case.law语料库的9亿个词元(Token)训练得到的BERT-base-cased模型变体;2. 一套精选语料库,涵盖从美国国税局(IRS)与美国税务法院官网爬取的私人信函裁决、税务法院备忘录及简易判决意见书,该语料库经光学字符识别(OCR)处理后按需完成了内容校正;3. 基于上述语料库及case.law语料库中的税法专属案例训练得到的、涵盖税法术语(含多词搭配)的500维word2vec词向量表征。




