TaxVectors.zip
收藏DataCite Commons2021-02-04 更新2025-04-16 收录
下载链接：
https://archive.data.jhu.edu/file.xhtml?persistentId=doi:10.7281/T1/N1X6I4/SZF0HP
下载链接
链接失效反馈官方服务：
资源简介：
500-dimension vector representation for tax-law terms and collocations (e.g. “tax year”, which is represented as “tax_year”) derived using (Mikolov 2013)’s word2vec implementation using skip-gram with negative sampling; words with a frequency of less than 10 were discarded; 5 iterations through the data; 15 negative samples were used per focus word; words with a unigram probability above 10^-3 were probabilistically discarded; only static windows were used; the training data was all tax-law documents, specifically the curated tax corpus (PLRs and Tax Court unreported decisions) plus tax-specific cases in the Federal case.law corpus.
提供机构：
Johns Hopkins University Data Archive
创建时间：
2020-07-23
