Harvard Inventors Benchmark

The manual Harvard disambiguation used as one of the benchmarks in the paper. In all columns, if more than one ID is found, the elements are comma separated. 6 columns, 587 rows, "|" delimited; the columns are:--Pub\_number: the publication number of the patent--ManualIDs: The manually disambiguated IDs of each inventor on the patent. --OurIDs: the output of our algorithm on this patent. --LiLow: The "low" disambiguation of Li et. al. on this patent. --LiHigh: the "high" disambiguation of Li et. al. on this patent. --RawNames: the undisambiguated names on the patent, with case and punctuation dropped. <br>

本数据集采用手动哈佛消歧方案，作为论文中的基准测试集之一。所有列中，若存在多个ID，元素间以逗号分隔。该数据集共包含6列、587行，以竖线（|）作为分隔符，各列说明如下：
--Pub_number：专利的公开编号
--ManualIDs：该专利上每位发明人的手动消歧ID
--OurIDs：本研究算法针对该专利的输出结果
--LiLow：Li等人针对该专利的「低阈值」消歧结果
--LiHigh：Li等人针对该专利的「高阈值」消歧结果
--RawNames：该专利上未经过消歧的发明人姓名，已去除大小写与标点符号。