Hengzongshu's picture
Update README.md
7142fdf verified
|
raw
history blame
No virus
958 Bytes
---
license: mit
task_categories:
- token-classification
language:
- en
tags:
- math
size_categories:
- 10K<n<100K
---
这是一个基于维基百科制作的轻量训练集,主要用于数学线性代数领域的英语命名实体识别,以及知识图谱构建。
根据维基百科词条对链接进行串联选取,共有234句,无标点符号、数字,不区分字母大小写。
标签标注使用Doccano,有jsonl与txt版本。(jsonl可转换为json)
This is a lightweight training dataset created based on Wikipedia, primarily used for English Named Entity Recognition (NER) in the field of mathematics linear algebra, as well as knowledge graph construction.
Link selections were made based on Wikipedia entries, resulting in 234 sentences. The sentences contain no punctuation, numbers, and are case-insensitive.
Label annotations were done using Doccano, available in both JSONL and TXT formats. (JSONL can be converted to JSON).