文本相似度
UpBox优盒工具箱，免费在线文本相似度对比：粘贴两段文字，按相邻字组算出片段重合与字符重合两个百分比，快速判断改写稿件是否还留着原文骨架。
无需注册 · 数据不上传No sign-up · nothing uploaded
两段文字Two texts
已比对
相似度按相邻两字的重合片段计算（Dice 系数），对中文改写比较敏感、对换语序不敏感；它衡量字面重合，不判断意思是否相同，也不能当查重报告用
重合程度Overlap
—
—
片段重合（Dice）—
字符重合（Jaccard）—
有效字数 A / B—
Dice 越高说明两段文字用了越多相同的相邻字组；字符重合率会把「的」「了」这类高频字一起算进去，通常比片段重合更高，看趋势不看绝对值
本工具所有运算在本地浏览器完成，你的文件、文本不会上传到服务器
工具介绍
改完稿子总觉得还像原文，却说不清哪儿像；逐句比对两三百字的段落，眼睛先花。 A rewrite can still feel like the original without telling you where, and line-by-line comparison of a few hundred characters wears your eyes out.
把两段粘进来，按相邻两字的重合片段算个百分比，重复多少一眼就知道，改写前后各跑一次最直观。 Paste both texts and the percentage comes from shared adjacent character pairs, so one pass before and one after a rewrite shows the difference plainly.
- 两种指标Two metrics
- 忽略选项Ignore options
- 分档结论Graded verdict
- 即时计算Live result
使用说明
- 粘贴原文。Paste the original.
- 粘贴改写稿。Paste the rewrite.
- 按需勾去空白标点。Toggle whitespace and punctuation.
- 看百分比与结论。Read the percentage.
常见问题
相似度是怎么算的？How is the score computed?
把两段文字切成相邻两字的片段，统计共同片段占全部片段的比例，即 Dice 系数。Both texts are cut into adjacent character pairs, and the shared share of all pairs gives the Dice coefficient.
为什么换了语序相似度还是很高？Why does reordering barely lower the score?
相邻片段只看字面组合，句子顺序调换后字组大体不变；这是本页的已知局限。Adjacent pairs look only at local combinations, which survive reordered sentences; that is a known limit here.
能当论文查重结果吗？Is this a plagiarism check?
不能，真正的查重比对的是公开文献库，本页只在两段文本之间做字面比较。No, real checks compare against published corpora, while this compares only the two texts you paste.
多长的文本比较准？What length works best?
几十字到几千字都可用；短到十几个字时百分比会被少量片段放大，参考价值有限。Anything from dozens to thousands of characters works, though a dozen characters make percentages jumpy and less meaningful.