文本相似度
使用 TF-IDF 餘弦相似度、Jaccard 相似度和編輯距離衡量兩個文本的相似程度，全部在本地計算。
Enter text in both fields to compare them.
Token rule: if a text is mostly CJK characters it is split into character bigrams; otherwise it is split into lowercase words.
使用方法
- Enter the two texts you want to compare.
- Three metrics are computed automatically: TF-IDF cosine, Jaccard and Levenshtein similarity.
- Read the explanation on each card — they measure different things, so disagreement between them is normal.
注意事項
All three metrics measure the surface (words and characters), not meaning, and everything is computed in your browser. Tokenization uses character bigrams when a text is mostly CJK, otherwise lowercase words. Levenshtein uses at most 5,000 characters from each text. Paraphrased texts can score low here; for meaning-level similarity use an embedding approach such as the Semantic Search tool.
常見問題
- Can I use this to check for plagiarism?
- It shows surface overlap, but it cannot detect paraphrased or translated copying. Use it as an indicator only.