similarity

Why we included this project

Similarity is a Java toolkit for judging how close two pieces of text are, from single words up to whole paragraphs, and it runs as plain code with no model server to stand up. Alongside the usual edit distance, cosine, Jaccard, Jaro-Winkler, and SimHash measures, it adds Chinese-specific methods built on the Cilin synonym lexicon and HowNet, plus word-level sentiment tendency scores and approximate-word lookups. Teams doing Chinese NLP in particular will appreciate having those tasks covered inside a single JVM dependency instead of wiring together separate services. The dictionaries ship as plain text and models load lazily, so you can point the library at your own corpus without much setup.