Ipadic model is about 45M. so we created a tiny ipadic in zip.

• Download language model

curl -L -o mecab-ipadic-2.7.0-20070801.tar.gz "https://github.com/lindera-morphology/mecab-ipadic/archive/refs/tags/2.7.0-20070801.tar.gz"
tar xvf mecab-ipadic-2.7.0-20070801.tar.gz

• Remove csv files in folder

• Put files in ipadic/raw into folder

• Edit matrix.def, reset last column(weight) into zero, except first row.

• build

lindera build --dictionary-kind=ipadic mecab-ipadic-2.7.0-20070801 main

• build user dict

lindera build --build-user-dictionary --dictionary-kind=ipadic user_dict/userdict.csv user_dict2

Important: The binary user dictionary format (userdic.bin) is version-specific and needs to be regenerated when upgrading lindera versions.

• Current version: lindera 0.44.0
• Last regenerated: 2025-09-09
• Binary format changes between versions will cause deserialization errors