Text Analytics Toolbox does not ship the tooling to compile an extended MeCab dictionary. But if you have one for your field (I know there are such compiled dictionaries for medical purposes, for example), you can use mecabOptions to have tokenizedDocument use it.
Alternatively, if you only have a handful of words you want to preserve, and are not worried about inflections, you can use "CustomTokens" to pass them to the tokenizer:
5 tokens: 日本 睡眠 学会 の ガイドライン
3 tokens: 日本睡眠学会 の ガイドライン