Unlike the Western languages supported by iKnow where the basic sentence structure is Subject-Verb-Object, the basic structure of the Japanese language is Subject-Object-Verb. Moreover, there is no word boundaries in Japanese text. To overcome such differences, iKnow Japanese uses a different indexing approach. We identify Entities, which are equivalent to Concepts in Western languages, based on the grammatical information (but not always Relations) within the sentence, and an algorithm we call Entity Vector determines the Path.

Here are a couple of White Papers that discuss how we do this a bit more in detail. (NOTE - they are available in Japanese only.)

https://www.intersystems.com/assets/sites/6/iKnow_JapaneseVer1_4P.pdf

https://www.intersystems.com/assets/sites/6/iKnow_Japanese_Summary_ver1_6P.pdf