# Grammatical Agreement with Unicode Inflection AI high computational cost and network latency top tier languages are covered but torso and tail lack data bias based language ##### Where AI can help - Offline processing where latency or resources are not critical, like grammar fixing, lexicon generation/expansion - smaller, older, less costly models could be used for higher coverage or better accuracy, like LSTMs - Coverage for language we don't have experts to generate rules or with high grammar complexity - Client side support is **slowly** improving with nano models ##### **Concept of Lemmaless inflection**