Unfinished version: Faulty bot responses Left with several problems. training reached at most low 90% training accuracy word lemmatization and phrase identification only partially working
