On some arguable claims in B. Shneiderman’s evaluation of natural language interaction with database systems