Maybe I missed something but I don't understand this criticism. The author states that 'The number of hits for each year was divided by the number of hits for "

Maybe I missed something but I don't understand this criticism. The author states that 'The number of hits for each year was divided by the number of hits for "machine learning".' Wouldn't that control for the inflation of data in exactly the way you are proposing?