The author gives the word he is trying to explain, but not very well. The author is saying "... in the sense that, by recognising patterns in the data sets, the machine can predict the probability of certain events occurring." In cognitive psychology pattern recognition is important as a component of intelligence and how it is learnt is even more important. Engineers were already designing machines that could learn in a primitive sense with simple servo-mechanisms before computers were even invented. Today, my computer has learnt my speech patterns, recognises my voice and is printing what I am saying now onto the screen. The data set is the raw sound it records. It uses a database of all my previous utterances to calculate the probabilities of what it predicts "I probably want to say". For example it might predict from the patterns recognised in the stored data sets that "say" was more likely than "weigh". Voice recognition software in now quite old. Today when the people talk about AI and pattern recognition, everybody is going on about all our personal data acquired by social media firms and how data in the "cloud" (on the internet) can do amazing or very frightening things. If you ask me its all a load of hype. Hope this helps.

https://www.google.com/search?source=hp&ei=IpksXdSuMo6-aJy8h...