A computer model was programmed to call M. cedros “imperile,” or weak. When the model was operated again, it produced the word “evaporator,” a word with another meaning in the word “lymbidium.”
“Of course we called it ‘evaporator,’ ” said Fabiano Racca, the paleontologist who built the model. “We didn’t think it was something that didn’t exist.”
Dr. Racca and his colleagues from the University of Trieste analyzed one of the 20,000 words and symbols that were placed into a lexicon of new software algorithms that scientists use to parse, interpret and synthesize scientific language. The study, published Thursday in the journal PLOS One, is part of a growing field of research called “semantic exploration.”
“What we are able to do is to look at words, understand the meaning and identify related words and similar phrases,” said Jonathan Kim, a professor at MIT, who studies the workings of algorithms. “Because we can do this, if we want to understand what algorithms are doing, we can simply ask people what it is.”
Dr. Racca said it was important to have some historical documentation of the DNA of words. Sometimes artificial intelligence programs generate words that make no sense.
“If a word is never used on the web it is not likely to be used in biology,” Dr. Racca said.