Much of Tuesday’s news cycle was dominated by stories about how Melania Trump’s GOP convention speech was far too similar to one given by Michelle Obama in 2008.
But a cognitive computing company with an office in Rockville is floating an alternative hypothesis: no, it wasn’t.
Well, not entirely, at least.
Expert System, which has developed software designed to not only to recognize words but to understand meaning and context, has used that technology to analyze transcripts of the two speeches. The result? Aside from the controversial passage (i.e., the apparent agiarism-play), the speeches were quite different in their main concepts, word choices and target audiences.
Trump’s speech seemed to emphasize her husband’s role as a single leader, while Obama’s speech portrayed her husband’s role as one of service to the American people, according to the report.
Trump also mentioned her husband 35 percent more often often than Obama referred to hers.
Obama’s speech also seemed to target a more highly-educated audience, with an average word-per-sentence count of 28.08 to Trump’s 16.33 and use of “more articulated sentences.” But Trump had a higher “readability index,” indicating that her speech was generally easier to understand, according to the report.
Verb usage was largely similar, but Obama used “hope” — a central part of her husband’s campaign message — fairly often while Trump didn’t use it once. Trump’s most commonly used verb was “can,” while Obama’s was “work.”
“Hope” was also determine to be the most relevant emotion in Obama’s speech, followed by love and success; “desire” was the primary emotion in Trump’s speech, followed also by love and success.
Daniel Mayer, CEO of Expert System Enterprise, said in a statement that the analysis showed the power of data science to test assumptions — such as the idea that the speeches were largely similar — and prove them wrong.
Of course, when the specific controversial passages were analyzed, “there was indeed a 95% correlation between these parts, making the two, by data analysis standards, nearly exactly the same,” the company wrote in a statement.