Sadly, this election cycle, with its many dramatic moments, continues to expose the divide in the United States. Ms. Trump’s speech on the first day of the RNC was no exception. People’s reaction to the speech (and the decisions and the actions that led up to the speech and its aftermath) ranged from rational (using math and linguistics to show that the speech was/wasn’t plagiarized), legal (copyright infringement, election guidelines), moral (is plagiarism really a bad thing? ), social (is plagiarism inevitable and *unfortunately* irrelevant in this case? My partner Tanya Gupta thought so). And of course lots of ad hominem (from both the defenders and the critics).
I decided to ask someone who has a great track record of being unbiased and reasonable. This someone is a great chef, a wise medical guide, a fun music lover, a brilliant chess player and a Jeopardy champion. Yes it’s Watson, IBM’s billion dollar super AI.
The IBM Watson™ AlchemyLanguage service is a collection of text analysis functions that derive semantic information from your content. You can input text, HTML, or a public URL and leverage sophisticated natural language processing techniques to get a quick high-level understanding of your content and obtain detailed insights such as sentiment for detected entities and keywords.
I ran Watson’s concept extraction algorithm on the following four convention speeches.
- Laura Bush 2000
- Cindy McCain 2008
- Michelle Obama 2008
- Melania Trump 2016
Watson returned the concepts with a relevance score of (0 to 1) for each concept. I created the following Concept Cloud (using Luca’s JQuery plug-in). The bigger the concepts in the clouds below— the more relevant Watson thinks it is in describing the theme of the document.
Here’s my unscientific conclusion (I am a big fan of Scott Adams)
- Ms. Trump does not have much to say. Watson could only extract four concepts from her speech. Does the AI have a liberal bias?
- All the first ladies are all about their husbands. The husbands are not just better halves — they are the overwhelming dominant whole. A bit too much so for this day and age, no?
- The ladies’ speeches are “ladylike.” They avoid any concrete issues or political discussions. Wonder what Bill Clinton’s speech on Tuesday will be like?
Does this have anything to do with gender bias?
Here’s the raw data from Watson
I enjoyed the process of being a “ghost writer” for Watson. I may write a few more articles of this type. So follow me if you are interested in reading more “Unscience”.