Trying to analyze Donald Trump made a speech recognition bot literally CRASH – and the only way to make it work properly was to strip out its grammar and syntax tools
- Researchers assigned a speech recognition bot to analyze Donald Trump
- The bot struggled to make sense of his rambling and fragmentary sentences
- The team brought in a specialist to remove the bots grammar and syntax tools
- Since unlearning grammar, the AI bot has become an expert on Trump and is licensed by Amazon to help Alexa answer user questions about the president
The developers of a speech recognition bot assigned to analyze the public statements of politicians hit a major stumbling block when it tried to make sense of Donald Trump.
Built by a tech startup called FactSquared, the bot AI was assigned to go through more than 11 million words Trump has spoken or tweeted since 1976–in interviews, campaign speeches, media appearances, and social media posts.
According to FactSquared’s CEO Bill Frischling, the bot failed to understand Trump’s speeches until he brought in a specialist to strip out all of the bot’s grammar and syntax coding.
The tech startup FactSquared created an AI bot to try and catalog and analyze Donald Trump’s public appearances and interviews, but they were so incoherent and rambling the bot actually crashed. Only after the grammar tools were removed could the bot understand Trump
‘It was still trying to punctuate it like it was English, versus trying to punctuate it like it was Trump,’ Frischling told The LA Times.
One speech had actually caused the bot, named Margaret after a character on The West Wing, to completely crash.
Trump gave the speech on May 4, 2017 to commemorate the 75th anniversary of the Battle of the Coral Sea, which pitted US and Australian troops against the Japanese Imperial Navy.
‘We are privileged to be joined by many amazing veterans from our two countries, as well, and for really, from so many different conflicts,’ Trump said.
‘There are so many different conflicts that we fought on and worked on together. And by the way, in all cases, succeeded on. It’s nice to win.’
Programmer Bill Frischling founded FactSquared to create analyze and catalog the public speeches and appearances of politicians, which can then be used by journalists, researchers, and even businesses like Amazon
Without the obstacle of proper grammar and clear syntax, Trump suddenly started to make a lot more sense to Margaret, and today the bot has become an expert on the president.
FactSquared makes Margaret available as a resource to journalists and historians, and it also licenses the technology to Amazon, who uses it to help Alexa answer user questions about Trump in a more accurate fashion.
Margaret not only catalogs speech transcripts, but can fact check Trump’s claims by independently running queries of materials on public record.
Margaret has also been coded with machine learning tools that allow her to identify Trump’s various emotional and psychological states when he speaks.
According to Margaret, Trump speaks much more quickly when improvising than when reading off a script–around 220 words per minute compared to 111 words per minute.
According to Margaret, Trump displays few signs of physical stress when lying, while for most other people, lying is accompanied by signs of physical tension.
Margaret has found that the surest sign that Trump is genuinely angry is when he stops moving his hands, which he otherwise uses frequently when speaking.
‘When he stops making gestures, that’s the — watch out,” Frischling said. ‘Whatever’s about to happen, hang on to your tush.’
The one overarching insight Margaret has developed from studying Trump’s public statements is that there’s finally very little connection between what he says and does.
‘He can say, “I think this is the worst” about a policy proposal, and then quite literally sign an executive order 10 hours later that approves it,’ Frischling said.