Courtesy never goes out of style. Which is why researchers at Carnegie Mellon University developed an automated method to up the politeness quotient for various communications. The method, aimed at applications such as chatbots, emails and blogs, involves what is known as politeness transfer. Here's how it might be put to use.
A company working to create a chatbot – a computer program that simulates human conversation and interactions with customers – may use this approach to train its chatbot to respond more politely. So, if the chatbot makes requests or gives instructions that are impolite, or carry a neutral tone, this new method will rearrange and/or add words to the chatbot's responses to improve the tone.
For example, if a chatbot gives the response, "I need more information," the program might teach it to instead say, "Would you please give me more information?"
When it comes to emails and blogs, a businessperson or blogger may use this automated program to check the tone of the email or blog entry before sending or posting.
Language technologists have been training artificial intelligence for years to change negative communications to positive ones. But training for politeness was a distant goal, in part because it's not as easy as it may seem.
"It's not just about using words such as 'please' and 'thank you,'" Shrimai Prabhumoye, one of the program's creators said in a statement. "Sometimes, it means making language a bit less direct, so that instead of saying 'you should do X,' the sentence becomes something like 'let us do X.'"
There was also the problem of defining politeness, which differs across cultures. In this inaugural version of the program, researchers used North American standards of politeness.
Ironically, to create a politeness response database for their method to use, the researchers culled 1.4 million polite sentences from emails exchanged by employees of Enron – the Texas energy company that went bust due to fraud and corruption. Thanks to lawsuits that came out of the Enron scandal, some half-a-million corporate emails became public, and now are being used for various research projects such as this one.