New Automated Method Makes Sending Requests More Polite

Being polite to one another facilitate communication during an interaction, as it minimizes the potential for conflict and confrontation. It is something that people need these trying times as the pandemic continues to rage, and politicians wrangle for votes and protesters demand justice. A little bit of politeness can surely go a long way.

Researchers from the Carnegie Mellon University have developed an automated method for making interactions more polite, Science Daily reports. The researchers are set to present their study on politeness transfer at the Association for Computational Linguistics annual meeting, which will be held virtually on July 5.

From Impolite Directives to Polite and Well-Mannered Requests

The new method takes impolite or neutral directives or requests and restructures them by adding words to make them more well-mannered. Such as the phrase: "Send me the data," might turn into "could you please send me the data?"

The idea of polite transfer method is something that technologists have been doing for some time. Performing politeness transfer has long been the goal of Shrimai Prabhumoye, a Ph.D. student in CMU's Language Technologies Institute (LTI).

"It is extremely relevant for some applications, such as if you want to make your emails or chatbot sound more polite or if you're writing a blog," Prabhumoye said. "But we could never find the right data to perform this task."

Together with LTI master's students Aman Madaan, Amrith Setlur and Tanmay Parekh, they found a solution for that problem by generating a dataset of 1.39 million sentences labelled for politeness that they used for their experiments.

These sentences were derived from emails exchanged by employees of Enron, a Texas-based energy company that was better known for corporate fraud and corruption before its demise in 2001. But because of the lawsuits surrounding its fraud scandal, over half a million emails became public which are then used as a dataset for a variety of research projects.

Researchers find it challenging to define politeness even with a dataset. Prabhumoye said that it is not just the words "please" and "thank you," sometimes it means making language a bit less direct.

Moreover, politeness varies from one culture to another, such as for native North Americans who commonly use "please" in requests to close friends. But in Arab culture, saying the word is awkward or sometimes rude. So the CMU researchers decided to restrict their work to speakers of North American English in a formal setting.

It's More Than Just "Thank You" and "Please"

The researchers analyzed the politeness dataset they gathered and determined the frequency of words in the polite and nonpolite statements and then developed a "tag and generate" pipeline to do the politeness transfers.

It works by tagging first the impolite words then a text generator replaces it into well-mannered statements. The system is careful not to change the meaning of the sentence.

Over time, the scoring system became more realistic, and changes became more subtle. Like the changing the first person singular pronounce, such as I, me and mine, to first-person plural pronouns, such as we, us and our.

Furthermore, rather than put "please" at the beginning of the sentence, the system learned to insert it within the sentence. An example of that is this" "Could you please send me the file?"