Software Scans for Spin in Speeches

Humans may be susceptible to methods of persuasion that play on the emotions and circumvent logic, but computers are another story. Enter a software program that purports to detect “spin” in politicians’ speeches by using a complex (albeit man-made) algorithm to hunt for truth-stretching words and phrases.

New Scientist:

Technology is here to help. Software programs that analyse a person’s speech, voice or facial expressions are building upon the work of researchers like Ekman to help us discover when the truth is being stretched, and even by how much. “The important thing to recognise is that politicians aren’t typically good at out-and-out lies, but they are very adept at dancing around the truth,” says David Skillicorn, a mathematics and computer science researcher at Queen’s University in Kingston, Ontario, Canada. “The 2008 election has so far given us plenty of chances to see them in action.”

Skillicorn has been watching out for verbal “spin”. He has developed an algorithm that evaluates word usage within the text of a conversation or speech to determine when a person “presents themselves or their content in a way that does not necessarily reflect what they know to be true”.

The algorithm counts usage of first person nouns – “I” tends to indicate less spin than “we”, for example. It also searches out phrases that offer qualifications or clarifications of more general statements, since speeches that contain few such amendments tend to be high on spin. Finally, increased rates of action verbs such as “go” and “going”, and negatively charged words, such as “hate” and “enemy”, also indicate greater levels of spin. Skillicorn had his software tackle a database of 150 speeches from politicians involved in the 2008 US election race (see diagram).

Read more