“Sports scores, quarterly earnings, and the vicissitudes of a financial portfolio are easy enough to arrange into tables of numbers, says Yejin Choi, an expert in natural language processing at the University of Washington. Opinions, however, are bit more challenging to quantify. In addition, software programs have a hard time imitating the rhetoric of a serious op-ed writer.
“‘One is the content challenge; the other is the style challenge,’ she says.
“On the content side, many AI researchers are now looking into how a computer might recognize when an argument supports or opposes a particular position. Likewise, computer scientists have been working on qualitative reasoning or understanding the relationship between two variables (for example, ‘If I throw the ball harder, it will travel further’).”
Read more here.