As I write, I note that there are 2,668 comments to my posting yesterday about the Labour party and Jeremy Corbyn. And no, of course I haven’t read them all.
As so often, many commenters go off at a tangent, some debate with each other and several of them return in order to hone their arguments or because they have second thoughts.
Whatever the case, comment threads, especially those which attract many hundreds of participants, can be tough to read and to make sense of.
But help, courtesy of a group of Sheffield university computer scientists, is at hand. They have created a system to transform a large thread into a coherent and concise account of the commenters’ opinions.
By using a computer imitation of the skills traditionally used by newspaper reporters, the commenters are treated as if they are attending a virtual town hall meeting.
The comments are then grouped and summarised so that the writer or a newly arrived reader can quickly gain a sense of the issues raised by commenters. The accompanying video shows how it’s done.
The system, SENSEI, has been developed in collaboration with the Guardian through testing its viability on the comment-is-free section.
It is the creation of an eight-member team led by Rob Gaizauskas and Mark Hepple. “Reader comments are a mix of insight and inanity that few of us have the time or energy to digest,” says Gaizauskas.
“SENSEI has the very practical aim of helping readers to make sense of this mass of comment. We use advanced natural language processing technologies to automatically summarise social media conversations.
“We identify key issues discussed and the range of opinion. The result should be practical tools for on-line news providers... and deeper understanding of the nature of informal argumentative discourse.”
The study of how novel language processing technologies can help make us sense of large volumes of human conversational data has been funded by the European Union and involved four universities in three countries, plus two private sector partners.
