Automatic style spotter understands how a book is written

< its sensing equipment is inside the body and it is equally adept at grasping and moving with any of its four identical arms – each with seven joints. This gives it more options for navigating an obstacle course. Kennedy says they have talked about shimmying up a ladder section like a monkey, without using the steps. “It’s possible because we have this symmetry,” says Kennedy. A team led by Tony Stentz at Carnegie Mellon University is also taking an odd spin on a humanoid robot. Their robot, called CHIMP, has legs that can fold up and double as treads when the robot needs to traverse rough terrain. Like RoboSimian, CHIMP is designed to be statically stable, so isn’t good on the move, but doesn’t need to use complicated algorithms to balance a bipedal gait. “This means we can position ourselves in one place and manipulate things easily,” says Stentz. “We feel good about the tasks that require us to operate tools and manipulate objects with our arms.” The pressure is on. “This is a ridiculously difficult challenge,” says Hong. “There are going to be robots falling down, grey smoke, 30 minutes just standing

“If a robot can do all of these eight tasks, it means that robots can be used for everything” there doing nothing.” Next year, in the second year of the challenge, there will be a marked improvement, he says. “You’re going to see a lot of very impressive things from the teams – just like the autonomous car challenges.” Hong is referring to the DARPA Urban Challenge, which sparked research that has since delivered Google’s self-driving car. Roboticists expect the challenge to improve humanoid robotics, too. “If a robot can do all of these eight tasks, it means that robots can be used for everything.” n 20 | NewScientist | 14 September 2013

Christopher Jackson/Getty Images

TECHNOLOGY

–If you enjoyed reading this…–

Find your next book Software builds a picture of a novel’s style, helping to spot the next bestseller Douglas Heaven

FINISHED a book and want another just like it? A system that provides a quick view of a book’s narrative structure could soon help you fill that Twilight-shaped hole by showing you which other novels match Stephenie Meyer’s style. Joseph Reddington at Royal Holloway, University of London, and colleagues developed a system that compares different bodies of text, looking at the relative frequency of individual words and converting the data into visualisations. Subtle correlations in word use between novels or sections within a novel make it possible to compare style, says Reddington. At its simplest, chapters written in the first person, “I”, will stand out from those written in the third person, and those in the past tense will look different to those in the present. “It helps to get a picture,” says Reddington. “It says this is what your novel looks like.” The tool could be used to monitor consistency in

collaborative writing projects such as film and TV scripts or magazines, where multiple writers must keep to a given style. Reddington has already tested it in a project called TooManyCooks, in which groups of students get together to co-write a novel in a week. Each contributes a section and the tool allows them to tell if any parts do not fit. As the novel develops, the tool displays different sections in a tree-like structure. Those identified as similar appear clustered together, making outliers easy to spot. “At no point is it making a value judgement,” says Reddington. It could also help novice writers see how their style compares with bestselling authors, and give publishers an indication of where a manuscript might fit in the market. The work will be presented in October at a data analysis conference in London. Ultimately, it could form the basis of a recommendation system that makes suggestions based solely on an automatic

assessment of the text. Pandora Internet Radio does this with music: users pick a song or band they like and the system plays others that its Music Genome Project algorithms consider to be similar. “It would be like Pandora for books,” says Reddington. Maintaining consistency is important in collaborative projects, says scriptwriter Brendon Connelly, who is based in Oxford, UK. Producers of the US TV show ER, for example, worked particularly hard to ensure that characters spoke in the same way across seasons even as writers came and went, he says. Software could help flag up inconsistencies. But Connelly thinks such tools could be most useful in tracking the contributions different writers make to a project. The Writers Guild of America, for example, is frequently involved in disputes about who should get writing credits for screenplays. “It could help with those Lennon and McCartney moments where you’re not sure who came up with what line,” he says. n