News360 today launches a redesigned version of its semantically-enabled news discovery application for the iPad. In the works for some eight months, the company’s goal has been to further its vision of building an AI assistant that understands what someone is interested in and how he consumes different types of news, to make ever-more solid personalized recommendations of content.

Personalization capabilities in the update last August of the service (which we first discussed here) followed an approach that worked best for people who knew what they wanted, according to Roman Karachinsky, CEO, News360. “Most people don’t want to spend time switching between categories and configuring things,” he says. “They wanted something more simple, so we focused on that.”

Before personalizing things further with Facebook and Twitter analysis of profiles, likes, demographics and social graph activity for more suggestions, there now are a host of sections from which users can easily select their interests – more than 1 million different things they can follow, from companies to topics to people to brands. There’s a new Good News section for getting your optimism on, and one on zombies, too.

“More or less there is an infinite list of things you are interested in, and instead of giving you different categories or switching, a single section integrates them all, which is the Home category,” says Karachinksy. “It takes all your different interests and combines into a single feed.” Cube rotation provides different actions and perspectives on pieces, showing the same story from multiple sources, along with relevant photos, videos and so on in a single, comprehensive view. For a story about the president, for instance, that’s covered by more than 950 sources, News360 will try to choose the most relevant coverage and interesting updates.

Interests meshed together in a single feed means that users can have political stories, for example, brought right into their general news stream. “That takes the pain out so you don’t have to switch context all the time,” he says.

News360 also steps in now to give users a hint when perhaps they’ve read enough, to keep from having its scrollable list of stories become frustrating because it seems to go on forever. Once the service feels a user has read enough, it awards him a star; up to three can be earned each day. “We’ll see if this translates to a larger audience but for our testers it mitigates anxiety about having too much news to read,” he says. “It’s not about making this a game, but about giving a sense of here’s where it is ok to stop.”

More stars help with further personalizing feeds, as well, and eventually they may be used to create a history around a person’s News360 use. For instance, from one month’s use users might be able to get an overview of how they consume information, what they missed, and here’s what they should focus on more, such as reading more general-purpose pieces to be a better-informed citizen.

Karachinsky says the work on the front-end has been combined with quite a lot of work on the back-end as well. “Sixty or 70 percent of this release comes from the semantic analysis side. This is the first version with adaptive behavior to gather information on what you read and interact with – via Facebook, Twitter and inside News360 – and adapt as we go along,” he says, versus the once-and-done approach. Now the service actually tries to adapt and redraw users’ portraits every half hour or so to find things it thinks will be more relevant to them. For instance, if it sees you are interested in technology it will try to understand if that interest is mostly about startups vs. some other connection. And it tries to also understand location and time parameters – what users like to read, where and when – to make presentation around personalization better.

He also notes that there have been a lot of evolutionary changes in the service’s semantic analysis platform, including scaling up the amount of different entities it tracks. It’s somewhere at the 1.2 to 1.3 million entity-mark in its ontology right now, he says. The Good News topic from the semantic analysis standpoint is interesting, Karachinksy notes, ultimately requiring a huge list of rules to cull appropriate stories. “We used a combination of linguistic analysis where we tried to identify basically sense and word combinations that signify something positive. That’s hard from the very beginning to even specify what we want to see in Good News, because it’s not just an absence of negativity, but also things like stories about animals being rescued,” he says. Also in the mix is social analysis of which stories are shared a lot – these tend to be positive, but there are appropriate alarm bells to avoid those that are scandalous or unsavory, which also tend to be shared a lot.

There have been more than 1.5 million downloads of the application on different devices so far, he says. The application remains free, and users can download the latest version from Apple’s App Store starting today.