Open news analytics: what you see when the whole stream is labelled

Individual news posts answer the question «what happened». They do not answer the questions that usually matter more: what the news around us is mostly about, who ended up at the centre of events and in what tone they were written about, which channel reported first and which retold it two hours later. To see that, the news stream has to be labelled as a whole, not read post by post.

We have opened that labelling to everyone: news analytics for Russia and for Uzbekistan. The page asks for nothing — no sign-up, no setup — and updates itself as the corpus is collected. At the bottom of each page you can see how many posts it is built on and how much history has been accumulated.

Five views of one stream

Topics — the last 24 hours. The size of a label reflects how often the topic occurs in the corpus; click it to see the posts behind it, with links to the sources.

Main topics of channels — also for the day: which sections form the core of each channel, and how «hot» a section is right now.

Newsmakers — the last week. Who exactly is mentioned, with a type (Government body, Company, Organisation, Person, Product), the number of stories and mentions, a tone bar (negative, neutral, positive) and a chart by day or by week. Open a card and you see the headlines themselves.

Story map — the last 48 hours. One news event assembled from the posts of different channels: you can see who reported first and with what lag the others picked it up. Every story has a «framing breakdown» — how differently channels place the emphasis: Problem, Cause, Judgement, What to do.

Channels — three tabs. «Topics» — what the channel writes about; «Publication speed» — where in a story it lands on average (1 = reported first); «Sourcing» — how often the channel's posts state things flatly, without saying where the claim comes from.

Why the numbers do not match the feeling you get from a feed

Labelling a stream is not the same as counting posts. Three things are done deliberately, and they change the result.

Duplicates of one news event are merged into a story. Otherwise a single event that spread across twelve channels would look like twelve different events, and the most reprinted news of the day would always come first purely by volume.

Round-up posts are not attributed to topics. A table of contents like «The main news of the day» formally belongs to the first item on its list — and would inflate the counter of somebody else's story. Such posts are excluded everywhere a post is credited to a topic, a story or a newsmaker.

The sample is named honestly. Under the channel tables there is a line «N of M channels shown», with the reason the rest did not make it. Weekly trends stay hidden while there is nothing to compare against.

And the main limitation: the labelling is done by AI, which is stated in a banner above all the blocks. The «Sourcing» tab is an assessment based on the text of posts, not an audit of an editorial team; it does not judge political sides.

What it gives the reader

You see a country's agenda in a minute — without scrolling through dozens of channels. For a particular company or person you can look at a whole week: how many mentions, in what tone, in which stories. Channels are compared not by gut feeling but by two measurable properties — who is faster and who is more disciplined about citing sources. That is already enough to decide whom to read yourself and whom to cite in your own work.

The page is available in Russian, English and Uzbek.

The same engine, but on your channels

The analytics is built on public news channels. Exactly the same machinery — collection, labelling of topics and entities, tone, merging of duplicates — can work on your own set of channels: instead of a country-wide view you get a personal feed that delivers only what you described in your own words as important, with analytics on your own corpus next to it.

You can look at the analytics right now: Russia, Uzbekistan. To set up your own channels — start for free.