Archive

Tag Archives: news

yahoolabs:

By Suju Rajan

Data is the lifeblood of research in machine learning. However, access to truly large-scale datasets is a privilege that has been traditionally reserved for machine learning researchers and data scientists working at large companies – and out of reach for most academic researchers.

Research scientists at Yahoo Labs have long enjoyed working on large-scale machine learning problems inspired by consumer-facing products. This has enabled us to advance the thinking in areas such as search ranking, computational advertising, information retrieval, and core machine learning. A key aspect of interest to the external research community has been the application of new algorithms and methodologies to production traffic and to large-scale datasets gathered from real products.

Today, we are proud to announce the public release of the largest-ever machine learning dataset to the research community. The dataset stands at a massive ~110B events (13.5TB uncompressed) of anonymized user-news item interaction data, collected by recording the user-news item interactions of about 20M users from February 2015 to May 2015.

The Yahoo News Feed dataset is a collection based on a sample of anonymized user interactions on the news feeds of several Yahoo properties, including the Yahoo homepage, Yahoo News, Yahoo Sports, Yahoo Finance, Yahoo Movies, and Yahoo Real Estate.

image

The news feed on Yahoo’s homepage

Our goals are to promote independent research in the fields of large-scale machine learning and recommender systems, and to help level the playing field between industrial and academic research. The dataset is available as part of the Yahoo Labs Webscope data-sharing program, which is a reference library of scientifically-useful datasets comprising anonymized user data for non-commercial use.

In addition to the interaction data, we are providing categorized demographic information (age range, gender, and generalized geographic data) for a subset of the anonymized users. On the item side, we are releasing the title, summary, and key-phrases of the pertinent news article. The interaction data is timestamped with the relevant local time and also contains partial information about the device on which the user accessed the news feeds, which allows for interesting work in contextual recommendation and temporal data mining.

The Personalization Science team at Yahoo Labs has had a ton of fun working on a full-scale version of the Yahoo News Feed dataset, which has sparked some compelling ideas (e.g. Birds, Apps, and Users: Scalable Factorization Machines and Science Powering Product and Personalization: Going Beyond Clicks) in the areas of behavior modeling, recommender systems, large-scale and distributed machine learning, ranking, online algorithms, content modeling, and time-series mining.

We hope that this data release will similarly inspire our fellow researchers, data scientists, and machine learning enthusiasts in academia, and help validate their models on an extensive, “real-world” dataset. We strongly believe that this dataset can become the benchmark for large-scale machine learning and recommender systems, and we look forward to hearing from the community about their applications of our data.

Happy (large-scale) machine learning in 2016!

Note on our approach to user privacy: Our users place their trust in us each and every day, and we work hard to earn that trust. We zealously protect our users’ privacy, and responsibly and transparently use and protect our users’ personal information. Accordingly, the dataset that we’re releasing as part of this project has been anonymized.

This is a ludicrous amount of data covering billions upon billions of interactions with news stories – fascinating to think what analysis might be made from and ideas sparked by this dataset.

all of this goes back to James Hamilton [a professor of communication at Stanford University]. He gets tremendous props for caring about this. His story of how he came to study this is really interesting. I heard him describe it as, he was in a convenience store, and he saw a newspaper that was basically just made up of people’s mug shots—super weird. And it was one of the only newspapers in this convenience store, and he’s like, “What the hell is this? How is there a market for this and not a market for news? If people are willing to buy this, what are they not being served by traditional media?”

The research that he does is really interesting because he notes that even when low-income news consumers are taking in media at very similar rates to people who have more money, they’re not being served by the media because the media is obsessed with their target audience. I know that to be true. I’m sure you know that to be true. In public radio, there’s this person we consider, called “Mary.” Sometimes, when people are pitching stories, somebody will say, “Well, why would Mary care about that?” And Mary is in her 50s, she’s well-educated, she’s white, she’s affluent. And Mary is not Maria, you know?

It’s not that low-income news consumers are not interested in being served by media, but there are these huge information gaps that result from targeting higher-income consumers. So the stories aimed at them, especially issues in low-income communities, those stories are more like, “Look at what’s happening on the other side of town.” And there’s this very behind-the-museum-glass mentality. If you’re in a low-income community and you see that story, that might be validating if it’s done well. But it’s not informative. It’s not helpful.

Brûlé said Monocle is considering adding new shops or cafes, but its next project, launching this spring, will be a Monocle newsstand near London’s Paddington Station. The kiosk will offer a curated selection of magazines and newspapers as well as global newspapers printed on-demand. Customers will be able to register online, request a printing of newspapers ranging from Norway’s Aftenbladet to The Australian, and then come to the newsstand to pick it up.

Then, why not build a “News Profile”? It could have all the components of my news diet: The publications I subscribed or registered to, the media I visit on a frequent basis, the authors I searched for, my average length of preferred stories, my propensity to read large documented profiles of business persons, the documentaries I watched on You Tube, the decks I downloaded from SlideShare… Why not adding the books I ordered on Amazon and the people I follow on Twitter, etc. All of the above already exists inside my computer, in the form of hundreds, if not thousands, of cookies I collected in my navigations.

The News Lab at Google empowers the creation of media that improves people’s lives. Our mission is to collaborate with journalists and entrepreneurs everywhere to build the future of media with Google. We do this through product partnerships, media trainings, and programs that foster the development of the news industry as a whole.

In case you’re wondering what Steve Grove and Olivia Ma of the News Lab at Google are plotting, here’s a brief description on the registration site for their News Impact Summit Brussels
Follow

Get every new post delivered to your Inbox.