How technology is helping provide 'personalised' news feed, content: Dailyhunt's Umang Bedi explains
New Delhi, Mar 9: The number of people who access the internet in India has grown by leaps and bounds in the past 10 years and with that, the way people of the country follow news has changed radically. Dailyhunt president Umang Bedi explained in detail how the user behaviour has evolved in recent times which has led the news and content providers to formulate business strategy by making best use of the technology.
In an interview to CNBC TV 18, Bedi said that Dailyhunt's core strategy revolves around providing content in regional languages with "deep personalisation" using technology. He said that it would be unfair to say that Dailyhunt is a content aggregator, but rather a "content discovery destination for Bharat".
"It would be unfair to say that we are an aggregator, we have moved well beyond that. When you think about our sources of content, we work with large organisations, about 1,500 of them, we work with 15,000 professionally generated content creators," he said.
He said Dailyhunt has the ability to use content to 'drive deep personalisation' without compromising on user privacy or details.
"This is not a user-generated content, its professionally generated content. On the news side, we also have 10,000 stringers across India who give us news content. Think of us content discovery destination for Bharat in local languages, supporting 14-15 different languages where we have the ability of using our content to drive deep personalisation without compromising on user privacy or user details. So, now we are starting to curate our own content, we are setting up studios and moving on to curating a few shows in the future," he further said.
When asked about the difference between curating and creating, Bedi said, "We haven't got to truly creating ourselves, we have a publisher called OneIndia that we have acquired and they create content."
On how technology and machine learning is being effectively used:
"We are a tech platform and not a content company per say. On the tech platform, you will be surprised but the algorithm and the machine learning decides."
"We ingest content from our partners onto our platform such that the user experience is always engaging," he said.
He said when Dailyhunt gets content in 15 different languages, "We examine what that content is saying, we look at the headlines, we look at the content in the article, we look at the images or the videos. We break up the video into images. We look and try and understand what is the genre, what is the depth, is it a technology article? Is it about mobile phones? Is it (about) comparing two low-cost models? We then tag that in local languages, and put across 25,000 different interest groups in a taxonomy which is based on 5-6 years of machine learning data on over 75 algorithms."
Further explaining how the machine learning helps in enhancing user experience, he said the content graph takes various variables into account while personalising the feed for a particular user.
"In real time when a user comes in, we have got a content graph that continuously improve depending on how he/she is consuming content. What kind of content does he/she click on, What is the time spent, What are the like/dislike signals, what are the shares and comments? On the based on that we curate a news feed that is deeply personal for them," he added.
How is it different from Facebook model"
"The first thing is that in Facebook or any social network, there is your personalisation, based on which there is a social graph, so they know everything about you. The content strategy at our end is that you could be even anonymous, you don't need to sign in. You can sign in overtime for different use cases, but what we do is that we create a content feed that is based on purely unidentifiable personal variables. Automatically the machine learns that there are certain hot topics that are coming up."
"Let's say its the GST or the demonetisation or the elections in 2019. Now, every publisher that we aggregate content from may have an alignment, they could be left aligned or right aligned, we are dead centre and neutral. We create destination zones wherein if a person has read or seen a particular type of video, we mix it with alternate points of view from another publication," he added.
On future plans:
"Coming to the TV platform that you are talking about. We are launching now or launched about 30-40 channels thats going to go up to 543 channels serving 543 constituencies in India within the app. Now, there we are very careful to ensure that there is no form of editorial bias, the content is mixed between stringers, between large publishers and we give a holistic point of view. In fact, that is going to be the lynchpin of our TV campaign that is coming out in the near future," he said.
On curbing fake news:
"Fake news is a huge menace today. Where does fake news start? Its either just fake or propoganda. Or its started by a user. We do not allow user generated content," he said.
"So all the content that comes from the stringers, goes through editorial review in-house. So we have 450 people in our company Dailyhunt, and another 400 people in OneIndia which is our entire editorial team, its all content and editorial. So we vet that content, we validate where is it coming from, what's the reputation of the source," Bedi further said.
"The second thing that we do is, that large publisher content can also sometimes get flagged as fake news and there we do three things. We have a publisher quality score, a content quality score for each article , and if there are two flags from the users saying that it is fake, then we manually examine it and go back to the publisher. The most important thing is that for political content we do not take it from small publishers, we don't take it from PGC providers unless it is stiger who is on our rolls who has a contract with us. We only surface political content fromlarger more established publishers who have a certain repoutation," he added.