Fake news, human behavior analysis and prediction, filtering and serving content in the context of semantic technology were among the main topics, discussed in a recent interview for Bloomberg TV, in which Velko Kamenov talked to Ontotext’s CEO Atanas Kiryakov.
Read the entire interview below to learn more about the nature of fake news, their relationship with Donald Trump and Brexit and what keeps Ontotext’s CEO up at night.
Bloomberg TV: The fake news topic gained popularity in 2016 and big companies such as Facebook and Google declared war on it as the most affected by the impact of fake news. How does semantic technology address this problem?
Atanas Kiryakov: There is no way to beat fake news and to completely eradicate the spread of misinformation. If someone is inclined to believe that it will snow in July, for example, or that they can lose 50 pounds in a week without any effort, nobody can help them. Parts of the population will always believe such things, no matter what.
Artificial intelligence simply cannot provide the ultimate answer whether something is true or not, whether an article is misleading or not.
It is true that there are things that are obvious nonsense and technology can help sifting those out. But what will happen then? The next generation of fake news will be more subtle in creating misleading information. So, there will always be fake news. Only next time they will be better crafted.
The Benefit of Semantic Technology in the Fight Against Fake News
Bloomberg TV: Do you mean to say that technology is not the answer to fighting fake news?
Atanas Kiryakov: What technologies can do is make it easier for people to check the veracity of what they read. Because very often, it takes us up to 20-30 minutes searching in Google to check whether what we have read is so or not. And since this process of verification is so time consuming, we just skip the checking and move on, taking what we have read for granted.
What we can use machines for is to do a better analysis. They can examine a piece of news or a post in the social media and tell us how this particular content has been referred to, to what extent the information in it has been verified, whether it can be backed by reputable sources, etc.
Bloomberg TV: So technology can save us these 20-30 minutes for verifying and fact checking as you explained?
Atanas Kiryakov: Exactly. And on top of that, technology can give you additional context so that you can get other pieces of content related to the topic, with one click. This is something that can realistically be done with semantic technology.
All else is the same as believing in a silver bullet for solving all problems with fake news. Click To Tweet
Data Analysis as a Tool for Media and Politicians
Bloomberg TV: There is another interesting aspect aside from the fake news trend. Let’s look at Donald Trump’s campaign. It gained popularity with the strategy devised by his team of social media analysts and what they did to better target voters. What is your take on that? How did they take advantage of data analysis to get better results? And also, do you think this approach will become more widely spread among politicians and companies?
Atanas Kiryakov: The techniques used by the company helping Trump’s analysts are well known. They are based on accessing large volumes of personal data for psychological profiling and using it to microtarget specific messages to specific people.
What I find interesting about this campaign is the way these techniques were used and the result achieved with their help. They led to a big change and showed that if you have $25 million and funnel them properly to a few states where you need to turn around public opinion by 1%, you can achieve this effect. In a sense, this was a very good use case, demonstrating what digital marketing can do with the help of data analysis technologies. It was a public use case and anyone can learn from it.
Bloomberg TV: This approach obviously worked. Can we expect to see the use of such technology more and more in the future ?
Atanas Kiryakov: Actually, these techniques are currently used very actively for commercial purposes. All vendors who try to sell you one more thing do precisely what the people from the Trump’s campaign did. They do it all the time as they try to make you spend $5 or $10 more on something. But if this or that large online store made, say, $3 billion more last year, they don’t brag about it. For them it’s just business as usual.
Politics and the Inevitable Shift Towards Data-Driven Strategies
Bloomberg TV: Isn’t it dangerous that these technologies enter politics now or was it inevitable?
Atanas Kiryakov: I think it was inevitable. For me, the question was not whether, but rather when it was going to happen as we see that the price for creating such an effect is well justified. The Trump’s campaign had this powerful effect because, on the one hand, it was the first time this technology was applied on such a scale. And on the other, because sociologists are still not ready for it. This, perhaps, is the most important thing and it is valid for Brexit as well. The fact that sociologists still dwell in a comfort zone where they believe that interviewing thousand people the way they have done it for many years in the past can lead to reliable opinion polls. It can’t.
Bloomberg TV: Is that the reason why two of the major events in 2016 – the US elections and Brexit – were so incorrectly predicted by many of these sociologists.
Atanas Kiryakov: Yes, exactly. I’ll give you an example. Shortly before Brexit, Ontotext made an analysis of Twitter data. It showed that those who wanted Britain to leave the European Union were much more active and had significantly greater impact on people in Twitter. And this was on all levels – from the number of posts to the number of views and retweets, to the number of interactions and likes. We measured this influence using different indicators and the analysis showed that the campaign supporting Brexit was considerably stronger. About twice as strong.
Bloomberg TV: So you are saying that what happened was not a surprise although it came as a shock to many?
Atanas Kiryakov: Yes, absolutely. And we published this analysis a week before Brexit so there was no way we could have known what was going to happen with the referendum.
I gave you this example because this wasn’t even the most complicated thing we’ve done in terms of data analysis. It was just a use case that was extremely visible. Many companies offer technologies for analyzing social networks and to some extent this is now a commodity, something generally available to everyone.
Bloomberg TV: You describe these data analysis technologies as a commodity. Is there a market demand for such products and will we see more accurate predictions based on these technologies?
Atanas Kiryakov: Yes, for sure. The market just needs to adjust to them. The polling agencies as well as those who commission them, i.e. politics and the media, will need to adapt to the new realities.
The Role of Technologies in Reinforcing Beliefs
Bloomberg TV: Before we get into the details of what Ontotext does, let’s talk a little bit about the technologies that reinforce polarization on certain topics. As you know, people tend to click on information they already believe in. Isn’t it dangerous that, through targeting, technologies enable people to read about what they initially believed in? For example, if a person has a strong opinion on a given topic and technologies show this person the same information everywhere, which reinforces this opinion, do you see a problem in it?
Atanas Kiryakov: No, I do not see a problem in it. I think people need to learn and understand how this works so they are better equipped to counter it.
Mankind has found ways to react to far worse things than what now threatens us in the social media. Click To TweetWe just need some time to adapt.
The Internet itself brought many problems but with time society found a way to overcome them. The effect you are talking about is called “echo chambers”. Although we have access to a lot of information, the way we are recommended content by search engines or social media isolates us in what we have previously read and believed. Usually, this is done with good intentions, but the effect is – just as you said – reinforcing and amplifying certain beliefs.
Actually, Ontotext took part in a research project funded by the European Commission whose aim was to develop techniques for diversifying information more easily and for weakening self-reinforcing information. Fortunately, there are many techniques for overcoming this.
The most powerful weapon in this direction is to make it easier for people to get informed about more things. For example, when you read an article to be easy for you to get more information or to be easy for you to judge the degree of truthfulness of this article or its credibility.
It’s the only way, although there is a lot of work still to be done. Our typical customers are today’s biggest publishers such as the Financial Times, the BBC, Euromoney, scientific publishers, financial agencies and generally companies that work with the most expensive content, with the most expensive information.
The Value Semantic Technology Brings for Publishers
Bloomberg TV: How does your semantic technology add value for them?
Atanas Kiryakov: We do semantic analysis of their content and add more information to it by describing it with special tags. Instead of just adding a string that says “this article is about Paris”, we add an identifier that states uniquely whether this article is about Paris in France, or Paris in Texas, or Paris Hilton, or something completely different.
In this way, semantic technology makes the content of a certain topic clearly identifiable. It also allows content to be clustered around a given person or a topic. These are the so-called story lines – a series of media materials related to the same story.
So we help publisher by making it a lot easier to retrieve and recommend content on the same topic, or related stories, or trends. This makes it easier for journalists to prepare content and for editors to select content. For example, they can see which topics are the most popular in an edition, based on the consolidation of the number of reads from several articles related to a certain topic. And this is done automatically, it is not something that they have to do.
The Financial Times, for example, uses our technology mostly for personalized recommendations. When you read articles on their website, our technology bases its recommendation on what you’ve read before, what topics you were interested in, how connected they are, as well as what are you reading now and this leads to much better personalized recommendations.
Semantics and Social Media
Bloomberg TV: Are you planning to develop and expand your portfolio to assist not only big publishers but also social media, like what you did with the Brexit analysis? Was this a one time exercise or do you want to make a product out of it?
Atanas Kiryakov: We are currently working on a product that combines social networks and news. It offers following up stories on the one hand and on the other, it facilitates checking the veracity of facts. Most likely, we’ll announce its release in April.
However, such products are still intended for media publishers, not for end users. End users will benefit from these technologies through the media. But, yes, for the last three years we have been working very actively on social media. And last year, the level of maturity of our technologies allowed us to start making targeted analysis, as with Brexit. Now, we are in a position to announce a product that can make a similar analysis on any subject.
Bloomberg TV: What are your plans for the future? Do you have anything exciting to share with us at the end of this interview?
Atanas Kiryakov: Ontotext is actually part of Sirma Group and on the level of Sirma, we have serious plans in terms of cognitive services and other products in the area of artificial intelligence.
As for Ontotext, we started working with these technologies back in 2000. In the first 7-8 years we invested time and effort in research, much of which fundamental, in the fields of text analysis and data interpretation. Then, in 2008, we started commercializing and productizing this technology. This was the period of the so-called early adopters. Very often, we worked on projects that weren’t cost-effective for us just to see how this technology could help in a given domain.
We have worked in many geographical locations and in almost all industries you can think of until we found where these technologies added most value. It really took a lot of resources. But now we know our semantic technology is paramount to organizations working with expensive content. Click To TweetThese are media and newspapers, scientific publishers, financial agencies. We know how to work with this type of clients and this business direction is quite stable.
What keeps me up at night these days and what would be an interesting direction for Ontotext is how to democratize the use of semantic technology.
What do we need to change in the way we create these systems so that smaller publishers, including bloggers and customers could have access to semantic technology? So, our goal now is to make what we do for the Financial Times available as a cloud service for your blog.
If you want to learn more about media content and semantic technology, take the opportunity to learn from the best and listen to a recording of our webinar about Smarter Content with a Dynamic Semantic Publishing Platform.
For more information, contact Doug Kimball, Chief Marketing Officer at Ontotext