Interview with Nima Akram, Data Scientist and Founder of NewsCord

I recently had the opportunity to sit down with Nima Akram (a data scientist), the founder of NewsCord, a news platform that aims to revolutionise the way people consume news. Built just over one year ago, NewsCord has already garnered over 48,000 followers on social media. 

In our conversation, Nima shared insights into how NewsCord leverages artificial intelligence (AI) to compare and summarise news articles, highlighting differences in coverage, tone, and potential biases across various news outlets.  

What is NewsCord?  

NewsCord is a news platform that aims to disrupt the way that people consume the news. It uses technology such as artificial intelligence (AI) to compare news articles which are talking about the same story and flagging the biggest differences in coverage between various news outlets.  

The AI in NewsCord highlights what is omitted in one article compared to another, the tone of an article and whether there are any contradictions between the articles covering the same story (amongst other things). These things make it clear to the user what the potential biases are of differing news outlets.   

Another aim of NewsCord is to make the news more accessible to people as we realised a lot of people do not have time to read in depth news articles, but they do want to keep up to date with the latest news. The AI in the NewsCord website condenses the news articles down so that it is digestible for our users.  

Can you explain how the AI in the NewsCord platform compares and summarises news from various sources?  

Where a news article comes out, let’s say from the BBC, the AI extracts the headline of that article itself, and it generates four more headlines based off that article (keeping the variations broad). These new headlines become additional search terms, which then leads to the AI searching against each of the headlines generated (in addition to the BBC headline). Then before you know it, you have around 200 articles all processed and extracted which are talking about a similar or the same topic.  

The headline search is the unfiltered stage, so we then ask the AI out of these 200 articles, how many of them are talking about substantively the same thing.   

The AI will answer and then we start to try to filter it further, to ensure the pool of articles the AI has selected are talking about is the same story.  

After this stage we start to do what we call a retrieval augmented generation (RAG). A RAG chain is where we say to the AI, here are all the articles, store each of them as documents.  

We embed them and now it is a database. We then ask the AI several questions, using various prompts such as “what are the biggest differences in coverage”, “what information is missing from article A compared to article B”, “what are the contradictions between the articles” or “what is the difference in tone between the articles” etc.  

The AI then generates a news article on the NewsCord website which flags the differences and condenses it down into a digestible article for the user.  

What is the motivation for selecting “tone” as one of the prompts?  

I believe tone is powerful in terms of trying to draw out certain emotions of the user.   

The motivation for selecting “tone” as one of the prompts is rooted in its powerful ability to evoke certain emotions in the reader. For example, if you have two news articles where one says, “a man attacks a boy in cold blooded murder” compared to the same story saying, “a boy died after attack by a man”, this inevitably draws out different emotions from the reader. Different news outlets will try to evoke different emotions depending on what their editorial guidelines are and this in turn creates a bias in terms of how the news is being reported.  

The benefit of having the AI identify the differences in tone is that it highlights the biases between news outlets which in turn enables the user to make an informed decision on how to digest a story, rather than being subconsciously influenced by the reporter. 

Currently, NewsCord is a free to use website, how do you intend to monetise it?   

We recently decided to open a new donation section on the NewsCord website. This section enables users to donate money to NewsCord to assist us in our work. We could have selected to have a paywall on certain features, however, as a management team we felt that at this stage we want to make all features of NewsCord accessible to as wide an audience as possible. You see some prominent newspapers, such as the Guardian adopt this model.  

We hope NewsCord’s call to action resonates with enough people to support our work. However, it is ultimately an experiment, and we may revisit this in the future.  

Finally, what would you say is NewsCord’s USP compared to other news comparison websites?  

At NewsCord we focus on objective differences rather than formulating an opinion in respect of what side of the political spectrum a news article lies. For example, some news comparison websites set out that a particular article leans closer to the left or right, whereas at NewsCord our focus is on the objective differences between the articles. As a result of this, we believe NewsCord empowers the user to make their own decision.