return Home

JabberBrain's Johan Ahlund on how AI will deliver a new breed of intelligent chatbots

Language based artificial intelligence is currently a very hot area for the media as more and more companies experiment with chatbots, text analysis and intuitive approaches to 'search.' It appears that the big dream for some companies is to be able to create intelligent chatbots that can maintain context-based dialogues with the user.

Johan Ahlund was a founder of Artificial Solutions, which specialises in creating intelligent virtual assistants and interactive customer support. He is now on the verge of launching JabberBrain a cloud-based solution that makes it possible for publishers to build high quality, advanced chatbots.

At DIS Johan will feature on a panel that will bring delegates up to speed on machine learning, chatbots, automated journalism, localisation and intention management.

 

Johan Ahlund ()

Here he talks about the potential of Natural Language Interactions and how media companies should be starting to consider how interactive content could transform their editorial approach.

***Registration for DIS 2018 (19-20 March in Berlin) is now available. Book by 13 March to avoid late registration fees. Secure your place here***

 

DIS logo ()

 

You have been involved in natural language applications for a while now? What first excited you about the technology? 

I could see that NLI (Natural Language Interaction) could make technology more accessible for normal people. Rather than the user having to adapt to technology, trying to figure out “how is this structured”, “what do I need to press” etc, NLI could work as an interpreter between users expressing themselves in everyday language, and the technology.

With a good NLI solution, the user does not need to adapt to technology. Instead the technology adapts to the user.

And how has artificial intelligence helped to develop NLI?

AI is a very broad area, whereas NLI is one very specific subsection. The main progress in NLI during the last few years has meant we are now seeing solutions that take advantage of both rule based NLI and machine learning algorithms. But the biggest advance might be that the capabilities of speech recognition have improved a lot. Thus, allowing for good quality speech driven NLI solutions.

JabberBrain ()

 

Can you briefly explain the difference between the technologies that make up the term artificial intelligence? 

This would require quite a few pages of explanation… However, focusing on where NLI fits into the AI picture and simplifying it quite a bit we can talk about “General AI” that includes areas such as e.g. image recognition, big data analysis, self-driving cars and language related AI.

Within the language related AI we have speech related AI and text related AI. Speech AI includes sub areas such as emotion detection (based on the tone of voice), user identification and speech to text conversation. Text AI includes sub areas such as text analysis (eg to analyse an article), language translation, simple natural language Q&A, natural language based search and then finally natural language interaction - which is my specific field of expertise. Natural Language Interaction allows computers to maintain intelligent context based dialogues with humans, based on everyday language and expressions.

And what is the key way that media companies and publishers should be using artificial intelligence broadly and natural language interfaces specifically?

Media companies should, as any company, take advantage of NLI to automate part of their customer support and customer interaction.

But perhaps more interesting is the rapid change in how content is consumed. Today’s readers typically don’t have the patience to read long texts. They also want immediate feedback if they have any doubts. Their expectations are based on ‘short,’ ‘instant’ and ‘interactive.’

Communication between people has also moved from emails to short text messages, again with the expectation of receiving immediate feedback.

We have seen many so-called chatbots, and there is a lot of interest surrounding them. However most of them are extremely simple without any real language capability, turning them into ‘clickbots’ rather than intelligent chatbots. By using the best available NLI technology it is however possible to build intelligent chatbots that can maintain context-based dialogues with the user.

Based on all that I mentioned above, I believe there is a great opportunity in ‘Interactive Content. Content that is served in shorter, more readable chunks of information and where the user can ask questions or request more information, using everyday language.

An example would be a person that reads a short “article” and then asks questions such as “Who is Theresa May?”, “When did it happen?” “Why is that important?” “Tell me more...” etc. Based on such questions, the content consumption becomes more of a dialogue. Thus ‘Interactive Content.’

Another opportunity for the media companies and publishers is to engage with their readers via expert chatbots that can give answer topic specific questions, or give general advice related to that topic. For example for health, wine, fishing, etc. 

 

mobile header ()

 

What solutions does JabberBrain offer and can you give me examples of the work you have done recently?

JabberBrain is a cloud-based solution that makes it possible to build high quality, advanced chatbots very quickly, cost efficiently and without the need of technical or linguistic expertise. As an NLI engine, JabberBrain uses the award winning Teneo platform that has been used by hundreds of companies worldwide.

JabberBrain is the result of my 16 years’ experience in the NLI field and three years of development and testing in multiple languages. We are now making the final tuning and will offer JabberBrain to the market at the end of March.

Are you optimistic about how publishers can use machine learning driven chatbots?

Pure machine learning driven chatbots are today too complicated to build and maintain and are also not very good at understanding the user. This is why so many of the so called ‘chatbots’ should be called ‘click bots’ as the users need to click on options rather than have a real conversation.

However, with solutions such as JabberBrain that use Intention based NLI in combination with machine learning, publishers are able to build quality chatbots without the need of a huge investment. 

***Registration for DIS 2018 (19-20 March in Berlin) is now available. Book by 13 March to avoid late registration fees. Secure your place here***

More like this

How Egmont is reaping rewards from creating ALT.dk

How traditional media know-how helped Visual Statements become a social publishing success

How to monetise product referral schemes - Breton Fischetti on Business Insider’s innovation Insider Picks

Blockchain: can it transform the media? Yes, says Burda CTO

How this startup plans to use blockchain to power a new media business model

How Cheddar revolutionises business media

ePrivacy: A loss of more than 30 per cent in digital advertising sales for journalistic media

How digital tools can transform newsrooms: cultural shifts and the growth of collaborative fact checking

How Libelle, The Netherland's oldest womens' weekly, finally made the most of digital

How technology can empower publishers to solve complex issues

Go to Full Site