• Jens Dahl Møllerhøj, Lead Data Scientist at BotXO

What makes our chatbots intelligent?

Danish version here

We are proud to say that here at BotXO we create the smartest chatbots in Denmark. Our chatbots are intelligent because they understand what you tell them.

If you for instance ask one of our chatbots: “What does it cost to send the products to my address?” Then the chatbot will understand that you are interested in knowing the price of shipping.

The artificial brain

Our chatbots are equipped with an “artificial brain”. It’s an artificial intelligence (AI) that use algorithms and mathematics to figure out what you write to it. When you write a sentence to the chatbot, it sends the sentence to the artificial brain, which then tries to figure out what it is you mean.

Our chatbots are equipped with an “artificial brain”

Our chatbots are equipped with an “artificial brain”

An artificial brain uses several different algorithms. The algorithms are designed to perform Natural Language Processing (NLP).

NLP is the common term used when describing the methods that are used to make computer programs that understand what you tell them. It’s very difficult for a computer to understand human language but the data scientists are working on making it possible.

A computer is very easily confused

A lot of research has been focused on how the grammar of our language was built. By writing grammar into computer programs the researchers tried to create computers that could learn how to read.

But because we use language in an infinite number of ways, it turned out to be incredible difficult to write grammar into a computer program. Humans are good at understanding what people mean, even if they don’t use the correct grammar but a computer is very easily confused. For instance, if you tell a chatbot:

“I eat spaghetti with a fork”.

Then the computer knows that you use a fork to eat your spaghetti with. But if you say:

“I eat spaghetti with meatballs”.

Then it thinks that you use meatballs to eat your spaghetti with – which isn’t smart.

What is AI and Machine Learning?

Unfortunately, using grammar to understand language isn’t the best way forward – the forks and meatballs are too easily mixed together. Instead we use Machine Learning.

If the computer has read enough text about Italian food, then it’s easy for it to understand that forks are used to eat spaghetti with and the meat balls are mixed in the spaghetti.

With AI and Machine Learning you feed the computer a lot of text and then it’s up to the computer to figure out which rules to make. For instance, if the computer has read enough text about Italian food, then it’s easy for it to understand that forks are used to eat spaghetti with and the meat balls are mixed in the spaghetti.

How we use Machine Learning at BotXO

We use the newest technological research and we work closely with the University of Copenhagen, which are world-renowned for their research in NLP. So, by using the best algorithms on systems that create value for our customers, we have built Denmarks smartest chatbots.

Our chatbots use several algorithms to understand what you say to them but also to figure out what to reply. When you send a message to one of our chatbots the artificial brain begins its analysis and the following are examined:

  • Sentiment analysis: The artificial brain uses advanced algorithms to decide what mood you are in.  
  • Spelling corrections: The artificial brain checks for misspellings and try to correct the spelling, so it can better understand what you mean.  
  • Named Entity Recognition: The artificial brain looks for named words in the sentence you have send. For instance, person names, names of companies or places.  
  • Dependency Parsing: The artificial brain divides the sentence into different substitutes and finds out what the link between them are.  

By combining a number of these algorithms our chatbots end up understanding what you tell them. That’s why we say that our chatbots are the smartest in Denmark.

If you are interested in hearing more about how to get started, please contact us at  hello@botxo.co – or give us a ring if you need more customer cases +45 26 71 58 45 .

2018-08-31T18:36:42+00:00By |Company Corporate news|

About the Author:

As Lead Data Scientist at BotXO I'm in charge of ensuring that our state of the art Artificial Intelligence lives up to its full potential. I bridge the gap between academia and the industry by implementing commercially applicable algorithms from the latest Machine Learning research. By doing so, I ensure that our Natural Language Processing keeps up with the cutting edge without diverting into a research project. I am a polyglot with regard to programming languages, but for my data science work, my tools of choice are python libraries such as Pandas, Numpy, Pytorch, Tensorflow, Spacy, Gensim, NLTK and Scrappy. I also do quite a bit of ad-hoc data extraction with SQL, various unix tools and the infamous occasional javascript snippet. Before BotXO, I worked as a freelance engineer for various startup companies. I hold an MSc in Computer Science from the University of Copenhagen.