Polish startup created the best chatbot technology in intent recognition

A recent research paper from SentiOne, a Polish conversational AI startup revealed that their proprietary algorithm achieved an intent detection accuracy result of 0.9889, far outclassing its competitors. This result firmy places SentiOne at the top of the Intent Detection Leaderboard – but what does it mean for business?

Intent recognition is the backbone of any successful chatbot. Being able to accurately pinpoint the meanings of words, phrases, and sentences based on varied, unstandardised user input is no easy feat. SentiOne’s near-perfect accuracy result means that challenge has been conquered. Chatbots are now able to easily understand customer motivations and intents – almost as well as humans!

Chatbot providers, NLU engineers and machine learning researchers around the world are competing to achieve the highest possible level of intent recognition accuracy. The goal, of course, is to eventually deliver technology which possesses natural language understanding capabilities mirroring that of human beings.

Papers with Code is a global repository of machine learning papers which ranks their results. The recent paper by SentiOne (Obuchowski, Lew, 2020), “Transformer-Capsule Model for Intent Detection”) reached #1 on the Intent Detection using the ATIS dataset with an accuracy result of 0.9889. 

Our AI team has developed a model with effectiveness in this task exceeding previously existing solutions, which was shown on the well known ATIS dataset. The solution is based on a Transformer-type neural network that has recently triumphed in the field of machine learning. We have developed this approach in order to obtain better results by combining it with so-called capsule networks, previously used in the field of image recognition. – says Michał Lew, Head of AI responsible for the project.

It is not the first time SentiOne made the results of their research public. The paper was initially presented at the largest conference on artificial intelligence – 33rd AAAI Conference on Artificial Intelligence 2020 in New York – and was selected as one of the top 20 in their category.

Aleksander Obuchowski, the AI researcher behind the project commented on this achievement: ‘This model is able to achieve state-of-the-art results, outperforming other solutions by a large margin. Our architecture is a combination of the most prominent solutions in language modelling and capsule neural networks that are said to be the future of AI. This system was developed at SentiOne and is used with the latest chatbot technology’.

Will chatbots replace human employees within the workplace as the technology progresses further? Absolutely not – AI exists only to assist humans, not to replace them. Unless a giant technological leap straight out of science fiction occurs, even the most cutting edge technology available will never be able to replace a human being. Instead, the best results are achieved when AI and humans work together. Processes that are best suited for AI automation are those that from a human perspective are tedious and repetitive. It should be noted that not all processes can be effectively handled by AI and chatbots. If the implementation of the task involves making decisions based on expert knowledge or requires creativity or empathy, only a human being will be able to deliver.


Founded in 2011, SentiOne supports brands with AI-powered customer service automation. Our solutions range from online listening and fully-integrated customer contact centres to new-generation conversational bots. 

Our team of world-class professionals, engineers, and analysts has been recognised by many awards, such as EY Entrepreneur of the Year and Deloitte Technology Fast 500 EMEA. We build natural language understanding technology in cooperation with top European universities. Using datasets from social listening, our system can achieve high accuracy in processing conversations from any field. This allows developing exceptional conversational bots (both in text and voice channels). 

SentiOne solutions have been used worldwide by brands such as Procter&Gamble, McDonald’s, Nike, Unilever, Beiersdorf, Tesco, Starcom, Mindshare, Mediacom, Saatchi&Saatchi, and many others. 


Download assets

SentiOne's logo and product images…. by using the SentiOne brand assets you agree to follow the guidelines.

download full media kits download icon

Media contact

SentiOne's logo and product images…. by using the SentiOne brand assets you agree to follow the guidelines.

PRESS@SENTIONE.COM download icon