Chatbots are the new black, but for a trend to succeed it needs to work efficiently, and until now, 70% of the bots available in Facebook Messenger fail to understand user requests. Based on this rate, it is not surprising that many companies decide not to trust them to be the image of their businesses to customers.
While major players chose not to take the challenge of solving this issue, a team of linguists decided to look at the problem from a different perspective to the one dominating Silicon Valley: machine learning.
Bitext new technology, based on Natural Language Processing, is able to triple the accuracy rate achieved by any development platform like Api.ai (Google), LUIS (Microsoft) or Lex (Amazon). While these services offer their users an accuracy of about 35%, independent tests run by experts have shown that including Bitext middleware in the chatbot development process increases the accuracy to 100%.
According to Paul Korzeniowski, member of the well-recognized ArnoldIT: “Bitext chatbot service is one of the most effective content processing systems to deal correctly with variability in human statements, instructions, and queries.”
Achieving this level of accuracy at understanding customers' demands is great news for the sector. Companies will have an enormous power to talk to their clients and fulfill their needs faster than using traditional channels like call centers or even emails. This technology also opens the door to a broad range of operations that can be put in action to increase sales and revenues using messaging apps without much effort on the company's side.
One of the biggest causes of low accuracy is the lack of data to train the bot. Imagine you need to ask a retailer to cancel your order, there are many ways to express the same thing, and each of us will surely use a different one. If a chatbot needs to understand all of them, someone must tag all these variants manually in the system, and that takes very long time.
If there is one thing we can all agree on, it’s that in technology time is always of the essence. If someone stops for a second, they are out of the race. Bitext Natural Language Generation service solves this problem by generating automatic data to train your bot. Besides saving time, companies also avoid human-made errors and inconsistencies in the data, and there is no violation of users' privacy because the data is created automatically.
Another key feature of Bitext NLP Middleware is the possibility of analyzing and extracting the most out of users' requests. Bitext offers the opportunity to connect its middleware with other NLP and linguistic services provided by the company. Sentiment analysis can be provided on demand. It is also possible to extract all types of entities and concepts and create knowledge graphs that provide related information.
On top of all these benefits, imagine this service available in more than 50 languages. As of now, there is no company offering a chatbot in more than two languages. Alexa, for example, only works in English and German and its performance is poor in many cases. The current technology at Bitext allows to create high-performance chatbots in a wide variety of languages.