How we improved natural language understanding at Zana Health Assistant.

Zana is an interactive artificial intelligence-based health assistant that can respond to questions with intelligent answers, concise informative articles, and personalized recommendations.

Zana is available online for conversation in two channels, Facebook Messenger and Amazon Alexa. In the next days, we will release a conversation interface in our web portal zana.com as well.

Recently, we upgraded the architecture of our natural language understanding (NLU) system and trained Zana further.

First, we improved the understanding capabilities of Zana, giving her the power to resolve and understand more medical conditions. We analyzed carefully all the queries that users engaged with Zana and captured the semantics of these requests in order to give back the best results.

We improved our entity recognition approach for expressions containing medical conditions. Zana is now able to better resolve in the sentence expressed by the user medical conditions (diseases) or medical topics (such as “blood pressure”, “child fever”, “paracetamol”, etc.) that are composite of some words (tokens), for example: “How can a heart block develop?”, “diagnosis of heart failure”, “what are the symptoms of acute bronchitis”etc.

The second approach for improvements consists in training Zana for different user intents when searching medical information e.g. ask for general information, ask for symptoms, causes, treatment options, or diagnosis of a medical topic. Zana then suggests informative articles on that topic. These articles are created by Zana’s Network of Doctors or licensed sources.

We also did a lot of work on training Zana for heart-related medical conditions. Zana can now give you detailed information for the most common heart diseases like: heart block, heart failure, heart stroke, cerebrovascular disease, hypertension, etc.

We are delighted to see an exponential growth of users lately. Zana is smarter in comprehending more queries expressed in various ways, and help you with the concise information in a wide range of medical topics.

We want to thank users that has sent their feedback and explored various skills of Zana. We will continue to enrich and keep on working to feed Zana with knowledge, so she can understand the various concerns and questions that users might communicate.