Who would buy a Space Travel Insurance?
Which movie should I watch this evening? Which household insurance should I buy for my new apartment? Which customer would buy the new insurance package for space travel?
All these questions can either be answered by human experts in movies, insurance, and space travel – or by artificial intelligence. While recommender systems algorithms have been widely used by Netflix or Amazon to recommend movies and books, recommending insurances is much harder due to the “small data” problem. While we buy and watch movies on a regular basis and thus provide “big data”, we only rarely change the household insurance or travel to space. Hence, an artificial intelligence solution has much less information to learn from and thus to make relevant recommendations.
Teaming up with ZHAW and UZH
To tackle these challenges, Veezoo has teamed up with ZHAW Zurich University of Applied Sciences in a project called “NQuest – Natural Language Query Exploration System“ — led by Prof. Dr. Kurt Stockinger and funded by Innosuisse, the Swiss agency that supports applied research between universities and companies. The goal of this project is to make natural language interfaces to databases more suitable for widespread adoption in real life situations. One of the fruits of the project is published in a recent paper at one of the leading international database research conferences. In this paper the team from Veezoo, University of Zurich and ZHAW perform a detailed evaluation of how well recommendation systems perform in finding answers to the above-mentioned questions. As the team points out in the paper, there is no optimal algorithm that works well for all kinds of recommendation problems, e.g. movies and insurances. Hence, the optimal algorithmic choice highly depends on the underlying data and the amount of information we have about the users.
Combining natural language and advanced analytics
And this is where Veezoo comes in. Veezoo is one of the first analytics tools in the world that combines querying databases in natural language as well as enabling more advanced analytics, like recommendations, in natural language. The foundations for the technology have been laid by the above-mentioned research collaboration between Veezoo, UZH and ZHAW. Currently Veezoo supports a subset of the algorithms presented in the new paper. However, the idea is to allow data scientists inside companies to plug in their own models and make them accessible company-wide over Veezoo’s natural language interface.
Even though this new feature is still in private beta, you can start using Veezoo for free today with your own data with up to 5 users.