Artificial Intelligence and data access: challenge accepted!

by | Apr 19, 2021 | Blog

Almost four months have passed since Ellycode was first announced, a time in which we have repeatedly hinted at the work we are doing on our product. In this period, many have asked us for details on what we are creating and what to expect: and now I can say that we can begin to reveal the true focus of our work.

Indeed, we have often emphasized the processing and understanding of natural language and how the technologies made available by artificial intelligence can help us improve the User eXperience in our applications. But this is only the means to an end: we want to solve a more important and specific problem for business scenarios: access to business data.

The idea grew out of discussions with my fellow entrepreneurs, who like me had to face and still face the challenge of having to make decisions for their business. If it’s possible to take the time to do it, it is possible to retrieve all the data necessary to get an idea of the problem that needs to be solved, and on the basis of them evaluate the advantages and disadvantages of possible solutions. Unfortunately, however, it is not always easy to find these data, even if most of the time they are found in the databases we have in the company. And increasingly, the amount of time we have to make these decisions is incompatible with the time it takes to find the information.

I remember the early years of Blexin well. That’s the company I founded 8 years ago and of which I am still the CEO. There the invaluable assistance of my sister, who was in charge of accounting and administration at the time, was able to pull out the data that was necessary for planning the following months. It was only when she left us that, having to deal with the retrieval of data alone, I discovered how complicated it is.

Consider that having a technical background, I was also able to benefit from my ability to write algorithms and use the database query languages we use. However, there are many companies in which managers do not have technical training, so they must ask the IT department to extract data, often using Excel, and they do so with the tool they have at their disposal: natural language.

Elly  (this is the code name of our product) tries to solve this exact problem: how to use natural language to access the information necessary to make decisions for our company. So yes, we are building a virtual assistant and yes, we are using artificial intelligence for our project, but the ultimate goal is to make data access simple and natural.

To do this, we are creating a service that can have a conversation in natural language with the user, to understand and learn what data they want to access and provide this data in the form necessary to make decisions. We have to keep in mind that, most likely, the data are saved in a form that is not what the user needs. Moreover, the navigation of this data and the manipulation necessary to change them into a form which the user finds useful is very clear to those who must use these data, but often it is not so for those who have to extract them.

For this reason, one of the two focuses of our work is based on the techniques of NLP (Natural Language Processing) and NLU (Natural Language Understanding), in order to understand the requests and instructions for manipulating and navigating data in natural language (see articles by Salvatore Merone on these issues).

Once we’ve understood what the user is asking of us, attention shifts to the second fulcrum: transforming these requests into data queries. Here there’s a small technical detail: there is a phase where we import the data in order to transform and save them in a form which we can then manipulate based on user requests. This phase is different depending on the original sources and both manual and one-off imports are foreseen, as well as continuous and / or timed updates from the source to our databases.

And here is another interesting revelation, which sees two possible uses of our technology: we offer our solution both to end users, who can use our client to access the service, and to software houses that want to integrate our services into their applications. These are two different scenarios from the point of view of accessing the data: in the first case a phase of defining the imports will be necessary, while in the second the access will be simplified by the fact that the software house has direct access to the data it wants to make available to the service.

In any case, the goal will be the same: you can ask Elly to access this information, to correlate, manipulate and navigate it, all by using natural language. We can use this data to assess the risks of a contract or the procurement of resources, but also to monitor the status of critical information over time, saving our queries and keeping them available to observe how they evolve as new data arrives in the system.

A big challenge for sure, but it doesn’t end there. We want access to the service to be as transparent as possible and can be a tool to help make decisions when needed. This is why we are also focusing on multi-channeling: integrating Elly into everyday communication tools, which are now necessarily digital. So imagine being in a meeting and being able to have the virtual assistant as a member of the meeting, to whom you can ask for information relevant to the discussion, or being in the car or at home and using the smart devices we already have to support some ideas that are just starting to form in our mind. All of this of course in total security and ensuring the necessary privacy when accessing information.

I confess that my technical background and passion for software development technologies are extremely satisfied by this project: we are talking about a (real) microservice architecture, which uses the latest versions of all the frameworks we usually work with to exploit every bit. A solution that finds its natural place in the Cloud, but which thanks to containers and Kubernetes can also be brought on premise or configured in a hybrid solution.

For our readers who are developers, we are about to open a technical blog where we will tell you about the technologies we use and how you can integrate Elly into your solutions. For readers more interested in the functional aspects and the opportunities that our solution can offer, we will continue to share how Elly can help you make better decisions every day, providing your instinct with data support to better evaluate which way to go.

If you have not already done so, subscribe to our newsletter: you will be able to access special content and be notified when we open subscriptions to the beta to try Elly in advance.

Too good to be true?

The challenge has been accepted: keep following us!

Written by

Written by

Michele Aponte