A virtual assistant bridging Earth observation and user needsĀ 

Title image with a satellite and planet Earth

Originally published on https://domino-e.eu.  

Raivis Skadins, Director of Research and Development at Tilde, shares insights into their work with virtual assistants and their role in the Domino-E project. He highlights the journey from animated assistants to robust conversational systems and explains how theyā€™re bridging the gap between users and Earth observation technology.Ā 

Weā€™re happy to be here at your office in Riga today. Can you introduce yourself and where we are? 

My name is Raivis Skadins, and I work at Tilde, a language technology company in the Baltics with offices in Tallinn, Vilnius, and our headquarters here in Riga. We have always been focused on language technologies. In the 1990s, we started with proofing tools, and then we moved to dictionaries, machine translation, speech technology, speech synthesis, and recognition. For the past ten years, weā€™ve been working on virtual assistants. 

When we started, nobody spoke about virtual assistants. We began with animated heads because we thought that was where the industry was headed. Later, we discovered that people care more about the quality of the conversation.

Users prefer virtual assistants that feel natural and smart, and that has guided our direction ever since. 

What makes a good virtual assistant? 

Thatā€™s an interesting question. From my perspective, itā€™s not just about technology. An engaged customer is key. If the customer or patron of the virtual assistant is truly invested in creating a smart, advanced solution, the project is more likely to succeed. The same goes for having a motivated, engaged team that believes in the project. These are the most important success factors. 

From the end userā€™s perspective, itā€™s more complex because users are very different. Some prefer brief interactionsā€”just a single keyword and a one-sentence answer. Others enjoy chatting with the assistant. Some expect deep understanding, while others just need quick reactions to keywords. But one thing is clear: users dislike virtual assistants that canā€™t help. A simple ā€œI donā€™t understand, please rephraseā€ frustrates them. They expect smartness, even in basic systems. 

How does the Domino-E project make use of a virtual assistant? 

In Domino-E, weā€™re building a virtual assistant to help users interact with a sophisticated system. The system connects to satellites that take pictures and make observations, but itā€™s not easily accessible to most users because of its complexity. For example, if someone working in a municipality wants a satellite image of their area, they might not know where to go or what information they need. They might not even know the right questions to ask. Thatā€™s where the virtual assistant comes in. 

The assistant acts as a bridge between non-expert users and the Earth observation system. It simplifies the process, guiding users through questions about locations, timing, and conditions, and translates their needs into technical language that the system can understand. 

How does the assistant complement human operators? 

We donā€™t see virtual assistants as replacing people but as enabling them. Currently, customers rely on domain expertsā€”operatorsā€”to consult with them, specify needs, and book satellite observations. These experts are limited in capacity. By automating routine tasks, the virtual assistant allows operators to focus on creative, complex situations that require their expertise. 

For example, the assistant collects key information from clients, such as geographical details and timing preferences, and organizes it for the operator. This reduces repetitive work and ensures operators can concentrate on providing valuable advice for non-standard cases. 

How has development progressed, and what challenges remain? 

When we joined the project, Earth Observation was an entirely new field to us. Initially, we struggled with the terminologyā€”what is ā€œground segment,ā€ ā€œmissions,ā€ or ā€œantennasā€? This gave us a unique perspective because we were in the same position as many users. That helped us better understand their needs. 

Now, weā€™ve developed the first version of the assistant. It organizes dialogues, asks users about their requirements, and collects all necessary parameters. The next step will involve real-world testing. When actual users interact with the system, they may approach it differently than we expect, and that will present new challenges. 

How do you approach testing? 

We start by imagining how conversations will flow, creating scenarios, and testing them with users to see if the system performs as expected. Then, we invite testers who are unfamiliar with the system to interact with it from scratch. The assistant needs to handle unexpected or out-of-context questions gracefully. 

The assistant also helps with two primary tasks: browsing existing catalogs for images or placing new image orders. Catalog searches are critical, but they fall outside the direct scope of the Domino-E project. To simulate this, we use services like those from the Copernicus project. If the desired image isnā€™t available, the assistant helps users define the parameters needed to order a new one, sending the request to the coverage service. 

What role does knowledge management play in the assistantā€™s lifecycle? 

We think of virtual assistants as employees. Like any new hire, they require initial training and ongoing updates.

Over time, as the system evolves and new data becomes available, the assistant must adapt. Outdated knowledge must be replaced, and new capabilities must be added. This continuous improvement is crucial to ensure the assistant remains useful and relevant. 

What has Tilde gained from Domino-E? 

For us, so far itā€™s been an exciting learning experience. When we started, we were essentially working from a blank slate. Every interaction with our partners has expanded our understanding of space technologies and helped us grow into this industry. Joining the space domain has been challenging but also deeply rewarding. For those willing to step into unfamiliar territory, itā€™s a journey worth taking.

Curious about implementing AI in your organisation?
Contact us today and see how our solutions can improve your workflow.