ESA Project OTEG for the GMES Space Component Data Access

The ESA – European Space Agency commissioned Epistematica to develop a software application for the GMES portal within the Space Component Data Access functions.

The software application offers two services that allow a more efficient access to the data ESA makes available through the portal:

The data identify images of the Earth coming from a constellation of satellites launched by ESA within its Earth Observation activities.

The ESA provides some of its partners with semi-processed images they treat to obtain products tailored to specific sectors, for instance, land monitoring, marine environment, etc… These products are then sold through the ESA website.

What were the ESA problems in selling its products?

Customers bought always the same products.

For example, customers dealing with land monitoring, bought only products made by ESA partners operating exclusively in that area. However ESA – being the producer of the images – knew that other partners, operating in other sectors, produced images interesting also for land monitoring, such as images of coastal lands, islands and so on, produced by partners operating in the marine environment field.

Therefore the problem was to ensure that customers interested in images coherent with the needs of specific sectors became aware of the existence of other interesting images treated by partners operating in different fields.

The ESA’s need was to show the entire catalog of products in a less cryptic way and to make the access easier to maximize the sales of the Earth’s images.

What solutions can be designed to face these problems?

To resolve this problem usually each product is associated to a number of relevant words – the tags. For instance, if the terms “land monitoring” and “marine environment” are associated to a specific image, who does the search using “land monitoring” as a keyword can also find images belonging to the “marine environment” category that are interesting for the land monitoring sector.

This approach does not completely resolve the problem because it is necessary to describe the images with a lot of tags, and that causes a lot of “noise”. As soon as the number of images raises, the searches produce too many results.

Anyone using a normal search engine knows what “noise” means. By performing keyword searches, the most part of the results obtained have nothing to do with what we are looking for. The e-commerce systems generally run in this way: a search by keyword produces a list of products likely to meet the needs of the customer. In fact, systems based on tags always return a list whose header is: “Since you are interested in that product, maybe you’re looking for this one too …”

In the case of a B2C e-commerce business, a search returning a list of products that just “probably” meet the customer needs – according to a statistical model – does not have a serious negative impact. At most, the business loses a sale if the list offers many products that are not interesting for the customer.

In the case of a B2B e-commerce business the risk is rather to lose the customer. In fact, the customer may believe that the supplier doesn’t live up to satisfying his needs, since it proposed products that have nothing to do with what he wants.

The approach based on tags has also another limit: the customer performs the searches using the keywords and the vocabulary he knows. He orients himself within the “noise” thanks to what he knows and understands. In fact, if you do not know a specific subject, you certainly don’t know the vocabulary used to describe that subject and therefore you can’t enter any keyword in the search window. The customer can’t find products that he doesn’t know because it ignores the terminology necessary to search for them.

For a B2B e-commerce system, as the ESA’s one, it is therefore preferable to use a deterministic approach. Thanks to it, the e-commerce system is able to suggest to customers only the products that actually meet their needs and not those that “maybe” meet their needs. This approach is also independent from the lexicon used in the search process.

How did Epistematica resolve the ESA’s problem?

Espitematica has designed a solution that makes use of technologies based on logic.

Thanks to these technologies, in addition to the use of pertinent terms, properties and relations of products have been described in a formal way. Indeed, the knowledge ESA has on its products has been described through the formalisms of logic. This made the use of automated reasoning possible leading to the development of a software able to use the ESA knowledge to drive customers towards products logically, not statistically, more pertinent to the search criteria.

This solution enables the customers to perform searches using all the knowledge ESA has on its products, rather than using only their knowledge, which is partial and sectoral.

What did Epistematica implement?

Epistematica created two software applications: one to perform searches within the ESA’s catalog of Hearth Observation products and the other one to navigate within it. Users are generally government organizations and enterprises that use the images of the Earth for their business or institutional purposes.

To realize this software application, the Epistematica’s experts used an original methodology for designing and developing of knowledge-based systems.

Epistematica’s experts in Knowledge Representation designed a system that makes use of technologies based on logic. The result of this activity has produced the application requirements and the design of the knowledge base. Then, the Knowledge Engineers performed the analysis and developed the system’s software components.

The application architecture is based on XML / Java. The knowledge base was built in OWL / DL and consistency tests were carried out using the services RacerPro. To carry out the functions of the automated reasoning some classes produced within the Protégé projects has been utilized. The graphics functions for the navigation within the knowledge base have been implemented with the Prefuse visualization toolkit.

The project is described in and

What have been the operational phases?

  1. So we interviewed ESA experts, and created a knowledge base containing all the information about the applications, like “land monitoring” or “topography”.
    This part of the knowledge base has an hierarchical structure, showing that, for example, “topography” is a sub-application of “land monitoring”, which means that the reasoner – a type of inferential engine used to query special databases where data are described according to the formalism of logic – knows that all the satellite products relevant for “topography” are also relevant for “land monitoring”.
    The picture below shows a little excerpt from this knowledge base.
    Water pressure
  2. Next, we created a knowledge base to organize all the satellite products offered by ESA: each product has been described in from a logical point of view, thus in a way that is independent from technical and low-level details (like storage locations). Also, relations between products has been described in the same way.
    Here you can see an excerpt of this knowledge base represented in tabular form:
    Climate change
    So at this point we had a formal representation of the end-users’ mind (their application domains) and of the products offered by ESA (satellite data).
  3. The last step was to create a knowledge base to provide a mapping between satellite products and application domains, by using automated reasoning techniques.
    Now the user could search and browse its application domain, and automatically obtain all the relevant satellite products.
  4. With this use case in mind, we designed and implemented the software system, and integrated it with the existing ESA portals.
    We added an interactive visualization tool to browse the knowledge base, and provided, for each satellite product, a link to the ESA portal where the user can actually purchase the product.
  5. We also designed another similar application, this time using a simpler visualization scheme but adding a textual search functionality.
  6. This application allow the user to use the more familiar concept of a search tools, while retaining all the knowledge based functionalities of our system.

Why is the ESA project a good example of what Epistematica can do for any company and organization?

What has been done with the space products can be done with any other type of product.

Epistematica uses all the knowledge that businesses and organizations have on their products, services and information, and organizes them according to the paradigms of logic to build software applications that act as an expert in the flesh.

Using these software applications to display products, services and information, it is as if the organizations make at the customers disposal a real assistant to help them chose what fits with their needs. In fact, this assistant is only a “robot” able of reasoning thanks to the knowledge that has been expressed specifically for him in the form of logical instructions and not just in the form of words of a natural language.

To say it with a metaphor, the ‘robot’ performs the duties of a librarian who, while possessing previous knowledge and skills, interacts with a library’s user offering information that is pertinent and consistent with the user’s book search. That means the librarian offers knowledge that the reader doesn’t have yet.