What linked data are and why they create value

Companies today have at their disposal a mass of data that would have been unthinkable ten years ago. We can now know everything about our customers, about our suppliers, about the supply chain and about the territories in which companies operate. Much of this data is also available as open data, that is, accessible to anyone free of charge. On paper, the possibilities opened by this information are enormous — but all too often banal technical obstacles and management difficulties limit the chance to exploit it in order to grow the company’s business.

A first, simple solution for cutting the time and cost of managing a company’s own information is the use of linked data: data collected and stored intelligently, easily integrable with other data systems and often available free of charge online in the databases of institutions or non-profit research centres.

What are linked data?

The concept of linked data was theorised by Tim Berners-Lee, the inventor of the World Wide Web. Linked data are fundamentally data designed to be easily linked to one another.

Think of an address book: it is as if the names, numbers and addresses of a series of contacts, instead of being filed alphabetically in a notebook, were piled up at random, each jotted on a separate scrap of paper. In that situation, though, each piece of paper — that is, each number, each address, each name — carries additional information pointing to the other elements linked to it. With a search engine I can therefore type in the name of the person whose telephone number I need and immediately find the name and the number (just as I would have done with the notebook-address book); but I can also run the reverse search, starting from the telephone number, or ask for and obtain in moments the list of all the telephone numbers, among my contacts, of the people who live in Milan less than 5 km from Cadorna station. I can also add to that initial list of contacts, in moments, a series of email addresses retrieved from another address book. I can do the same with birthdays, information on height and weight, musical preferences, the football team supported, and very much more. If linked data are used, the information that can be aggregated is virtually infinite — and if open data are used, it is also free.

Returning to the definition, linked data have four characteristics:

1. they are often freely accessible (that is, they carry an open licence);
2. they are collected and catalogued through a resource description framework: for every element — whether name, address or number — the way in which that datum can be linked to related data is indicated, which makes integration and comparison between different types of data easier. They are, in other words, smart data;
3. they have a Uniform Resource Identifier: every single element has its own unique identifying code, and redundancies and inconsistencies are thereby avoided;
4. they are online (that is, accessible through an http protocol).

How can a company exploit linked data?

To understand how linked data can cut the time and cost of a company’s information management, it is useful to start with an example.

Imagine a company that produces food products, dairy products in particular. This company — let us call it “Bianchi Formaggi srl” — sells to wholesalers active more or less throughout Italy. The list of its customers, with their billing addresses, is in one database. In another database it holds information on those customers’ orders, so as to manage production. Then there is a third data system: another directory, this time with the telephone numbers and email addresses of the commercial contacts of each wholesaler. The list of products, finally, with all the information on costs, ingredients and production times, is in a fourth system, held by the production department.

Now, what happens if Bianchi Formaggi wants to know in detail the orders of its customers in a particular geographic region — say Puglia — in order to decide whether to promote a special offer that increases its market share in that territory? It would have to integrate all this data: information on the orders, the physical addresses of the wholesalers and the production costs of each item, in order better to decide which products to discount and by how much. It could also gain access to freely available open data held on an institutional website of the Puglia region, where tourism-arrival and visitor-consumption data for recent years and forecasts for the current year, broken down by province, may have been uploaded. On the basis of that data, Bianchi Formaggi could obtain as exact a measure as possible of the demand associated with its wholesalers in the provinces of Lecce, Taranto and Bari during July and August, and so cut down the share of unsold stock. It could also decide on special offers for the products that — thanks to data supplied by the wholesalers themselves and to the open data found on the region’s institutional site — it knows to be particularly in demand by tourists during the summer months, thereby winning a larger share of the market.

It is clear that in the situation described at the outset, integrating all this information is impossible: it is long and costly work. But everything becomes simpler if this data is accumulated in the form of linked data in a single database. Better still if the data needed is available free of charge, online, already in the form of smart data ready to be downloaded and integrated: that is linked open data.

Figure: the linked open data databases connected to DBpedia.

The solution for linked open data

Many people do not know that more and more data is available in the form of open data online, free of charge, in institutional databases such as the European data portal (which holds information on agriculture, energy and transport), the Istat data portal or that of the Italian government. There are also generalist portals such as DBpedia, which holds the data contained in Wikipedia, and more specific ones such as OpenStreetMap, which holds data useful for the geolocation and georeferencing of addresses.

How do you search for the data you need? Often these portals are simply not known. And often the same category of information may be present in several portals with slight differences — the data on tourist arrivals in Milan, for example, may be held both in the Istat portal and in the Lombardy region’s portal with slight numerical divergences. How, then, once that data has been integrated, can I keep its source under control, so that I can decide which source to rely on each time?

The solution is to use a state-of-the-art tool: the enrich solution from LinkedData.Center.

The enrich solution is used to download data from any web source, structure it and use it together with the data of your own company. Open data can be useful, for example, for marketing (data on consumers, prices, products, companies), for sales (data on competitors, markets) and for production (data on consumption, traffic). The enrich solution retrieves the useful data automatically and structures it so that it can then be combined with the data already available. The solution also includes a check of the usage licence of the extracted data and identifies its source, so that you can choose each time, in case of conflict, which data to rely on — and whether you are allowed to distribute it to third parties or not.

The enrich solution uses the most powerful technique available today for managing data, and it allows you to:

• be natively based on the reference standard for publishing open data in the world (recommended by AGID);
• keep the extracted data under control, tracing the provenance of every single datum;
• give data a meaning by verifying the correctness and consistency of the data at every moment through semantic rules;
• adopt machine-learning techniques to structure unstructured data;
• repeat and update the extraction at any moment.

For the time being it is the best solution for managing corporate data as effectively and efficiently as possible.

For more information about this solution, see the enrich solution from LinkedData.Center.