Artificial Intelligence has become a buzzword in recent months. However, Semantic Visions, a company that processes, sorts, and evaluates news content from around the world, has been utilizing AI from the very beginning. They sell their findings to companies. "Currently, we see great potential in developing services for financial markets. But we have about a hundred actively developing ideas," says Jan Balatka, Managing Director of Semantic Visions, in an interview with newstream.cz.
Who are your main clients, and what do clients most often request nowadays?
Our clients are practically from all over the world, most frequently from Europe and the United States, and recently the Middle East. They usually request data either in the form of monitoring selected areas in near-real-time or verifying historical events published in global media, going back up to five years.
From which sectors do your clients most commonly come?
I wouldn't say that any one sector dominates. Our products are used by state institutions, as well as companies in the manufacturing and retail sectors, pharmaceuticals, finance, global consulting firms, and recently, for example, publishing houses. However, what we are currently focusing on in the development of new products is the development of services for financial markets.
What information can you find for clients in open sources?
Basically, anything. In practice, in articles, we first search for so-called named entities. Under this technical and somewhat mysterious term for laymen, we can imagine legal entities, individuals, locations, or commodities. We are also preparing additional named entities such as products, currencies, or regulations. In the next phase, we detect the themes and events mentioned in the articles. There are usually several of these in each article, relating to various entities. I would illustrate this with an article about violations of rules in the fight against money laundering and terrorist financing. The themes could be, for example, "AML rule violations," "AML regulations," or "investigations by the relevant authority." Finally, in the last step, we find the relationship between the named entity and the theme or event.
The result is very precise information that speaks of a topic, event, or entity that truly interests the client. We don't overwhelm them with a mass of irrelevant articles where the theme is mentioned but is no longer relevant to the entity. This may look good on paper, but it's of zero or even negative value to the client, as they still have to expend energy to filter out what they need.
Understanding the Past Shapes a Better Future
How do you further work with the data?
Historically, finding a relationship between a named entity and a theme or event was the furthest possible level of processing. Such processed data were offered to clients in the form of continuous monitoring or verification of past events. However, we go even further and develop a number of additional steps – for example, we build knowledge graphs from the results, which will allow for very quick querying of information context in the future, and we work on backtesting against various variables. This can be very interesting for financial markets, where based on historical data, it will be possible to simulate how a trading or pricing strategy would have worked in the past, thus reducing the risk that the company will "burn" in the future.
Your business is built on AI and machine learning, which has become a hit in recent months. How does your AI work?
AI is a magic word that has only recently become widely used, but it's a field that has been developing since the 1950s. We have been working with artificial intelligence at Semantic Visions since its inception, using different methods and algorithms for various stages of data processing. For recognizing named entities and clustering, for example, we use machine learning, for categorizing themes we use a knowledge system, deep learning for filtering inappropriate content, and I could go on. This allows us to optimize accuracy, yield, speed, and ultimately the cost of processing.
What other services are you currently preparing?
As I mentioned, we are currently focusing a lot on developing services for financial markets, where we see great potential. At the same time, we are developing the base of our so-called signals, i.e., themes or events tied to the named entities. We are continuously adding new themes and events in all twelve languages in which we can search and sort data. We are working on developing cooperation with other Czech or foreign companies, through which we can now detect the leak of login details on the dark web or recognize the authorship of an article. We also plan to integrate our data into other global software, similar to how we are one of the data sources for managing supply chain risks in the SAP Ariba system. With one of our clients, we are discussing expanding searches in other languages…
When we compiled a list of requirements from existing clients and our own ideas three-quarters of a year ago, we came up with almost a hundred items. So, it could be said that we are in a constant catch-up phase. This is a huge motivation for the whole team, and I believe that thanks to this, we remain at the top of our field.
Jan Balatka
A specialist in technological projects. After his studies, he began working in the cybersecurity risk management department at Deloitte Czech Republic. Over time, he focused not only on cyber threats but also on analytical and forensic technologies for fraud investigation services. He then led the Analytics, Artificial Intelligence, and eDiscovery division for the Central Europe region at Deloitte. During his 18 years at Deloitte, he participated in and later managed numerous projects both in the Czech Republic and abroad, especially in Switzerland, Canada, and Germany. In addition to project activities, he was also responsible for leading and developing divisions related to data analytics and machine learning. This included building alliances and partnerships not only with big brands but also with local teams of subcontractors, startups, and innovative technologies.
FLOWBOX Energy Cooperates with Town of Jindřichův Hradec in Innovative Municipal Energy Community Project
