What is it you actually do and what difference do you make for a business organisation?
Identrics operates in the area of text mining through artificial intelligence. We develop technologies by which machines can start to understand the natural language of people so as to be able to categorise various documents. We work with natural languages, and machines do not have a clue how natural languages work and they need to be trained in order to be able to find their way around. Another thing we do is named entities recognition - we train a machine to identify a very specific set of named entities determined by a client out of a database of hundreds of thousands of documents. This is actually
one of our major competitive advantages to other companies that offer “ready-made” entity recognition and cannot meet a client's needs to single out a specific legal or physical person.
Today, a number of people employed in the IT industry are trudging through vast amounts of uninspiring work on a daily basis. This process could be automated, which would make it quicker and much cheaper, allowing people to do more important things which are still off limits for machines, such as data analysis. Our mission is to give people more room to be creative and to tap the real potential of creative and associative thinking, thereby restructuring the cost of an information product and making more funds available for creative work.
What practical application do your solutions have?
Our technologies could be of use to anyone who is doing data research involving an extensive body of information. For example, if a recruitment agency whose main asset is its pool of CVs has a client seeking to hire someone for a very specific position, we can relatively quickly allow the company to find a suitable candidate. To match some very specific job requirements to a large pool of CVs, we only need to write the ideal candidate's CV, to index it and then compare it to the body of info that we have.
Another practical application of our solutions would be with companies which have hundreds of thousands of customers and often need customer support centres to answer their questions, i.e. they often have a relatively high cost for customer support services. However, customers often ask identical or similar questions, and we could develop an algorithm with the help of which as soon as a customer starts typing a question, it would be matched to the body of already answered questions and a possible answer offered.
How can you help media organisations in particular?
Over the past decade media in the region have been investing a lot in developing attractive multimedia platforms that can be accessible on various types of device, and now these media have state-of-the art digital platforms via which they can disseminate information and serve ads. However, the media were slow to acknowledge that the information which they create is unintelligible to machines and that some basic tasks such as the classification of this information and the creation of linked data can be done very quickly and without human involvement. If an article mentions a politician, this article can automatically lead to a profile of this politician, or to other articles with similar content, as the whole process of selection of this linked data can be done by a machine. If a media plans to create a website that brings together content from already existing websites, this too can be done fairly easy and quickly.
We offer the media access to technologies which they would find it very hard to develop on their own. These technologies would relatively quickly improve major parameters of the behavior of a media's audience such as the duration of readers' stay and their engagement. This is particularly important for the media in the post-crisis context in which they operate - where the entry of heavyweights such as Google and Facebook has changed fundamentally the ad market, leaving the media with a very limited financial resource which they can invest in innovations related to artificial intelligence. If a media wants to diversify its business model and sell not only ads but some sort of research or analyses, it needs higher audience engagement.
What type of clients do you target?
Our clients are companies based in the US and the UK that have specialised in data intelligence and media. However, we would like to specialise in the region of Southeast Europe, which means being able to work with the languages of the countries in the region. In fact, language itself is one of the biggest challenges in computational linguistics.
Making sense of a language requires a huge investment which is not within the capacities of a single company, you need to work with the established scientific institutions that have been doing it for years, and this is precisely what we would like to do.
What is the rate of penetration of machine learning in Southeast Europe and how does our region compare to the rest of the world in this department? What reception do these technologies see on the part of the business organisations and the state institutions?
The region of Southeast Europe (SEE), and Bulgaria in particular, is quite well positioned in machine learning. The Bulgarian Academy of Sciences and the other regional scientific organisations are investing in computational linguistics, it is being taught as an academic subject, various events are being organised and the scientific community itself is investing in this type of technologies.
There are also some startups that operate in the area of computational linguistics and artificial intelligence - some from a more scientific perspective and other not, some using ready-made solutions and others developing their own ones. In SEE, the IT and outsourcing sectors in general are very well developed, as is the development of websites, however, there is still untapped potential in the area of artificial intelligence development.
Regrettably, a large part of these solutions are being perceived as science fiction and that is why our clients are mostly western companies, usually media companies or companies involved in data processing and intelligence. For the time being, the regional business, apart from the startups, relies mainly on conventional technologies. However, as the skills and technologies develop, this is bound to change.
Automated processes are quicker and cheaper, unblocking companies' creative potential. Out technologies increase readers’ engagement, allowing media to diversify their business models.
Identrics is an innovative technology provider of automated journalism, data mining and semantic web solutions, whose dedicated team builds and trains technology models based on a client’s own data set and fitting his needs.