What are the symptoms of social dissatisfaction, how AI finds abusers in social networks, prevents fraud, helps to manage the local budget effectively, and avoid information attacks.  We discussed this and much more with Margulan Alkuatuly, the CEO of the analytics company ALEM RESEARCH.

— It’s always interesting to find out how business ideas are born. How did it all start for ALEM RESEARCH?

— The company was founded by Doctor of Technical Sciences Sergazy Sakenovich Narynov in 2008, and it operated in the IT sphere. The idea for the product, which now serves as ALEM RESEARCH’s calling card, emerged in 2010-2011. At that time, I worked in the internal policy management of the Akimat of Kostanay region, specifically in the media monitoring department. I came to Almaty and met Sergazy Sakenovich. He invited me to join him as an analyst. That was when the idea to create an automated media and social network monitoring system came about. At that time, everything was done manually. Monitoring departments recorded TV programs, read all newspapers and magazines, and browsed websites by hand… it was a titanic amount of work that, moreover, did not guarantee that something would not be missed in the process.

I had practical experience in monitoring, Sergazy Sakenovich was an excellent programmer-architect; at that time, Rasul, also a programmer-architect, was working in the company, and the three of us began to implement the idea. It took more than a year to create a finished product, and in 2013 we launched the first Alem semantic automated monitoring system.

— Did it take long to prove the development’s prospects?

— It was enough to demonstrate the system’s capabilities. It was immediately clear that automated monitoring allows for faster and more comprehensive collection of information that the client needs

The first client was the General Prosecutor’s Office, followed by the internal policy department of the Akmola region. Naturally, we continued to develop the system, as real work provided food for thought and ideas. As a result, by 2018 we presented a radically new version of the system — Media Monitoring. In essence, only the idea of automated monitoring remained from the old version, but the functionality and possibilities for clients improved significantly.

— There was a time you worked on a government grant, how fruitful was that cooperation?

— We won a grant from the Science Committee of the Ministry of Information and Public Development to identify depressive and suicidal posts on social networks. We developed a model that identifies such posts semantically, which led to the creation of the Alem mental system. In 2019, as part of the grant, we tested our system, and it proved to be successful.

About 12,000 accounts were monitored, and psychologists worked on approximately 5,000 accounts that raised suspicions. As a result, 67 children received appropriate help. At the moment, the system is fully refined and ready for large-scale deployment. However, there have been regulatory and legal issues that government agencies are currently resolving.

Our system accurately identifies accounts that fall into the risk zone, but questions arise: how to identify them, how to handle the protection of personal data, who has the right to contact account owners, which state bodies should interact in this work, and so on?

The problem is that no such model exists in any country in the world; essentially, this is an innovation of Kazakhstan, and therefore work on the legislative framework for identifying suicidal posts is starting from scratch. But based on this model, we have developed another popular product — Social Monitoring, which has been successfully operating for the fourth year now.

Does this product not require a legislative base?

— No, because if we simplify it, the model’s functions consist of identifying complaints and negativity in the media and social networks for any city or entire region.

— How is it different? After all, Media Monitoring essentially does the same thing…

— Media Monitoring is more about business because it’s tailored to accomplish specific tasks. For example, identifying the impact of certain company campaigns, information attacks on the brand, tracking image-related issues, and the ability to respond quickly to negative feedback. Several banks have been using our system for a long time, allowing them to respond promptly to negative issues that could affect their image. For example, a bank client complained on social media that the air conditioners were not working in a certain branch and that the room was very stuffy. Within an hour and a half after this post appeared, the bank’s relevant service had already taken measures to address the deficiency because our system caught this post and showed it to the client.

Or there was a case when a message started circulating on WhatsApp, suggesting to quickly withdraw money from deposits because the bank was about to close. Thanks to our system, the bank quickly learned about this information attack and promptly issued a refutation, which helped to avoid potential unhealthy excitement among depositors.

— So Media Monitoring works on specific keywords, for example, the company name…

— To put it simply, yes, but the functionality of Social Monitoring is much broader because this system has to track posts by fairly vague criteria. We’ve divided the society’s activities into many spheres: public administration, utilities, security, education, transport, and so on. And the client, mainly governmental bodies, gets a complete picture of public dissatisfaction. This allows them to understand which areas have the biggest problems, where there are gaps in informing the population. And not only that. The system also enables the monitoring of the level of resonance to the statements of individual officials or posts by bloggers. Understandably, a negative post by a blogger with a hundred followers — regardless of the intention behind it, will have less reach than one by a blogger with a million followers. And of course, the resonance from the second blogger’s post will be tremendous. The system also quite easily identifies the original source of any information being spread on social networks. This helps in the fight against fakes. In general, the functionality and capabilities of the system are very broad.

In addition, since 2015 we have been collecting an archive that records who wrote about what topic and when. That means we can perform retrospective analysis, but so far this service is not in high demand. Such analyses are mostly requested by individual scientists or institutes who use the data for scientific and research purposes.

— Can it be said that based on the data obtained, it is really possible to predict the growth of tension and a social explosion?

— In principle, yes. For example, since last year, an explosive increase in dissatisfaction with public administration, that is, with local authorities, has been observed in almost all regions. Our system is capable not only of identifying thematic posts but also of determining their tonality. That is, a post or comment might seem completely neutral in content, but at the same time, it can have a pronounced negative or positive tonality.

— How accurate is the system in determining the nature of a publication, and what is the probability of error? Say, a comment like ‘the mayor is cool’ seems positive, but, for example, given the context, it’s clear sarcasm…

— In general, the accuracy of our products is not less than 80-90%, a lot depends on the set tasks. For example, for firefighters, a extinguished fire is a positive, but for a local administration, it’s still a negative. That is, for different clients, the same information can carry completely opposite sentiment. Initially, the system operates in semi-automatic mode, meaning we assist it in learning. From experience, I can say that after processing about five thousand information materials on a specific topic, the system switches to fully automatic mode, continues to learn, and, accordingly, the accuracy of the data increases.

— Is the competition in your market high?

— It’s difficult to call it competition yet. Yes, there is a local company that recently launched their system. But we provide more comprehensive data because we have been in the market for a long time and cover the entire spectrum of the information field. For instance, we document everything, even local newspapers, and enter everything published in them into the system. There are Russian and Western companies trying to work in Kazakhstan, but we have the advantage that we monitor in the Kazakh language, which they cannot do, so their clients do not get the full picture. Our systems monitor in three languages — Kazakh, Russian, and English, but on the whole, we can monitor any websites in Latin or Cyrillic, meaning at least 60% of the global internet.

The highly professional analysts of our company are capable of performing virtually any tasks related to the information field of the country and the world.

— What are the plans?

— Technologies are developing, social networks are changing their algorithms, so by the new year, we will present Media Monitoring 2.0. This version will be even more functional and fully meet the realities of the time.

Then, based on the model for detecting suicidal posts, we are preparing a system for detecting bullying on social networks. We understand the criteria for identifying victims and aggressors, and besides, there is already legislation on bullying, so we will soon present this model as well. The principle is the same, just different settings.

We are working on improving the model for detecting advertisements of illegal financial pyramids on social networks for the Financial Monitoring Agency. For the Ministry of Internal Affairs, we are enhancing the system for detecting illegal online casinos and betting.

We are also developing Social Monitoring, through which we have a partnership with approximately a third of regional administrations and ministries. And the more clients we have, the more ideas arise for the development of the system.

In general, we just keep working.

There is also an interesting project on detecting physical bullying in schools. Essentially, it is software that, through surveillance cameras, identifies and records suspicious behavior of students and sends signals to the Public Service Centers (CSCs). The problem was that many schools had analog cameras and needed to compile signals from both analog and digital cameras into a single stream. We fully integrated all the cameras into a single system, moreover, CSC operators can now review archives. The pilot project in several schools yielded good results, and most likely, the state will scale up this experience.

Based on this project, we have already created a system for medical institutions that allows monitoring the condition of patients, for example, if someone feels ill in the corridor, or someone falls from their bed in the ward, and so on. For the police, we have developed a pilot project that allows monitoring who, when, and for how long turns off their personal cameras. Manually reviewing 8-9 hours of patrol duty is difficult, but our system immediately lets you know if the camera was turned off and for how long. And finally, by the end of the year, we plan to launch a chatbot based on artificial intelligence for psychological assistance. The principle is quite simple: by passing a special questionnaire compiled according to the recommendations of leading psychologists, a person can learn about their psycho-emotional state and receive qualified advice. In certain cases, the chatbot will offer contacts of psychologists who work in the person’s location, phone numbers of emergency services. We are experiencing a shortage of psychologists, and the mentality of our citizens does not always allow them to seek help from a psychologist. And we provide the opportunity for completely anonymous consultation. I think this will be a good tool.

— Thank you for the conversation.

——————————————————————————————————————————————————————————

ALEM RESEARCH analytical company provides Media Monitoring Systems and social network monitoring based on artificial intelligence.

Cases:

The case of Akmola region, which uses media monitoring tools, is indicative. Timely feedback allows to reduce the level of residents’ dissatisfaction and to respond promptly to requests. It should be noted that monitoring not only collects negative publications, complaints, but also identifies the interests of the residents of the region.

For example, monitoring has shown that sporting events — Spartakiad, tournaments in Kazakh wrestling (Kazaksha Koores), karate, various exhibitions arouse the greatest interest among the residents of the region. Thus, valuable information obtained during the monitoring of social networks has become a guide in planning the budget for sports and cultural events.