Textual emotion analyzer received the first prize in the “Best Matching Engine for Cryptocurrency Trading Venue” category. It was as part of the 5th edition of the TradingTech Insight Awards USA conference, which took place on June 8, 2023. Thanks to the analysis of emotions in online discussions, the algorithm can predict exchange rates, financial instruments, or cryptocurrencies.
The heart and brain of the activities of the Sentimente company is Katarzyna Maszyna. IT is a proprietary solution developed in cooperation with scientists from the Wrocław University of Technology and the Institute of Biology of the Polish Academy of Sciences. However, investing is only one area where it can be used. First, it helps companies recognize customer feelings and use this knowledge in practice. For example in media monitoring, social listening, or matching the advertising message to the recipient. It also allows analyzing customer comments and issues related to brand safety or hate and hate speech. Thanks to it PR, HR, and Contact Center departments can effectively use it. No wonder that technology of Damian Gremling has already been used by many companies. For example by telecoms, banks, media houses, advertising agencies. And even by Allegro, Żabka, or IKEA, as well as state offices and political parties.
For business, it is used to examine emotions in customer comments on internet forums. It is an incomparably better way than other tools available on the market. The latter are based only on whether the sentiment is positive or negative. While Sentimenti identifies eight emotions and emotional arousal, not on a scale of 0-1, but as a percentage. Not surprisingly, distinguishing such emotions is much more effective in telling which client is worth dealing with in the first place.
Business model
The company earns from the sale of reports and access to a tool in which marketers enter content to be checked. The best year for the company was 2021 when it recorded over $0.15M in profit with $0.2M in revenue. A year later, revenues fell to $0.1M with zero profit, so new ideas had to be sought. At the beginning of this year, the company launched an algorithm that suggests how to play the stock market based on investors’ emotions. The monthly subscription cost for private investors is $6. But the company is looking for people willing to launch an algorithmic investment fund (Quant Fund).
Cryptocurrencies and more
Since the offered solution can measure emotions with an accuracy of the fifth decimal place, the company decided to try it in the investment industry. It started with bitcoin, the least anchored in the real economy and highly susceptible to emotions. Sentimenti extracted from Twitter all mentions of this cryptocurrency from the last two years (14M entries). Together with quotations from the Binance exchange, AI found a strong correlation between emotions contained in tweeds and cryptocurrency quotations. Prediction efficiency was of 80%. In 2022, the bot launched by the company earned about 40%, while bitcoin lost over 60%. If the bot operated from 2021, it would earn 600% with a 60% increase in bitcoin quotes simultaneously.
Based on Twitter and Reddit entries, the algorithm predicts how much the exchange rate will change within an hour. Then the bot decides to buy or sell on this basis. Work is still underway to shorten the prediction period, and additionally, industry internet forums are to be connected to the analysis. Initially, successful analyzes were also carried out for the Forex market, Amazon shares, and several dozen companies from the Warsaw Stock Exchange. The next step is to launch a bot for the most popular shares and currency pairs.
Textual emotion analyzer is also a strategic partner of Statista, for which it recognizes so-called emotive data. Recently, it has also started cooperation with one of the chat platforms for customer service. During the operator’s conversation with the customer, emotions from both sides of the conversation are monitored on an ongoing basis. It allows the former to react more effectively to the customer’s emotions. And his supervisor how he dealt with them. Sentimenti also plans to offer HR departments a tool that will enable them to measure the level of employee satisfaction while working remotely.
Textual emotion analyzer – competition for Watson
Textual emotion analyzer was founded a few years ago. Its owner decided to create a tool that tells how the person on the other side of the cottage feels. It quickly obtained a grant that allowed him to start cooperation with two Polish universities. They were to conduct advanced research for three years. They consisted of analyzing 30,000 different lengths of texts to identify specific emotions in them. The data went to AI/LM process, which propagated it first into Polish, and then into 17 other languages.
The company started building the model in 2018 when it received $1.1M funding from the NCBiR. In the meantime, it was co-financed by the SpeeUP fund with $0.25M. The company currently employs 20 people.
Many companies worldwide monitor mentions of brands on the Internet but still determine the sentiment on a scale of 0-1. IBM’s Watson also focuses on specific emotions. However, Watson breaks down the sentence into individual words, analyzes them in terms of emotions, and then takes the average. The Sentimenti AI understands the context, irony, and sarcasm. And it goes deep into the analysis of text comprehension, and only then reads and measures emotions.
Although 83% of companies measure customer satisfaction, 49% do not use these results sufficiently to change and improve. Examining the level of specific emotions can be crucial in better prioritizing such initiatives. But neither Amazon nor Google offers it in their business solutions. Sentimenti appeared on the last Google TechTalk, and maybe it will increase the number of customers.
Read more about Sentimenti here.
Read more about SentiOne, European ChatGPT for business, here.