Students of Samara University and National Research University “Higher School of Economics” will develop a neural network service that will simplify and speed up the process of writing reviews on a wide variety of websites where customer feedback is important: on marketplaces, in the field of various services, in the hotel and restaurant business, housing and communal services and healthcare.
Using this development, those who do not like to formulate their thoughts in writing will be able to fast record their feedbacks verbally, by dictating it to the telephone as a voice message in a messenger. The neural network service will not only translate the oral review into the written text, but also analyze the speaker’s intonation and emotions, which in future, will make it possible to more accurately determine the sincerity of the reviews left.
The service project, called Your Feedback, became the winner of the Competition “Student Startup”, organized by the Ministry of Higher Education and Science of the Russian Federation and conducted by the Innovation Promotion Foundation as part of the Federal Project “Platform for University Technological Entrepreneurship”.
“Our project is a neural-network voice service. The universal widget of this service can be easily installed on any website where a customer’s feedback is important. There are people, many of them, who find it difficult or simply too lazy to express their thoughts in writing, it is easier and faster for them to express their opinions verbally. By clicking on this widget on a website, for example, a marketplace, the buyer will be able to leave a voice message, his feedback about the purchased product. This review will be automatically decoded by our service and sent to the appropriate marketplace service. This method will save the user time and simplify the process of leaving feedback, but this is not the whole task that we set for ourselves. Our service will not just translate audio messages into text: we set the task for the service to identify and analyze the speaker’s intonation and emotions using neural networks. It is impossible to understand from text reviews what intonation the person wrote them with and what message he sent his review with. The simple phrase “Good product” can be pronounced in completely different ways: you can really praise the product, or you can say the words sarcastically or ironically”, said Anastasia Surkova, the project author, a 4th-year student in specialty “Applied Mathematics and Informatics” at Samara University’s Institute of IT and Cybernetics.
By decoding the audio and identifying the person’s intonation and emotions, the Your Feedback service will select special keywords so that in future, it will be possible to sort the reviews into different categories: negative, positive, neutral, improvement tips, and so on.
“This development should help business owners get to know their product better, pay attention to its weaknesses or to those points that are not revealed through regular reviews or feedback. Moreover, in future, this service can help potential buyers, who study reviews before buying, form a more accurate opinion about the product quality. If, for example, a product has a lot of positive reviews left by using our service, then such reviews seem to me to be more trustworthy than regular ones: you can write anything, but it’s much more difficult to fake intonation, even using different voices from various accounts. In this case, the neural network will act to some extent as an additional “feedback reliability controller”. Though, then we will need to think about some kind of labeling, like “The review was created by using Your Feedback”. But for now, this is just an idea. Later, we will think about what other additional options can be implemented so that these voice reviews provide the most useful feedback for businesses and users”, noted Anastasia Surkova.
According to her, a thorough market analysis was carried out before starting the work on the project, but no such service operating in Russia or in Russian could be found. In course of creating Your Feedback, the developers adapt neural network models for real-time streaming recognition of spoken phrases and relevant vocabulary of the Russian language. The reviews will be processed on the server, so the widget itself intended for installing on websites or in applications will “weigh” quite a bit.
Besides, in order to leave a review by using Your Feedback, you will not need to download the website or open the application. In ordinary “offline” institutions and establishments, for example, in cafes, hotels, clinics, housing and communal services enterprises, etc., it will be possible to place the relevant QR code of the service in any convenient place, even directly on the wall. When you point your smartphone camera at such a code, Your Feedback will be connected, and you can express your opinion, which will be brought to the attention of the organization’s management.
The prototype of the service is planned to present in 2026.
The material is prepared with the support of Russia’s Ministry of Education and Science, in the framework of the Decade of Science and Technology.
