RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2
Project

Neural network helps Tricolor virtual assistant serve clients

Customers: Tricolor (National Satellite Company)

St. Petersburg; Telecommunications and Communications

Product: BSS: Virtual Voice Assistant

Project date: 2023/10  - 2024/04

2024: Implementation of the Neurobot in the Virtual Assistant

Neuronet helps virtual assistant Tricolor serve customers - a non-standard mechanic from the company, BSS which announced this on May 7, 2024.

In some scenarios, the assistant needs to take into account many factors: equipment and satellite antenna settings, channel statuses, actions already performed, and even weather conditions. There can be a lot of combinations in client responses, which implies a fairly large number of possible branches in the dialogue with the robot. In this case, the neurobot understands the speech and context of the dialogue and offers answers based on these data.

The neurobot consists of 4 modules in which human speech processing and formalization of the received information take place, updating the current state and determining the next step of the dialogue (the next question or the end of the conversation), forming and issuing an answer to the user. Since the neurobot consists of several separate modules, it is possible to combine solutions depending on the requirements and expectations of the customer.

The neurobot speech processing model takes into account both the user's replica and the assistant's question. This allows for a more accurate understanding of the user and more efficient determination of further service steps.

In this case, the virtual assistant can be configured, providing greater control if it is important to follow the script. When greater freedom of action is given, the virtual assistant is implemented using neural network technologies: the bot makes decisions based on cases from the training set. Even if the bot encounters a situation that was not in the examples, it will build further service based on the training model. This avoids manual scripting. You do not need to spend resources on entering different combinations of questions and dialogs into the script.

File:Aquote1.png
Our solution - the introduction of a neurobot in scenarios with a large variation in responses - provides a better client service to Tricolor subscribers. Already in the first days after the launch of the neurobot, the customer observed an increase in the quality of service. Our neurobot generates answers, focusing on dialogue and, if necessary, receiving information from third-party systems via API, "commented Alexander Krushinsky, director of the BSS voice digital technologies department.
File:Aquote2.png

File:Aquote1.png
We see that the bot and the client have become more understanding of each other, so requests are resolved faster. In general, the dialogues look more comfortable. Clients are satisfied with the help provided by the neurobot, - continued Ivan Kulagin, director of client service at Tricolor.
File:Aquote2.png