Українська правда   /   Межа

Ukraine is testing a neural network for automatic analysis of hostile media resources

Ukraine is testing a neural network for automatic analysis of hostile media resources
0

The Center of Excellence "Boryviter" has begun testing the "ODCR Assistant" neural network, which is designed to automate the process of passive collection and analysis of information from enemy open sources. Militarnyi writes about this. The project is implemented by the analytical department of the center and is aimed at automatically detecting useful data from the enemy media space.

The module works with unclassified data streams, including Russian military social networks and technical blogs. The system processes about a thousand messages every day. Algorithms sort them using dictionary markers — "green" and "red". If the message contains "green" keywords and does not contain "red", it goes further to form observations.

The next stage will include checking the data for novelty, as well as its systematization. At the same time, the developers note that full automation of the analysis is not yet achievable - human participation remains necessary.

In addition to the "ODCR Assistant", the team is working on a language model for the Ukrainian military that will operate on the basis of neural networks and have an interface similar to ChatGPT. The goal is to facilitate dialogue with experts and simplify situation analysis.

The development is carried out by the Boryvitr team on its own resources on powerful servers. The organization is open to cooperation, but emphasizes the importance of non-commercial support focused exclusively on the Ukrainian army.

Share:
Посилання скопійовано
Advert:
Advert: