An Intent Extraction Method for LEO Satellite Networks Based on a Rotational Position Encoding BERT Model

An intent extraction method is proposed to extract user intent in LEO satellite networks. Simulation results show that the proposed method accurately identifies user intent in LEO satellite scenarios and achieves lower latency and higher bandwidth efficiency under dynamic constellation conditions.