Low-resource languages remain challenging for cross-lingual semantic alignment because of limited parallel corpora. In addition, conventional symmetric alignment may distort the semantic space of a high-resource language through noisy low-resource updates. To address this issue, we propose Monolingual Anchoring for Cross-Lingual Semantic Alignment (MACA), focusing on Uyghur as a low-resource case study. MACA follows an asymmetric paradigm that treats the high-resource language as a fixed semantic anchor and transfers its semantic structure to the Uyghur side. The method consists of three components: (1) Anchored Embedding Initialization for newly introduced Uyghur subwords, (2) Cross-Lingual Neighborhood Anchoring for structural alignment between Uyghur and the anchor language, and (3) Monolingual Structure Anchoring for improving the internal semantic organization of Uyghur representations. Experiments centered on Uyghur-Chinese show that MACA outperforms LaBSE, the strongest off-the-shelf multilingual baseline in our comparison, by 7.67 points on cross-lingual STS. In an exploratory Uyghur-English zero-shot setting, MACA also surpasses LaBSE by 2.45 points without using Uyghur-English training data. These results provide evidence for the effectiveness of MACA in the evaluated Uyghur setting and suggest that monolingual anchoring may be further explored for related low-resource languages, such as Kazakh, Kyrgyz, and Uzbek.
Ruohao Yan, Huaping Zhang, Yuwen Niu et al.· Journal of King Saud Univers...· 0 citations
Accurate radar tracking of highly maneuverable unmanned aerial vehicles (UAVs) is a key enabling technology for low-altitude airspace surveillance, counter-UAS defense, and UAS traffic management (UTM). Once a non-cooperative UAV has been detected, estimating its motion state must cope with nonlinear polar-coordinate observations, unknown maneuver-mode switching, and multi-scale state variations driven by agile drone flight, making it difficult for classical IMM/UKF filters and deep sequence models to preserve local maneuver response and long-term temporal consistency. We propose MSGate, a multi-scale gated temporal network organized along an “observation-representation-fusion-constraint” pipeline. A non-learnable physical front end maps polar measurements into a Cartesian observation trajectory of the same dimension as the UAV state. Multi-scale gated convolution and RoPE-Transformer encoding extract local maneuver responses and long-range dependencies. A shared gated dual-path decoder fuses the two paths adaptively at each time step and channel, and velocity-smoothness and position-velocity kinematic consistency terms regularize the predicted trajectory. On the real-UAV datasets UZH-FPV, EuRoC MAV, and NeuroBEM, under a unified range-azimuth observation protocol, MSGate attains the lowest average position and velocity errors (Pos-RMSE 0.0486m; Vel-RMSE 0.1767m/s), outperforming the strongest time-series baseline TimeMixer, and generalizes to a separate nano-quadrotor dataset (NanoBench). MSGate provides an accurate, maneuver-robust solution for radar state estimation of highly maneuverable UAVs.
