Deep learning (DL) enabled semantic communication has been demonstrated to be efficient in speech transmission under fixed channels by equivalently transmitting all the reconstruction-related semantic features, which however, neglects the effects of distinct semantic features under dynamic channels, constraining the further improvement of communication efficiency. In this paper, we propose a signal-to-noise ratio (SNR)-adaptive multi-layer joint semantic-channel coding framework for speech transmission. Specifically, to achieve the consistency between the source speeches and the reconstructed ones at the semantic level, a multi-layer semantic representation framework is specially designed for speeches, where not only the content-related semantic feature, but also the acoustic-related semantic features are extracted to combat the semantic distortion caused by channel noise and attenuation.By jointly considering the effect of different semantic features and the dynamic channel conditions, an SNR-adaptive channel encoder is proposed to fuse the multi-layer semantic features, where an attention-based gating network is adopted to adjust the proportion of fusion with aim of efficiency maximization. Simulation results show that the proposed system can significantly improve the communication efficiency and robustness under dynamic SNRs.

Many recently proposed code clone detectors exploit neural networks to capture latent semantics of source code, thus achieving impressive results for detecting semantic clones. These neural clone detectors rely on the availability of large amounts of labeled training data. We identify a key oversight in the current evaluation methodology for neural clone detection: cross-functionality generalization (i.e., detecting semantic clones of which the functionalities are unseen in training). Specifically, we focus on this question: do neural clone detectors truly learn the ability to detect semantic clones, or they just learn how to model specific functionalities in training data while cannot generalize to realistic unseen functionalitiesƒ This paper investigates how the generalizability can be evaluated and improved.Our contributions are 3-folds: (1) We propose an evaluation methodology that can systematically measure the cross-functionality generalizability of neural clone detection. Based on this evaluation methodology, an empirical study is conducted and the results indicate that current neural clone detectors cannot generalize well as expected. (2) We conduct empirical analysis to understand key factors that can impact the generalizability. We investigate 3 factors: training data diversity, vocabulary, and locality. Results show that the performance loss on unseen functionalities can be reduced through addressing the out-of-vocabulary problem and increasing training data diversity. (3) We propose a human-in-the-loop mechanism that help adapt neural clone detectors to new code repositories containing lots of unseen functionalities. It improves annotation efficiency with the combination of transfer learning and active learning. Experimental results show that it reduces the amount of annotations by about 88%. Our code and data are publicly available1.