Keyphrase extraction (KPE) aims to automatically identify salient phrases that encapsulate a document's core concepts. While previous method based on prompt learning employs prompt-guided encoder-decoder architectures, its performance exhibits significant sensitivity to manually designed linguistic templates. To address this limitation, we propose a novel unsupervised KPE framework Pre-trained Prompt Rank (PTPRank). Specifically, PTPRank deploys our self-supervised model PromptT5 to dynamically generate domain-aware prompt templates through trainable parameters, eliminating the need for manual template engineering. Furthermore, we introduce a Discrete Wavelet Transformation (DWT) module to suppressing attention noise while preserving semantically critical patterns. During candidate ranking, we synergistically combine the model's generation probabilities with a novel Relevance Score metric that quantifies statistical salience through term distribution analysis. Comprehensive evaluations across six benchmark datasets demonstrate that PTPRank performs on a par with the latest large-language-model-based SOTA method.

Knowledge tracing focuses on modeling learners' past answer sequences to trace the evolving knowledge state and predict their performance in the future. Most of the existing GNN-based knowledge tracing model only considers the static pairwise relationship between concepts and exercises, but ignores the mining of edge features. Also, the dynamic and complex higher-order relationships hidden in the learners' answer sequence have not been fully exploited. In this paper, a novel hypergraph-based knowledge tracing model (HGKT) is proposed to address these limitations. Firstly, we exploit edge feature that indicates the frequency of exercise-concept's occurrence to extend the common bipartite graph. Then we use Node and Edge features based graph Neural Networks (NENN) to obtain the embedding representation of exercises and concepts. Secondly, a hypergraph with different weights on vertices is constructed during the learners' exercise-answering process and then it is transformed to a simple graph based on its similarity between hyperedges. Thereafter, we use the hypergraph neural networks (HGNN) and line hypergraph convolution network (LHCN) to obtain the learners' embedding and discover the higher-order relationships formed during this process. Thirdly, the difficulty of exercises and the average response time are utilized to improve the learning of LSTM's hidden states. Finally, all the embeddings are jointly added to the generalized interaction module of GIKT to draw attention to the useful information for prediction. Experiments demonstrate that the proposed HGKT outperforms previous classical methods in terms of AUC on the three widely used datasets.

In this paper, we study precise localization using Angle of Arrival (AOA) estimations by smart-antenna equipped beacons in Ad-Hoc networks. The node to be localized sends a signal to its surrounding beacons. The beacons estimate the signal directions with high resolution AOA methods and feed them back to the node for position calculation. In other words, position calculation is not required at beacons. When AOA estimates from three or more beacons are received, ambiguity occurs. Three resolution methods, namely (a) simple averaging (b) Minimax and (c) Precision-weighted averaging are proposed and compared. As estimation bias is heavily dependent on antenna orientations the center-facing approach is found to give better performance in a square field.