Feng Qiu

Sentiment analysis is significant for exploiting countless opinion-rich data from social media. However, it faces known challenges, such as informal expression and data sparsity. While large language models (LLMs) excel at dynamic contextual disambiguation, such deep learning models neglect to leverage prior knowledge calculated statistically and globally based on dataset-wide category tendencies to guide the encoding process. To overcome these challenges, this study proposes a pre-attention framework that includes a modified gated recurrent unit hierarchically integrated with a dual-level pre-attention mechanism (PA-GRU) and an LLM-PA-GRU model equipped with a strategic LLM fine-tuning method. The pre-attention mechanism extracts inter-category and intra-text statistical priors from the training set. Trainable coefficient matrices are proposed to adaptively fuse these priors within a unified formulation, enabling flexible allocation of global and local statistical signals. The proposed PA-GRU cell features a modified gating structure and a computation flow that integrates these extracted priors as explicit statistical anchors during sequence encoding. Moreover, we implement an LLM-PA-GRU model that connects the LLM’s deep feature representations to the PA-GRU. This strategic LLM fine-tuning approach is intended to mitigate gradient instability frequently encountered in models with heterogeneous architectures. Finally, a prototype-based alignment loss is employed to enforce feature consistency across modules. Extensive experiments on benchmark datasets demonstrate that the proposed approach achieves competitive performance compared with recent literature-reported models.

Short message services (SMS), microblogging tools, instant message apps, and commercial websites produce numerous short text messages every day. These short text messages are usually guaranteed to reach mass audience with low cost. Spammers take advantage of short texts by sending bulk malicious or unwanted messages. Short texts are difficult to classify because of their shortness, sparsity, rapidness, and informal writing. The effectiveness of the hidden Markov model (HMM) for short text classification has been illustrated in our previous study. However, the HMM has limited capability to handle new words, which are mostly generated by informal writing. In this paper, a hybrid model is proposed to address the informal writing issue by weighting new words for fast short text filtering with high accuracy. The hybrid model consists of an artificial neural network (ANN) and an HMM, which are used for new word weighting and spam filtering, respectively. The weight of a new word is calculated based on the weights of its neighbor, along with the spam and ham (i.e., not spam) probabilities of short text message predicted by the ANN. Performance evaluations on benchmark datasets, including the SMS message data maintained by University of California, Irvine; the movie reviews, and the customer reviews are conducted. The hybrid model operates at a significantly higher speed than deep learning models. The experiment results show that the proposed hybrid model outperforms other prominent machine learning algorithms, achieving a good balance between filtering throughput and accuracy.

This paper presents a new method for accurately measuring the vocabulary size of second language (L2) learners. Traditional vocabulary size tests (VSTs) are limited in capturing a tester’s vocabulary and are often population-specific. To overcome these issues, we propose an intelligent vocabulary size measurement method that utilizes massive robot testers. They are equipped with randomized and word-frequency-based vocabularies to simulate L2 learners’ variant vocabularies. An intelligent vocabulary size test (IVST) is developed to precisely measure vocabulary size for any population. The robot testers “take” the IVST, which dynamically generates quizzes with varying levels of difficulty adapted to the estimated tester’s vocabulary size in real-time using an artificial neural network (ANN) through iterative learning. The effectiveness of the IVST is factually verified by their visible vocabularies. Additionally, we apply a long short-term memory (LSTM) model to further enhance the method’s performance. The proposed method has demonstrated high reliability and effectiveness, achieving accuracies of 98.47% for the IVST and 99.87% for the IVST with LSTM. This novel approach provides a more precise and reliable method for measuring vocabulary size in L2 learners compared to traditional VSTs, offering potential benefits to language learners and educators.

Medical computed tomography (CT), nuclear magnetic resonance imaging (MRI), and other imaging facility produce a large amount of medical image data which has great diagnosis value. Traditional three‐dimensional reconstruction of medical images has high requirements for graphics acceleration hardware and the processing speed. In this study, VxWorks embedded real‐time process feature is used for CT or MRI DICOM data to real restoration and establishment of virtual three‐dimensional model for realizing volume reconstruction, maximum density projection, multiplane reconstruction, dynamic interactive cutting of any surface, dynamic display of three‐dimensional model and two‐dimensional sectional image, surgical path planning and interactive surgical simulation, and determine the best surgical scheme. The practical application shows that the virtual simulation environment supports the seamless transplantation of code, function debugging, and interaction and solves the issue of high requirements for hardware. It can meet the needs of scientific research and teaching for clinicians and medical imaging workers and can be widely used in the training of virtual surgical anatomy for medical students.

G-quadruplexes have drawn widespread attention for serving as a potential anti-cancer target and their application in material science. Molecular dynamics (MD) simulation is the key theoretical tool in the study of GQ’s structure-function relationship. In this article, we systematically benchmarked the five force fields of parmbsc0, parmbsc1, OL15, AMOEBA, and Drude2017 on the MD simulation of G-quadruplex from four aspects: structural stability, central ion channel stability, description of Hoogsteen hydrogen bond network, and description of the main chain dihedral angle. The results show that the overall performance of the Drude force field is the best. Although there may be a certain over-polarization effect, it is still the best choice for the MD simulation of G-quadruplexes.