Nowadays, social media networks generate a tremendous amount of social information from their users. To understand people’s views and sentimental tendencies on a commodity or an event timely, it is necessary to conduct text sentiment analysis on the views expressed by users. For the microblog comment data, it is always mixed with long and short texts, which is relatively complex. Especially for long text data, it contains a lot of content, and the correlation between words is more complex than that in short text. To study the sentiment classification of these mixed texts composed of long-text and short-text, this research proposes an optimized GloVe-CNN-BiLSTM-based sentiment analysis model. In this model, GloVe is used to vectorize words, and CNN is given to represent part space character. BiLSTM is used to build temporal relationship. Twitter’s comment data on COVID-19 is used as an experimental dataset. The results of the experiments suggest that this method can effectually identify the sentimental tendency of users’ online comments, and the accuracy of sentiment classification on complete-text, long-text, and short-text can achieve to 0.9565, 0.9509, and 0.9560, respectively, which is obviously higher than other deep learning models. At the same time, experiments show that this method has good field expansion.

Sentiment classification has become a significant research topic in natural language processing. As the most popular research method for sentiment classification, deep learning has been applied to various experimental datasets by many scholars and has achieved good results. Aiming at the problems of poor effect and insufficient accuracy of sentiment classification in the current vertical field, and to solve sentiment classification on Chinese mixed text, including both long text and short text, this work proposes an improved BiLSTM-Attention model that can extract features more effectively. The problem of insufficient dependence on long text is resolved by the Bi-directional Long Short-Term Memory (BiLSTM) model, and important information in the text is obtained by the attention mechanism. This study uses online shopping comment datasets for experiments and applies a multiclassification evaluation index to evaluate the model. Experiments support the proposed approach, the accuracy of sentiment classification on mixed text, long text can achieve to 0.9280, 0.9358 respectively, and its practical effect are more advantageous than other sentiment classification methods in terms of classification performance. The experimental results show that this study makes an important contribution to business development and has good domain extensibility.

COVID-19 has produced significant fluctuations and impacts on the Chinese stock market, and the sentiment analysis of stock reviews is important for the study of economic recovery. Due to the lack of a large amount of labeled data in the existing Chinese stock review data, and the currently popular Bert model mostly failed to consider contextual information according to different contextual backgrounds when extracting features, resulting in the lack of contextual information in the modeled features. To address the above problems, this paper proposes an innovative Chinese stock review sentiment analysis model BERT-BiLSTM-Attention, which encodes the stock review text by BERT to enhance the semantic feature representation of the text, BiLSTM is then utilized to enhance the contextual information of the overall context of the review as well as the model’s comprehension of the text sequences, and then Attention mechanism is utilized to obtain important textual information and get the most effective information quickly. Experiments show that the model is effective in sentiment analysis of Chinese stock reviews, with an accuracy of 93.98%, It can be proved that the proposed model well enhances the performance of stock review text classification, and has a strong generalization ability, which can be used for sentiment analysis in many fields.

Forecasting future trajectories of intelligent agents presents a formidable challenge, necessitating the analysis of intricate scenarios and uncertainties arising from agent interactions. Consequently, it is judicious to contemplate the establishment of inter-agent relationships and the assimilation of contextual semantic information. In this manuscript, we introduce HTTNet, a comprehensive framework that spans three dimensions of information modeling: (1) the temporal dimension, where HTTNet employs a time encoder to articulate time sequences, comprehending the influences of past and future trajectories; (2) the social dimension, where the trajectory encoder facilitates the input of trajectories from multiple agents, thereby streamlining the modeling of interaction information among intelligent agents; (3) the contextual dimension, where the TF-map encoder integrates semantic scene input, amplifying HTTNet cognitive grasp of scene information. Furthermore, HTTNet integrates a hybrid modeling paradigm featuring CNN and transformer, transmuting map scenes into feature information for the transformer. Qualitative and quantitative analyses on the nuScenes and interaction datasets highlight the exceptional performance of HTTNet, achieving 1.03 minADE10 and a 0.31 miss rate on nuScenes, underscoring its effectiveness in multi-agent trajectory prediction in complex scenarios.

NoSQL databases are vital for modern big data applications, yet traditional teaching methods struggle with lagging content, insufficient practice, and low student engagement. To address these issues, this paper proposes the WWH-integrated teaching method “Why learn, What learn, How learn” for a NoSQL database course. WWH combines three core approaches: the general–special method, which structures knowledge from foundational concepts to specialized technologies; the comparative method, which contextualizes NoSQL value via real-scenario analysis; and the theory–practice combination method, which links concepts to hands-on tasks, supplemented by the problem-guidance and key-highlighting strategies. A quasi-experiment with two cohorts (80 students each; 2023 cohort as control, 2024 as experimental) validated WWH. Quantitative results showed significant improvements: theoretical exam scores rose by 9.2 points (t(158) = 9.21, p < 0.001) and experimental scores by 10.3 points (t(158) = 7.92, p < 0.001), and classroom discussion rates increased from 45.2% to 82.7% (χ²(1) = 28.90, p < 0.001). Qualitative analysis of student essays and project reports further confirmed deeper conceptual understanding, stronger tradeoff awareness, and enhanced knowledge integration in the experimental cohort. This study provides an evidence-based, student-centered framework for modernizing NoSQL instruction, better preparing students for industry data management needs.

Various word game software is becoming more and more popular, such as the recently popular “Wordle” crossword game, which can entertain, develop intelligence, and improve word learning ability. However, there are little research on how to improve the challenge and innovation of word games. For this challenge, this paper focuses on the research of word games based on SIRS-ARIMA model and machine learning algorithm. The SIRS-ARIMA model is an innovative approach that combines the SIRS propagation model and the autoregressive integrated moving average model (ARIMA) to analyze and predict dynamic changes in the word game by taking into account factors such as social media propagation. This paper also uses the entropy method of machine learning algorithm and SVC model to classify the difficulty of words, so as to optimize the design and play of word games. By analyzing player behavior and word attributes, it can personalize the game experience and provide players with precise feedback mechanisms. This research provides new theories and methods for the development of word games and provides strong support for the design of more engaging and innovative games.