American literature has long served as a mirror, reflecting the diverse cultural, social, and political landscapes of the United States. This research investigates the representation of social groups in American literature by employing advanced natural language processing techniques. Specifically, it utilizes contextualized word embedding models to analyze how characters from diverse social identities, particularly in terms of gender, race, and class, are portrayed across a curated corpus of canonical and contemporary American literary texts. The dataset is compiled and preprocessed through tokenization and normalization to prepare the texts for contextual embedding extraction and bias analysis. Bias detection is conducted using a Bidirectional Encoder Representations mutated Weighted Support Vector Machine (BERWSVM) model designed to classify complex social representations. The Contextualized Embedding Association Test (CEAT) isemployed to statistically evaluate the strength of association between social groups and character traits by computing cosine distances between contextual embeddings. Bidirectional Encoder Representations from Transformers (BERT) are used to extract rich semantic representations from the texts, capturing character descriptions, group identity references, and associated traits. The WSVM component classified intersectional group embeddings, enabling the assessment of representational patterns that extend beyond single-identity categorizations. Implemented in Python, the findings show that the BERWSVM approach performs better than multimodal baseline architectures, achieving superior results, with accuracy, F1-score, recall, and precision ranging from 90% to 95%. The findings reveal that the BERWSVM achieved high accuracy in distinguishing characters belonging to intersectional groups, significantly outperforming traditional baseline models. It shows the effectiveness of integrating computational bias detection algorithms with literary interpretation in analyzing social ideologies, representation, diversity, and fairness in narrative structures.

The current English writing quality evaluation methods limit applicability and relatively low scoring accuracy. A quality evaluation model for English writing based on improved K-means clustering and machine learning is proposed to address the existing issues. The study first uses a continuous bag of words model to extract text vectors from the training set, and then applies an improved K-means clustering algorithm (KMCA) for text feature extraction. Afterward, the clustered text features are combined with deep learning features and scored using support vector machines to obtain annotated essay collections. The study applies this essay collection to the training of a writing quality evaluation model based on convolutional neural networks and long short-term memory networks (LSTM). The experiment outcomes indicate that the proposed English writing quality evaluation model has a fitting degree of over 0.9 with the evaluation results of professional teachers, and its recall rate and F1 score are 0.94 and 0.92, respectively. The accuracy of its evaluation results reaches 94.10%. Under different theme prompts, the Kappa coefficient value of the model reaches 0.81. The proposed writing quality evaluation model can achieve high-precision composition evaluation and provide reliable data feedback for students’ English learning.

With the development of art, people’s demand for visual experience is also getting higher and higher. In visual communication design, how to improve the visual effect of design works has become an important issue. In recent years, the research on visual communication design has been paid more and more attention. However, these methods suffer from the problem of low output accuracy and low time efficiency. Binocular vision refers to the principle and method of simulating human binocular vision. It can provide richer depth visual information. Therefore, this paper proposes a binocular perception model based on multi-module joint optimization and applies it to visual communication design. The designed model includes image preprocessing module, binocular processing module, optimization module, and perception module, thus improving the effect and quality of visual communication design. The experimental results show that the algorithm proposed in this paper improves the output accuracy by 9.35% and the time efficiency by 10.24% compared with other algorithms. The proposed binocular perception model based on multi-module joint optimization can be used to optimize the depth perception effect of design works. Through 3D reconstruction and depth perception of design works, the visual design scheme can more accurately grasp the spatial structure and layout of design works, thereby improving visual effects and user experience.

With the rapid development of information technology, artificial intelligence has entered every aspect of life, and the intelligence in the judicial field is an inevitable trend for future development. Therefore, this paper utilizes deep learning text classification technology to achieve crime prediction and legal recommendation tasks. The theoretical part elaborates on the structural principles and attention mechanism of the BiGRU model, and provides the detailed mathematical definition of the model in this paper. The experimental part is divided into four parts: data preprocessing, experimental model design, indicator evaluation, and result analysis. A BiGRU + Attention model based on BERT word embedding is proposed through the GRU and attention mechanism. Experiments are designed on the CAIL2018 small dataset and compared them with baseline methods. In the crime prediction task and legal recommendation task, its comprehensive F union value (calculated from the F1 micro-mean and F1 macro-mean, with the formula F union = ( F mi + F ma )/2) reached 85.05% and 80.14%, respectively. The results indicate that this method can effectively improve the semantic accuracy of criminal fact encoding. This research provides strong support for the intelligent development of the judicial field, improves the efficiency of judicial work, and helps judges make more accurate judgments and decisions.

At present, there are still weak links in the development process of sports tourism resources, which hinders its own development. The prediction of tourism resources development potential plays an important role in guiding tourism projects. Before the study, this paper first investigates the present situation of sports tourism resources, from the perspective of spatial analysis of regional distribution of sports tourism in China. Guided by the concept of sustainable development and regional economic theory of sports tourism resources development potential influencing factors analysis; Identify the main body of resource development and construct the work flow of sports tourism resource development. Based on the basic principles of genetic algorithm and BP neural network theory, the genetic algorithm is used to optimize the weight and threshold of BP neural network. Seven influencing factors, including Sports Policy Support Index, Traffic Access Index, Sports Tourism Eco-Environment Index, Regional Sports Online Tourism Development Index, Cultural and Historical Resources Index, Cultural and Tourism Industry Influence Index, and Characteristic Industry Output Value Scale Index, are selected as the nodes of the sports tourism resource development prediction model. Then, the sample data is sorted out and normalized. The hybrid genetic neural network is used for prediction and analysis. The results show that: This paper uses genetic algorithm to optimize the weight and threshold of BP neural network, It can effectively avoid BP neural network falling into local minimum and improve the classification processing ability of BP neural network, The genetic neural network prediction model constructed on this basis can accurately classify the development types of four sports tourism resources in the training samples, It provides a multi-dimensional objective decision-making reference method for the development and prediction of sports tourism resources in the future.

Real-time collaborative music creation requires dynamic systems that can understand musical sequences, tonal structure, and rhythmic flow in live contexts. Conventional sequence generation methods using static deep learning models often struggle with adapting to changing musical input. The limited use of audio features and non-optimized architectures leads to reduced fluency and stylistic coherence in generated improvisations. To enable adaptive and context-sensitive real-time music improvisation that responds fluidly to symbolic and audio-based inputs. A Biogeography-Based optimizer-driven stacked Long Short-Term Memory (BB-Stacked LSTM) is introduced, combining evolutionary optimization with temporal deep learning to improve improvisation quality and model adaptability. The BB-Stacked LSTM system uses evolutionary principles to optimize sequence modeling parameters, enhancing both accuracy and expressiveness in music generation. Performance-oriented datasets featuring paired audio and symbolic data are utilized, including genres such as jazz and classical. One-hot encoding is used for symbolic note sequences. Sequence smoothing is achieved through a Hidden Markov Model. Time-aligned symbolic and audio data are structured for temporal modeling. Mel-Frequency Cepstral Coefficients (MFCC) are extracted from audio to capture spectral and timbral properties. The Stacked LSTM learns sequence progression, while BBO tunes architectural parameters, including layer depth, unit count, and learning rate to maximize musical coherence. Generated sequences exhibit an improved armony score of 4.6. The BB-Stacked LSTM approach enhances real-time music generation by integrating evolutionary optimization with deep temporal modeling.

With the deepening of economic and cultural globalization and the popularity of cross-cultural communication, Mandarin, as a key carrier of Chinese culture, has become increasingly important for both domestic language education and foreign Chinese learning. However, traditional Mandarin teaching faces limitations such as difficulty in real-time detection of individual reading errors (e.g., missing reading and back reading) and heavy reliance on teacher experience, which restricts the efficiency of error correction and teaching quality. Meanwhile, with the rapid development of information technology, deep learning has shown strong advantages in speech signal processing, providing a new technical path for intelligent Mandarin reading error detection. Against this background, this study focuses on the detection of missing reading and back reading errors in Mandarin reading aloud, and conducts research based on the deep learning framework. To improve the accuracy and efficiency of error detection, this study takes the traditional Deep Neural Network (DNN) as the basic model, and optimizes the core Reading Quality Assessment (GOP) algorithm: first, it extends the GOP algorithm to the DNN-based error detection system, and modifies the GOP calculation formula by introducing the average posterior probability of non-target senones and weight coefficients, which solves the problem of unreliable phoneme segmentation caused by non-standard pronunciation; second, it addresses the issue that missing-reading errors of the current phoneme affect the GOP calculation of adjacent phonemes in the traditional framework, further optimizing the algorithm’s robustness. Additionally, this study introduces DNN adaptive technology based on KL divergence regularization to align the standard and non-standard reading models, enhancing the algorithm’s adaptability to different speakers. Experiments are conducted on two databases (MPE database for domestic Mandarin speakers and ICALL database for foreign Chinese learners). The results show that the improved GOP algorithm combined with DNN adaptive technology significantly outperforms traditional methods: compared with the GMM-CM algorithm, the accuracy and recall of error detection are increased by 13.4%; compared with the original DNN-GOP algorithm, the improved DNN-GOP2 algorithm reduces the Top1 error rate by 1.7% and the Top5 error rate by 2.0%. This study not only provides a more accurate and efficient technical solution for Mandarin reading error detection but also lays a foundation for the development of intelligent Mandarin teaching systems, which is of great significance for promoting the modernization of Mandarin teaching and the popularization of Chinese language education globally.

As universities promote the deep integration of ideological and political education with the Production-Oriented Approach (POA), accurately identifying the emotional tendencies in student feedback has become a key challenge in improving teaching quality. Ideological and political emotions differ from general emotional expressions in that they involve deeper psychological activities such as value guidance and cultural identity, and are characterized by implicitness, complexity, and contextual dependence. Existing research has largely focused on teaching model design, but lacks sentiment analysis of ideological and political education feedback text, particularly in the areas of semantic understanding and weighting of key sentiment terms. To address this, this paper proposes a sentiment analysis model that integrates Bidirectional Long Short-Term Memory (BiLSTM) with an attention mechanism to accurately discriminate sentiment in ideological and political education feedback for POA college English courses. Through text preprocessing, BiLSTM bidirectionally captures contextual semantic dependencies, and an additive attention mechanism dynamically weights key sentiment terms, the model ultimately achieves positive, neutral, and negative sentiment discrimination using Softmax. Experimental results show that the model achieves an accuracy of 93.7% and an F1 score of 0.93, both outperforming baseline models and demonstrating good robustness and interpretability. This research provides an efficient, lightweight, and interpretable technical solution for sentiment analysis of ideological and political education feedback.

In recent years, safety issues in intelligent driving have occurred frequently. Deep learning technology, through continuous learning, has provided new ideas for its development. The intelligent driving ecosystem is supported by deep learning technology, covering the whole process of data acquisition, processing, and driving safety prevention, effectively supporting the iterative upgrading of the technology. This paper employs the methods of evolutionary games and supernetworks to systematically examine different behavioral analyses and conduct numerical simulation verification The results show that: (1) The higher the initial willingness of intelligent driving enterprises to participate, the more significant the positive facilitation effect is exerted by deep learning technology. (2) The government’s policy of directing scientific research institutions to support intelligent driving enterprises with deep learning has not had a positive effect but instead may have disrupted or destroyed the existing market landscape. (3) There is an interaction effect between the strategies of each subject, and deep learning technology, as an important tool, can effectively guide the intelligent driving ecosystem to improve driving safety. This paper provides an essential methodological reference for exploring the integration path of deep learning technology and the intelligent driving ecosystem to effectively improve driving safety.

Skill-job alignment remains a significant challenge in vocational education, where students often graduate with limited guidance on career pathways matching their acquired competencies. The gap between vocational training outcomes and dynamic labor market demands underscores the need for intelligent systems that facilitate personalized employment recommendations. To address this, a Machine Learning-Powered Skill-Job Matching Recommendation System is proposed for vocational graduates, integrating a novel hybrid model named Scalable Slime Mould optimized-Adaptive Random Forest Tree (SSM-ARFT). The model utilizes a curated dataset combining vocational student profiles, academic performance, certifications, and job requirement metadata gathered from institutional databases and public employment platforms. Data preprocessing is performed using normalization techniques to ensure uniformity across varying data types. For feature extraction, Principal Component Analysis (PCA) is employed to identify the most influential attributes for job-role alignment. The core framework involves mapping students’ skill profiles to job attributes through multi-level filtering and learning stages. The proposed SSM-ARFT algorithm enhances the ARFT by introducing slime mold-inspired metaheuristics for dynamic feature selection and adaptive learning, ensuring robustness and scalability across varying datasets. This intelligent recommendation system helps guide vocational students toward employment options that are closely aligned with their competencies. The proposed method is implemented by using Python 3.10.1. The models were trained and tested with k-fold cross-validation data. The findings determine that the suggested model outperforms traditional methods in metrics such as precision, F1 score, recall, and accuracy, which range from 91% to 94%. The research concludes that the proposed model offers a practical, scalable solution for effective skill-job matching, thereby enhancing graduate employability in vocational sectors.