Research on the Popularization of Marxism by Big Data Based on Attention Mechanism

Marxism plays an important guiding role in China, and it is a dominant position in the ideological field of China. To let people know more about Marxism and firmly follow this road in China, the popularization of Marxism is an important way. Big data is on the rise day by day and widely used in all aspects of social practice. As a result, the popularization of Marxism based on big data technology has a good development environment and opportunities [1, 2].

With the advent of the era of big data, traditional media, such as newspapers, radio and television, etc., are facing elimination. Because of their lag, one-way and poor communication effect, combined with the current development trend, the situation is rather difficult. Then, under the background of big data, various emerging media are constantly updated, and advanced communication methods, resource sharing and integration all show a kind of information diversification. Among them, short video is the most popular mainstream way to obtain information. In recent years, the way of dissemination of Marxist theory has also changed greatly. According to a series of reports on China's Internet development, influencing works and new media are the main forces of its dissemination. The vividness and visibility of the new media's own communication form make Marxist theory present to the broad masses of the people in a more amiable form, visualizing the abstract, complicated and difficult knowledge in the past, improving its inherent impression and greatly improving people's understanding ability. Therefore, we can think that the development of big data has promoted the popularization of Marxism [3, 4].

If you want to know more about the spread effect of the popularization of Marxism, you basically rely on sampling survey. This quantitative research method has a large workload of tasks, and it is difficult to ensure the accuracy of the effect conclusion formed on this basis [5]. Up to now, big data technology has provided operability for accurately evaluating the spread effect of Marxism popularization. Big data technology collects, stores and processes relevant information about the popularization of Marxism for quantitative analysis, builds a feedback mechanism through online communication and questionnaire survey, and then provides data guarantee for qualitative research on the basis of quantitative research on these massive data, thus making the evaluation of communication effect more convenient and accurate. Therefore, by using big data technology to evaluate the communication effect of Marxism popularization, we can fully grasp the communication effect presented in an intuitive and accurate way [6].

The popularization construction and development of Marxism in China are influenced by many factors. In the process of comprehensively promoting the popularization of Marxism, according to the actual needs of ideological construction, accurately positioning the specific factors that affect the popularization of Marxism can improve the construction effect and promote the full play of the advantages of construction to the maximum extent [7-9]. Therefore, we should actively explore the effective application of big data technology in our work practice, and try to reform the basic thinking mode of Marxist popularization construction. On the basis of not giving up the research of causality, we should incorporate relevant research into the practice of Marxist popularization construction to ensure that with the support of data information, we can examine the contents and things related to Marxist popularization construction from a new perspective, so as to judge the relevance of specific factors affecting the popularization construction. In this way, through the application of big data technology, on the basis of scientifically predicting the development needs of popularization construction and guiding different influencing factors, we can ensure that they can jointly assist and promote the practice of popularization construction and development of Marxism, provide good support for the real popularization of Marxist theory, and promote Marxism to maintain a benign development state in the process of exploring popularization.

2

Analysis of Marxist communication of attention mechanism

2.1

Attention mechanism

Attention mechanism is to express a form of attention in an abnormal active situation. No matter what its form, you can attract attention to anything. Resource allocation system is to allocate computing resources to solve the problem of information overload. Top-down attention mechanism is also called focused attention mechanism, which is a type of attention. Focused attention is purposeful, project-dependent attention and active and conscious attention to the object. One is unconscious upward attention, called inevitability.

The earliest attention mechanism is applied to the field of machine translation. This mechanism is based on the Encoder-Decoder framework, in which the input sequence is represented as ^{X = {x₁, x₂, x_m}} and the output sequence is represented as ^{Y = {y₁, y₂, y_m}}. The input sequence is transformed into semantic coding C through the nonlinear transformation of the coding module, and then It is input to the decoding module and transformed into an output sequence Y. The specific calculation formula is as follows: (1) $c = F (x_{1}, x_{2}, \dots, x_{m})$ (2) $y_{i} = G (C, y_{1}, y_{2}, \dots, y_{i - 1})$

In this framework, attention mechanism is introduced to solve the problem that decoding accuracy decreases due to the increase of sequence length. In the process of coding and decoding, hidden vector sequence hi (i = 1, 2, …, n) is added, which can make the model selectively screen subsets for decoding with the help of the information transmitted by the input sequence. The calculation formula is as follows: (3) $C_{i} = \sum_{i = 1}^{n} α_{i j} h_{i}$

Suppose the query is Q, the data source is S, the keyword is K, the weight is V, the attention value is A, and the sequence is in the form of key-value pairs (K, V). In order to keep each element in the context of the source code, the Q engine uses a set of key values. For the mapping function of (K, V), the attention score e_i is obtained. The higher the similarity, the more important an element is to the query target and the larger the e_i value; Then use the softmax function to normalize the attention score e_i to highlight the weights of important elements. The specific calculation formula is as follows: (4) $e_{t i} = F (q_{t}, k_{i})$ (5) $α t i = s o f t \max (e_{t i}) = \frac{\exp (e_{t i})}{\sum_{i = 1}^{N} \exp (e_{t i})}$ (6) $A ((K, V), q_{t}) = \sum_{i = 1}^{N} α_{t i} v_{i}$

2.1.1

AER attention mechanism

In order to screen out effective characteristic information from various Marx evaluation effects, attention mechanism is introduced into the evaluation of Marx's popularization development effect. Introducing AER-Net is an ultra-lightweight attention module, whose main role is to consider each channel to interact cross-channel information with its neighboring channels and generate weight values. Increasing the weight of key features of the index and reducing the weight of irrelevant features make the model pay attention to useful information first, which has an excellent effect on improving the recognition accuracy. The AER-Net structure diagram is shown in Figure 1.

Assume that the input feature map of the network is defined.

(7)

F \in R^{H \times W}

First, a 1 × 1 × C global description feature is obtained. Secondly, where k represents the coverage of cross-channel interactions, i.e. how many phase nearest neighbors are involved in the prediction of a channel, and the sigmoid activation function is used to create the weight ratio of each channel.

ECA-Net only considers communication between each channel and its neighbors, which reduces model complexity and improves performance compared to communication on all channels. A correlation formula is generated for each channel using a unidirectional rotation of size k, i.e.: (8) $W = σ (C 1 D_{K} (y))$ (9) $k = {| \frac{\log_{2} (C)}{γ} + \frac{b}{γ} |}_{o d d}$

Where |t|_odd is the odd number closest to t, and the values of γ and b are set to 2 and 1, respectively

2.1.2

Contextual attention mechanism

The convolution features extracted by HybridNet to the I_x-th layer are represented by X_|x , (10) representing a set of C-dimensional vectors at W × H positions in the feature map. Firstly, three convolutional filters of different sizes are used to extract context information from the extracted feature maps, with the sizes of 3 × 3, 5 × 5 and 7 × 7 respectively. Each filter generates a feature map. For example, there are 32 convolutional filters of 3 × 3 size, which generates 32 W × H feature maps.

X_|x ∈ R^W×H×C (10) is formed by connecting all feature maps in series in that direction of the channel

To calculate the contextual attention mask, CNN parameterization is performed on X_c, using a single-channel output convolutional layer behind X_c: (11) $T = f (K \times X_{c} + b)$

Where: x represents a convolution operation; K represents a 3 × 3 convolution filter; b is a bias parameter; f is the ReLU nonlinear activation function; The size of T is W × H, which summarizes the output information of all channels in X_c. Then use a softmax operation to get the attention mask: (12) $E (i) = \frac{\exp (T (i))}{\sum 1' \in L}$

Where: the size of attention mask E(i) is W × H, which represents the attention response values at different positions of the convolutional feature map. If the features at a certain position are significant, there is a higher response value in E(i).

3

Multi-layer attention fusion

This article extracts the features of Conv3, Conv4, and Conv5 in the HybridNet network that have rich semantic information and better recognition performance in the HybridNet network, and applies the contextual attention mechanism to the three-layer convolutional feature map to obtain their respective attention masks E1, E2, E3. Thereafter, the weighted sum G of E1, E2, and E3 is obtained through a 1 × 1 × 3 convolution layer: (13) $G = \sum_{k = 1}^{3} W_{k} E_{k} + b$

Where: wk and b represent the convolution weight and bias term respectively, and G is a mask with multiple layers of attention of size W × H. Finally, the Conv5 convolution feature is modulated by: (14) $F (p) = G^{°} X_{15} (p)$

Where: p represents the p-th characteristic channel; A Hadamard matrix product representing the channel direction; F denotes the feature map of the convolutional feature Xl5 extracted by Conv5 after being processed by the attention mechanism.

3.1

Multi-classification training

In the training of attention network, the actual index parameters are obtained by multi-classification training process. In order to calculate effectively, the input indicators are processed by HybridNet and attention mechanism, processed by a 3 × 3 size 64-channel convolution layer, then passed through a fully connected layer, and finally input the softmax layer: (15) $S_{j} = \frac{\exp (A_{j})}{\sum_{k \in N} \exp (A_{k})}$

Where: A is the output vector of the fully connected layer, the size is N × 1, and N is the classification number; S is a vector of N × 1, representing the output result of the softmax layer, where S_j represents the probability that the input index belongs to the j-th category, and the value is between 0 and 1. The loss function is: (16) $L = - \frac{1}{m} \sum_{i = 1}^{m} (\sum_{j = 1}^{N} P_{j} \log S_{j})$

Where: m is the number of social indicators of a training batch; P is a 1 × N vector, and all the elements in this vector are 0 except the value of the real category label corresponding to the input index is 1.

3.2

Similarity calculation

This object uses the variance of the cosine to calculate the similarity of two different measurements. The formula for cosine of distance is as follows.

(17)

d = \frac{F_{Q} F_{R}}{| | F_{Q} | | | | F_{R} | |}

Where: F_Q and F_R represent the feature vectors of the query index and the reference index respectively, the numerator is the dot multiplication of the two vectors, and the denominator represents the product of the lengths of the two vectors.

In the experiment, the cosine distance between the i-th index Mi in the query index data set and each index in the reference index data set is calculated, and the one with the largest value is judged to match the actual social situation.

3.3

Deep residual attention model

When training the attention model, the accuracy rate often decreases. Therefore, a new structural depth residual network that deepens the number of layers of the network as much as possible is proposed, and a jump connection line that skips one or more layers is introduced, as shown in Figure 3. The residual unit structure is shown.

Where x and y represent the input and output of the L-th residual unit respectively, F is the residual function, which represents the residual learned by this residual unit, the activation function relu is represented by σ, and the residual learned in Figure 5 is: (18) $F (x) = W_{2} σ (W_{1 X})$ (19) $m_{k} = \sum_{i = 1}^{N} x_{i} / N$

H is the identity mapping, that is, H (x) = x, calculate the output y of the L-th residual unit: (20) $Y = σ (F (x) + H (x)) = σ (W_{2} σ (W_{1 X}) + x)$ (21) $u = \sqrt{u_{A 1} (σ_{1}) u_{A 2} (σ_{2}) \geq \overset{*}{u}}$

When ordinary networks transmit information, there is a problem of partial information loss, which protects the integrity of transmitted information. ResNet contains a series of network models with different layers, and the deep residual networks with different layers are shown in Table 1.

Table 1.

Depth residual model structure with different layers

lay	out sizes	18-lay	34-lay	50-lay	101-lay	152-lay
convl	112x112	7x7,64,stride 2
		3x3 max pool, stride 2
conv2_x	56x56	$[\begin{array}{l} 3 \times 3, 64 \\ 3 x 3, 64 \end{array}] \times 2$	$[\begin{array}{l} 3 \times 3, 64 \\ 3 \times 3, 64 \end{array}] \times 3$	$[\begin{matrix} 1 x 1, 64 \\ 3 x 3, 64 \\ 1 x 1, 256 \end{matrix}] x 3$	$[\begin{matrix} 1 \times 1, 64 \\ 3 \times 3, 64 \\ 3 \times 3, 256 \end{matrix}] \times 3$	$[\begin{matrix} 3 \times 3, 64 \\ 3 \times 3, 64 \\ 3 \times 3, 256 \end{matrix}] x 3$
conv3_x	28x28	$[\begin{matrix} 3 \times 3, 128 \\ 3 \times 3, 128 \end{matrix}] \times 2$	$[\begin{array}{l} 3 \times 3, 128 \\ 3 \times 3, 128 \end{array}] \times 4$	$[\begin{array}{l} 1 \times 1, 128 \\ 3 \times 3, 128 \\ 1 \times 1, 512 \end{array}] \times 4$	$[\begin{array}{l} 1 \times 1, 128 \\ 3 \times 3, 128 \\ 1 \times 1, 512 \end{array}] x 4$	$[\begin{array}{l} 1 x 1, 128 \\ 3 x 3, 128 \\ 1 x 1, 512 \end{array}] x 8$
conv4_x	14x14	$[\begin{array}{l} 3 x 3, 256 \\ 3 x 3, 256 \end{array}] x 2$	$[\begin{array}{l} 3 x 3, 256 \\ 3 x 3, 256 \end{array}] x 6$	$[\begin{matrix} 1 x 1, 256 \\ 3 x 3, 256 \\ 1 x 1, 1024 \end{matrix}] x 6$	$[\begin{matrix} 1 x 1, 256 \\ 3 x 3, 256 \\ 1 x 1, 1024 \end{matrix}] x 23$	$[\begin{matrix} 1 \times 1, 256 \\ 3 \times 3, 256 \\ 1 x 1, 1024 \end{matrix}] x 36$
conv5_x	7x7	$[\begin{array}{l} 3 \times 3, 512 \\ 3 \times 3, 512 \end{array}] \times 2$	$[\begin{array}{l} 3 x 3, 512 \\ 3 x 3, 512 \end{array}] x 3$	$[\begin{matrix} 1 x 1, 512 \\ 3 x 3, 512 \\ 1 x 1, 2048 \end{matrix}] x 3$	$[\begin{matrix} 1 \times 1, 512 \\ 3 \times 3, 512 \\ 1 \times 1, 2048 \end{matrix}] x 3$	$[\begin{matrix} 1 \times 1, 512 \\ 3 x 3, 512 \\ 1 \times 1, 2048 \end{matrix}] x 3$
	1x1	average pool,1000-d fc, softmax
FLOPs		1.8x10⁹	3.6x10⁹	3.8x10⁹	7.6x10⁹	11.3x10⁹

The deep residual network ResNet50 is used. As can be seen from Figure 1, ResNet50 has a total of 50 layers, including 49 convolutional layers and 1 fully connected layer. Conv1 is the convolutional layer, and Convi _ x (i = 2, 3, 4, 5) is the residual block. As the structure of the residual block in Conv2 _ x is shown in Figure 1 below, it contains 3 convolutional layers, × 3 represents 3 residual blocks, other residual blocks are similar.

4

Innovative suggestions and development promotion of Marx's popularization in the era of big data

The popularization of Marxism has new development opportunities under the background of big data. Under the background of the new era, in order to explore a rational development path, it is necessary to systematically analyze the development of Marxist popularization under the background of big data era, provide corresponding guidance for the systematic innovation of various communication work, and enhance the overall development efficiency of Marxist popularization communication work[9,10].

4.1

Create a precise and popular communication mode of Marxism

In short, the popularization of Marxism means that through the spread and popularization of Marxism, more people can understand, recognize, accept and practice Marxism. If any theory or idea wants to be accepted, recognized and have greater influence by more people, it needs to rely on precise media and methods to form an all-round and multi-level communication system. Big data technology can be used to accurately capture and obtain the habits of the audience of Marxist theory, make personalized push according to personal characteristics, and strive to deeply integrate classic literature, hot comments on current affairs and other information, so as to present the basic content of Marxism in a form that the audience likes to see and hear[11,12]. Second, establish a benign interaction between communicators and audiences. Based on the interconnection of big data and the dynamic trend of the audience in the data platform, coupled with the application and development of new media technology, changing Marxism from a single word indoctrination to a comprehensive output integrating words, images, sounds, films and television can optimize the communication content and enhance the interactivity and effectiveness of the popular communication of Marxism on the premise of understanding the needs of the audience. In short, it is necessary to make full use of big data technology, fully grasp the ideological situation of the audience through the collection, collation and mining of data information, and form a benign interactive situation of communication-feedback-re-communication-re-feedback.

4.2

Improve the big data supervision system and optimize the network ecological environment

Under the background of big data era, problems such as omissions in network supervision and crisis of trust in humanistic environment have followed one after another, which provides a convenient channel for hostile forces to infiltrate socialist ideology. Virtual netizens without real-name registration system wantonly publish and spread inappropriate remarks on the Internet, such as some "non-horse" people misleading the masses under the banner of propagating and innovating Marxist theory, all of which seriously endanger the guiding power of China's mainstream ideology with Marxist theory. Therefore, building a big data supervision system is a necessary move to optimize the network ecological environment[13,14]. First, we must strengthen supervision and punishment mechanisms. For the behavior of publishing and disseminating remarks that endanger China's ideology on the online platform. The second is to improve the legal system of public communication. On the basis of striving for timeliness value and commercial value, some new media use too vulgary, entertaining or even distorted words, which greatly reduces the scientificity and rigor of Marxist theory. Relevant departments should improve the legal system of public communication, and third, establish an ecological management mechanism of network communication. For the communication subjects and receiving subjects existing in the Internet today, it is necessary to implement a strict real-name authentication account system, formulate a network information audit and release system and a network crime punishment system, and educate, warn and even impose legal sanctions on remarks that pollute Marxism[15,16].

4.3

Improve the application ability of big data and do a good job in talent training

Whether you have the awareness of big data application and whether you can make good use of big data technology is directly related to whether the popularization of Marxism can achieve practical results[17,18]. On the one hand, the main body of Marxist communication should actively adapt to the overall deployment of ideological work in the new era. In theory, we should always study the important works of classic Marxist writers and constantly improve our Marxist theoretical literacy; In terms of technology, we should actively learn and apply big data analysis technology, improve professional skills such as big data application, development, storage and analysis, and at the same time be familiar with the skills of using various communication technologies[19]. In terms of propaganda, on the basis of mastering Marxist theory and media technology, we should learn all kinds of effective and legal propaganda means and skills, use people's easy-to-understand words, and push relevant information to the audience in a personalized way, so as to enhance the guidance and pertinence of communication. On the other hand, the popular audience of Marxism should often carry out self-education and self-reflection, actively study the basic knowledge of Marxist theory, further improve their theoretical literacy and academic level, constantly cultivate their ability to look at things with dialectics, and consciously resist all kinds of remarks that are contrary to China's mainstream ideology. This not only plays an important role in building and optimizing the network ecosystem, but also provides a good space for promoting and promoting the popularization of Marxism.

5

Empirical Method of Marx's Popularization Based on Attention Mechanism

5.1

Analysis and Comparison of Marx's Popularization Evaluation

In order to further verify the effectiveness of the attention mechanism introduced in this paper on social governance and management, this paper quotes the survey data of Marx Sinicization in colleges and universities in two attention mechanisms, and makes a comparative experiment with samples from different regions. There are many models in Sample 1 and Sample 2, and we select two of them to compare with our own Marx evaluation effect model, so as to find which model obtains more accurate Marx popularization data. Generally speaking, the higher the accuracy rate and the lower the standard deviation model is better. Through the table data, it can be concluded that the social model constructed in this article in Marx's popularization evaluation database is the most suitable for evaluating the effect of Marx's popularization process. Table 2 shows the accuracy and standard deviation of the graph constructed in this article, classification on samples 1 and 2.

Table 2.

Draw accuracy of neural network classification on sample 1 and sample 2 datasets

Dataset	Model	Accuracy	Standard deviation
Sample 1	DGCNN	90.04%	19.25
	DBN	86.08%	15.35
	CNN+ATT	93.73%	3.56
Sample 2	DGCNN	69.88%	25.68
	DBN	69.08%	35.21
	CNN+ATT	83.59%	8.92

As can be seen from Table 2, compared with the baseline method, the advantages of the graph attention mechanism model are very obvious. Because the simulation method based on the adjacency matrix of spatial topological relations has certain advantages, it improves the efficiency of Marx's popularization. See Figures 5 and 6.

Since the attention mechanism is mainly applied in ECG classification, this article conducts ablation experiments. The objects of comparison are attention-based Bi-LSTM and Bi-LSTM, using the same data set, and other parameters of the model are set as above. The experimental results are shown in Table 3 and Figure 7.

Table 3.

Comparison of accuracy indexes of ablation experiments(%)

Dimension	Bi-LSTM	CNN+ATT
valence	76.65	73.14
arousal	70.15	69.03

Through the experiment after adding attention mechanism, it can be seen that the classification accuracy of the model is improved. The possible reason for the analysis is that attention can adjust the weight of indicators in different situations.

5.2

Comparison of entity relationship recognition efficiency of different models

This article contrasts the proposed model with models such as SVM, LR, LSTM, BiLSTM, and CNN. Among them, LSTM and BiLSTM models are the mainstream models with good extraction relationship at present.

It can be seen from Table 4 that the SVM machine learning algorithm has good performance in experimental results because of its good adaptability and universality. In contrast, the logistic regression model is relatively poor in experimental results due to its own mechanism, and its accuracy rate is about 12% lower than that of the SVM model, and its recall rate is about 9% lower. At the same time, compared with the LSTM model, the accuracy of the BiLSTM model has a certain improvement, but the recall rate has decreased. The effect of CNN on extracting indicators in a single situation is not as good as that of BiLSTM. The reason is that convolutional neural network is a oneway propagation model, and the extraction of complex indicators in specific social situations is not as single as that obtained by LSTM and BiLSTM, and the output results are relatively independent. However, the CNN model with the addition of Attention mechanism improves the extraction accuracy by nearly 3% compared with the CNN model. The visualization results are shown in Figure 8.

Table 4.

Comparison of efficiency of different model experimental systems

Model	Precision	Recall	F1
SVM	0.9018	0.8743	0.8878
LR	0.7833	0.7642	0.7736
LSTM	0.9384	0.9353	0.9367
BiLSTM	0.9408	0.9321	0.9363
CNN	0.9264	0.9264	0.8641
CNN+ATT	0.9532	0.9401	0.9449

5.3

Comparison of extraction efficiency of experimental index relationships in different vector training sets

In order to verify the applicability of experimental index relationship extraction under different circumstances, this paper uses two indexes with or without concretization as comparison.

It can be seen from Table 5 that especially for the vector-trained model, the F1 value has increased by about 4%. It can be seen that the convolution neural network model based on attention mechanism can obtain better index relationship extraction effect in different Marx generalization extraction test sets, and has better adaptability.

Table 5.

Comparison of specific indicators

Vector	Precision	Recall	F1
CNN+Attention(Random indicators)	0.8761	0.8483	0.8619
CNN+Attention(Specific indicators)	0.9176	0.8875	0.9021

Judging from the specific social situation evaluation extraction effect of different models, the index extraction effect of attention mechanism model is generally better than that of machine learning model, with the recall rate increasing by 1.3% and the F1 value increasing by 8%; Compared with the two-way long-term and short-term memory network model, the accuracy rate is 0.9% higher, the recall rate is 0.8% higher, and the F1 value is increased by 0.8%. The index extraction effect is obviously improved under complex and popular conditions.

From the extraction effect of Marx popularization of different label types, the effect of entity extraction is in the range of 88%-96%, and the effect of relationship extraction is relatively uniform. The model has good feasibility for the recognition of popularization evaluation effect.

From the effect of vector pre-training sample set to extract Marx popularization situation, the sample pre-trained by word2vec has better experimental results, and the accuracy rate is over 86%, which shows that the attention mechanism model proposed in this paper has a good effect on different training sets.

From the perspective of global stability, although the model has certain differences in recognition effect for different Marx popularization situations, from the perspective of global performance indexes, the model proposed in this paper has better extraction effect of specific Marx popularization situations indexes, and with the increase of experimental iterations, the data indexes of the model tend to be stable, which shows that the model has certain feasibility.

To sum up, this paper proves through relevant experiments that the model proposed in this paper has a good effect of extracting Marx's popularization index, which is consistent with the expected goal. Convolutional neural network based on attention mechanism can learn more sufficient multiple complex index relationships in different popularization situations, and can extract more effective system construction features. Looking forward to the future, this article will also improve from the following aspects: (1) Further optimize the model, such as introducing multi-head attention mechanism, collaborative attention mechanism, etc., and try to compare it with the current mainstream BiLSTM model based on attention mechanism. Experimental comparison; (2) Try to expand the diversity and richness of experimental samples, try to explore the development of Marx's popularization in finer grain, and try to use technologies in other fields of big data to develop Marx's popularization technology.

6

Conclusion

To sum up, under the background of big data era, the popularization development and dissemination of Marxism has aroused widespread concern and become a crucial task. How to grasp the influence of big data era, give full play to the advantages of big data technology, and promote the innovation and advancement of the popularization development and dissemination of Marxism has become an important topic in the ideological field under the new era. Therefore, in the new period, we should objectively look at the influence of big data technology, master the application of big data technology, and accurately position the development direction of Marxist popularization, so as to better guide the people to participate in socialist modernization, and help the construction of socialist cultural system with Chinese characteristics and the development of ideological work with the help of the strength of Marxist popularization. In short, although big data technology has brought challenges and opportunities to the popularization of Marxism. We should actively grasp, keep pace with the times, blaze new trails and innovate, make big data technology a new engine to drive the popularization of Marxism, and constantly open up a new realm of the popularization of Marxism.

Langue:: Anglais

Périodicité:: 1 fois par an
Sujets de la revue:: Sciences de la vie, Sciences de la vie, autres, Mathématiques, Mathématiques appliquées, Mathématiques générales, Physique, Physique, autres

RSS Feed de la revue

Research on the Popularization of Marxism by Big Data Based on Attention Mechanism

Yuanyuan Sun

Mingya Yang

Publié en ligne: 27 févr. 2025

Reçu: 30 sept. 2024

Accepté: 15 janv. 2025

DOI: https://doi.org/10.2478/amns-2025-0133

Mots clésMarx popularization, attention mechanism, big data, convolutional neural network model

© 2025 Yuanyuan Sun et al., published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Mots clés
Marx popularization, attention mechanism, big data, convolutional neural network model