Study on the Path of Foreign Cultural Translation and Dissemination of Jiangsu Prefecture-level Cities under Digital Technology Support

Suzhou, located in East China and South China, is a national historical and cultural city and a famous tourist city, as well as a modern and highly internationalized city that should not be underestimated due to its unique timing and geographical conditions. Suzhou culture, also known as Wu culture, has a long history, spanning about 2,500 years of history, with many distinctive features and relatively stable historical and cultural forms with inheritance relationships.

With the advent of artificial intelligence and digital era, publishing in the post-printing era presents very different characteristics from traditional paper publishing, and the concept of digital humanities is deeply rooted in people’s hearts [1–2]. The massive amount of digital resources has changed the paradigm of translation and communication research, making the research of translation and communication research multi-dimensional, with the research object shifting from “text” to “hypertext”, the research content from a single paper medium to multiple media, and the research results from “causality” to “relevance” [3–7]. Relying on the support of digital humanities, translators can obtain massive resources and a large amount of empirical materials, accurately grasp the phenomenon of translation and the facts of translation, and analyze the phenomenal characteristics and changing rules contained therein [8–10]. Translation research in the field of digital humanities focuses on translation content and translation method, which involves a large number of network translation objects, presenting distinctive digital characteristics [11–13]. The research object translates text, pictures, videos and other resources quickly with the help of network translation and other means, and the speed of data calling and screening is increased exponentially, which provides effective support for the dissemination of humanities content [14–17]. Moreover, it also reveals the digitalization change of translation means. Databases, corpora, and related software effectively penetrate into the field of translation with the help of data-based tools, and relying on massive data analysis can quickly mine the translated content and determine the correct translation results and facts [18–20]. This can provide an important guarantee for translators to explain, describe and study the laws of translation, which completely breaks the limitations of personal cognition and ability [21].

This paper points out the importance of translation and dissemination methods in the external translation and dissemination of the culture of Jiangsu prefecture-level cities, builds a framework for computer-aided translation technology, proposes the use of text similarity algorithms to achieve a high degree of matching retrieval, and applies cloud computing and parallel operation methods to update the response speed and optimize the use of computer-aided translation systems. Combine computer-aided translation technology with multimodal theory to create a multimodal strategy for translating and disseminating the culture of Jiangsu prefecture-level cities to the outside world. The accuracy of text similarity calculation based on feature fusion is analyzed, and the channels of audience access to the translation of Jiangsu prefecture-level city culture are investigated and counted.

2

Digitalization of the translation and dissemination of culture to the outside world

2.1

The Three Factors of Interpretation

External translation, as the name suggests, is the translation and introduction to the outside world, which is essentially a dialog and exchange with the cultural other, and in the vast majority of cases, it needs to cross the language barriers and convey cultural information, which is a process of language translation, and moreover, it is a practice of heterogeneous cultural dissemination.

Comprehensive translation and communication theory and practice, for the local characteristics of the more special local culture of the construction of foreign translation and communication mode, the five elements of communication can be combined and summarized in three main factors. That is, the main body of translation, the way of translation, and the effect of translation. It is hoped that it can be systematically assessed based on the organic whole of interconnection and mutual influence, in order to better showcase the exceptional parts of local culture.

First, the main body of local culture translation. The main components of translation are the translator and the disseminator. In the process of translating the culture of Jiangsu prefecture-level cities, the translator is first of all the translator of Jiangsu prefecture-level cities or the translator of Jiangsu localities who is immersed in the spiritual and cultural atmosphere of Jiangsu. They have a clearer understanding and grasp of the generation, development history, and characteristics of Jiangsu culture, so as to ensure the respect and fidelity of the translation to the original. With local translators being the main source, cooperation between Chinese and foreign translators is the best way to ensure the readability and acceptability of the translation.

Secondly, the multidimensional process of translating local culture into foreign languages. Under the perspective of internationalization, it is even more important to enhance proactive awareness of external communication, broaden the ways and channels of communication, and enhance the effect of local culture’s characteristic communication. The integration of multiple forms of communication requires both government-centered common communication activities and private, individual exchanges and communication. Both are compatible and complementary to each other in order to achieve better communication effects.

Thirdly, the utility of external translation of local culture. The ultimate goal and focus of local culture dissemination to foreign countries is to make local culture play a practical role in Chinese and foreign cultural exchanges, and really play the role of cultural infection and sensitization.

2.2

Ways of translating and disseminating the culture of prefecture-level cities to the outside world

The mode of translation communication is an important issue worth exploring in external communication of Jiangsu prefecture-level culture. The study of translation communication modes includes the understanding of translation communication symbols and media. Symbols are carriers of meaning or information, and mediums are the material entities that transmit them. Translation is the process of transmuting symbols to convey information, and the symbols loaded with information can only be accepted by the receiver in the process of transmission by relying on specific material entities (media) such as paper, sound, audio tape, video, videotape and so on. The development of media forms has gone through different stages such as oral communication, paper communication, electronic communication, and network communication. Different media have their own characteristics and functions, and the cause of cultural interpretation in Jiangsu prefecture-level cities can only play a complementary advantage by comprehensively utilizing different communication media, including books, newspapers, broadcasting, film and television, and networks.

Books, magazines, and newspapers are the main communication media for foreign translation of the culture of Jiangsu prefecture-level cities. However, their publishing organizations should not be limited to Jiangsu province or domestic, but should strengthen international exchanges and cooperation, and make full use of excellent foreign publishing units and mainstream media to publish and promote them, so as to find a window for Jiangsu prefecture-level city culture to enter the international market. At the same time, it is necessary to create different forms of translation versions (such as full translation, abridged translation, compilation, annotated translation, and painting) to adapt to different readers’ groups, and to expand the scope of readers and social influence of ethnic cultures. Minority films from Jiangsu prefecture-level cities are another important carrier for promoting Jiangsu national culture to the outside world, showing the charm of Jiangsu people with its unique regional style and folk customs. Animation is also an important part of translating and disseminating Jiangsu prefecture-level cities’ culture to foreign countries.

The network is another powerful tool to spread the culture of Jiangsu prefecture-level cities to the outside world. With the help of network technology, the culture of Jiangsu prefecture-level cities can be presented in a way that is pleasing to the audience. Through the establishment of official websites, community websites, personal websites, blogs, website posting, forums, and other forms of external translation and dissemination of Jiangsu prefecture-level city culture. With the rapid development of cell phone media, such as cell phone books, cell phone journals, cell phone newspapers, cell phone radio, cell phone TV, etc., can be used as a good communication channel to promote the culture of Jiangsu prefecture-level cities. In addition, if the culture of Jiangsu prefecture-level cities can enter the social media used by westerners in Britain and the United States, it will boost the effect of dissemination, such as Facebook, Tumblr, Twitter, Youtube and so on, and the reprinting rate and sharing rate of these media are quite high.

2.3

External translation and dissemination of digital culture

2.3.1

Intelligent translation

Smart Translation mainly refers to the construction of a multimodal translation form combining text + voice/video on the basis of traditional manual translation, with the help of the Smart Media big data platform, to provide customized translation services with highly matched information. In this way, it promotes human-computer interaction and paper-screen interaction in translation, and meets the diversified needs of readers of different countries and regions, ages and classes by means of an accurate and efficient “wisdom of translation” communication mode, so as to strengthen the construction of international communication capacity.

The concept of “intelligent translation” mainly includes two fusions: the fusion of information dissemination and translation.

Integration of technical means and translation. With the emergence of new media technology, people’s reading habits and aesthetic acceptance habits have greatly changed, and the existing modes can no longer accurately and quickly meet the needs of the target language audience. As technology pushes translation to upgrade and iterate to digitalization and intelligence, the foreign translation of literary works should take into full consideration of the target language’s spatial and temporal characteristics, gender characteristics, textual characteristics, etc., so as to realize the adaptive conversion between the translated language and the target language, and to improve the international “consensus” of linguistic symbolic expression. Such translation “consensus” can be regarded as the core element of “intelligent translation”, i.e. “intelligent translation in three dimensions”. 1)

Time and space translation: Starting from thinking imagery, we select the narrative time nodes of Chinese literary works to explain the intellectual media translation reconstruction such as heterogeneous isomorphism, homogeneous isomorphism and juxtaposition and concurrence.

2)

Gender Translation: Starting from gender awareness, the translator’s gender language, gender identity and subconscious gender characteristics are analyzed by intelligent media big data to rewrite the sexist contents of Chinese literary works, and analyze the influence of gender factors on the external environment, such as the social, economic and cultural environment.

3)

Text Translation Map: Starting from the intention and meaning space of text presentation, we will focus on paratextual reconstruction, de-redundant reconstruction and hypertextual reconstruction of Chinese literary works to explain the narrative intention of translation.

In other words, on the basis of traditional translation theory and in line with the development of the big data era, “Smart Translation” incorporates information dissemination, translation technology and cloud translation into the framework of translation theoretical research in the era of smart media, so as to construct a new pattern of “wisdom of translation” and promote Chinese culture to the world. To spread Chinese culture throughout the world. Intellectualized translation is actually the organic unity of translators, intellectual elements of computers, physical elements such as tools and equipment, and human-computer interaction, and it is a translation and communication system based on the organic integration of the subject of translation, the object of translation, and the process of translation and translation technology.

2.3.2

Multimodal translation strategies

Jiangsu literature is mostly translated into foreign languages by China Foreign Language Press and foreign publishers. The works of famous Jiangsu writers are all involved, but they are more scattered. Writers such as Wang Zengqi, Gao Xiaosheng, Lu Wenfu, Zhao Benfu, Fan Xiaoqing, Ye Zhaoyan, Su Tong, Bi Feiyu, Huang Beijia, and Ye Mi have had their works translated more sporadically overseas. The “Jiangsu Literature Masterpieces” has undoubtedly provided a more coherent and focused perspective for the world to understand Jiangsu’s cultural heritage and stylistic characteristics. From the current situation, government-led literary translation is still mainly based on textual modality, ignoring the multimodal reading mode brought by modern technology, which makes it difficult to guarantee the effectiveness of translation. This paper aims to expand the scope of translation and enhance the translation strategy through the lens of multimodal translation theory.

In recent years, multimodal theory has been continuously applied to translation practice and translation research. Since entering the 21st century, multimodal texts have integrated linguistic and non-linguistic symbols such as films and television, comics, advertisements, websites, videos, and so on. Have been commonly used. With the development of digital technology, media and reading habits have changed. Mobile reading terminals, smart phones, and audiobooks have stimulated new demands from audiences. Therefore, the translation of contemporary literature in Jiangsu can adopt the intermodal translation paradigm, abandoning single-modal translation that focuses only on linguistic symbols. By adding narrative elements such as images, sounds, and pictures, Jiangsu literature can be brought to foreign audiences in a richer form, broadening the communication channels and improving the effect of translation and mediation. After multimodal translation, Jiangsu contemporary literature can be “molded” into hypertext, movies, operas, picture books and other cultural products, so that foreign audiences can accept the literary works in a more intuitive way, triggering emotional resonance, and further improving the effectiveness of dissemination.

2.3.3

Computer-assisted translation techniques

Computer-assisted translation technology is defined in both a broad and a narrow sense. In a broad sense, it encompasses all software that assists translators in their work, including electronic dictionaries, search engines, and word processing software. In the narrow sense, it refers to the computer tools used in the process of translation realization, including project management, translation memory, and other tools [22–23].

As a modern translation technology, computer-assisted translation technology (CAT) uses translation memory technology as the core of a translation solution. This technology requires translators to use a computer program to translate, and it is applied based on the corpus and thesaurus stored in the system. The system automatically judges the sentence, divides it into multiple phrases, and finds and provides translation suggestions from the system’s thesaurus. The system displays the reference translation by automatically searching the translation resources in the memory. The translator reorganizes and adjusts the information according to the translation suggestions, or directly adopts the reference translation, so that the translation is also completed. This technology, unlike machine translation, requires personnel to participate in the translation process, and this technology is responsible for the mechanical, trivial and complex work in the translation process, while the translators are responsible for reorganizing and arranging the translations, which enables the advantages of the two to be maximized, and improves the efficiency as well as the quality of the translation. 1)

Mathematical foundation

In computer-aided translation technology, a large number of mathematical formulas are used, and translation technology is developed on the basis of a large number of mathematical theories. Including:

2)

Random Events: 1 $P (A) = (0 ~ 1)$ $$P(A) = (0 \sim 1)$$

Frequency and probability: 2 $f (A) = \frac{v}{n}$ $f(A) = {v \over n}$

When n gradually increases and the frequency stabilizes, then there is a probability: 3 $P (A) \approx \frac{v}{n}$ $P(A) \approx {v \over n}$

Conditional Probability: 4 $P (A | B) = \frac{P (A B)}{P (B)}$ $P(A|B) = {{P(AB)} \over {P(B)}}$

Independence formula when: 5 $P (A B) = P (A) P (B)$ $P(AB) = P(A)P(B)$

Bayesian formula: 6 $P (B_{i} | A) = \frac{P (B_{i}) P (A | B_{i})}{\sum_{j = 1}^{n} P (B_{j}) P (A | B_{j})}$ $P({B_i}|A) = {{P({B_i})P(A|{B_i})} \over {\sum\limits_{j = 1}^n P ({B_j})P(A|{B_j})}}$

Random variables: 7 $x = x_{1}, x_{2}, x_{3} \dots \dots x_{n}$ $x = {x_1},{x_2},{x_3} \cdots \cdots {x_n}$

Probability mass function: 8 $P X = x k = p k, (k = 1, 2, 3... n)$ $PX = xk = pk,(k = 1,2,3...n)$ where 1 ≥ pk ≥ 0, and Σpk = 1.

Expected variance standard deviation: 9 $E (X) = X_{1} p (X_{1}) + X_{2} p (X_{2}) + \dots \dots X_{n} p (X_{n})$ $E(X) = {X_1}p({X_1}) + {X_2}p({X_2}) + \cdots \cdots {X_n}p({X_n})$

Variance formula: 10 $s^{2} = \frac{\sum (x - x)}{n - 1}$ ${s^2} = {{\sum {(x - x)} } \over {n - 1}}$

Standard deviation formula: 11 $x = \sqrt{\frac{1}{n}} \sum_{i - 1}^{N} {(x_{i} - μ)}^{2}$ $x = \sqrt {{1 \over n}} \sum\limits_{i - 1}^N {{{({x_i} - \mu )}^2}} $

Overall sample: 12 $n = {n_{1}, n_{2}, n_{3} \dots n_{n}}$ $n = \{ {n_1},{n_2},{n_3} \cdots {n_n}\} $

3)

Search accuracy and response speed are the two most important indexes of computer-aided translation software. The actual response of checking accuracy is the matching degree between the original text of the retrieval feedback and the original text retrieved by the user, that is, the retrieval feedback result can present all the translations with translation reference value in the corpus to the user according to the order of similarity. Different from the traditional fuzzy database query, the retrieved content is not a definite search keyword, and in fact, the retrieved target and the retrieved content are only textually similar or approximate. Therefore, fuzzy queries such as those realized by Like statements are not applicable in the retrieval of computer-aided software. In addition, as with other retrieval systems, response speed is the most important performance index of computer-aided translation software, and usually the translation corpus has a large amount of data, so the optimization of query speed is especially important for computer-aided software.

To address the above problems, the text similarity algorithm is adopted to achieve high matching degree retrieval, and methods such as cloud computing and parallel operation are applied to optimize the response speed, so as to ultimately realize the optimal user experience of computer-aided translation software.

(1)

Query based on text similarity algorithm

This paper proposes a multi-feature fusion short text similarity improvement algorithm model using attention and Bi LSTM. The model can be divided into the following parts:

a)

Embedding layer: for two short texts, the input is transformed into word embedding and character embedding at the embedding layer, and the two embedding results are spliced.

b)

Coding layer: the coding layer consists of two coding methods, Bi LSTM and Multihead Self-Attention, which extract the semantic features of the text from different perspectives, respectively.

BiLSTM coding is introduced first. Since BiLSTM contains two LSTMs, the hidden states generated by each time step of these two LSTMs are connected to represent that time step and its contextual features, as shown in Eqs. (13) and (14): 13 $a_{b i l s t m, i} = BiLSTM (a_{e m b}, i), \forall i \in [1, \dots, l_{a}]$ ${a_{bilstm,i}} = {\rm{BiLSTM}}({a_{emb}},i),\>\forall {\rm{i}} \in [1, \cdots ,{l_a}]$ 14 $b_{b i l s t m, j} = BiLSTM (b_{e m b}, j), \forall j \in [1, \dots, l_{b}]$ ${b_{bilstm,j}} = {\rm{BiLSTM}}({b_{emb}},j),\>\forall {\rm{j}} \in [1, \cdots ,{l_b}]$ where a_bilstm,l is the hidden state generated by the BiLSTM at the i th moment for the coding layer input a_emb. Similarly, b_bilstm,j is the hidden state generated by BiLSTM at the j th moment for the coding layer input b_emb.

In addition to using the semantic features with context dependency learned by BiLSTM, the model further utilizes the self-attention mechanism to encode two short texts to strengthen the attention to the key features of the text itself. The multi-head self-attention mechanism can essentially be viewed as an independent computation of multiple single-head attention, i.e., sentences are input into multiple linear layers for attention weight computation and connection respectively. The single-head self-attention calculation process, as shown in Equation (15): 15 $S e l f - A t t e n t i o n = s o f t \max (\frac{W \cdot W^{T}}{\sqrt{d_{k}}}) W$ $Self - Attention = soft\max \left( {{{W \cdot {W^T}} \over {\sqrt {{d_k}} }}} \right)W$ where W is the input vector of the coding layer, a_emb and b_emb are used for the self-attentive feature representation of the two texts, respectively, $\sqrt{d_{k}}$ $\sqrt {{d_k}} $ is the scaling factor to prevent the inner product from being too large, and d is the dimension of the embedding vector.

In the coding layer, the residual join is to splice the output vectors obtained from the two coding methods with the inputs of the coding layer respectively, which allows the model to provide richer features to the interaction layer. After residual concatenation, the final coded features of the coding layer are obtained as shown in Eqs. (16) and (17): 16 $\bar{a} = [a_{e m b}; a_{b i l s t m}], \tilde{a} = [a_{e m b}; a_{s e l f a t t}]$ $\bar a = \left[ {{a_{emb}};{a_{bilstm}}} \right],\tilde a = \left[ {{a_{emb}};{a_{selfatt}}} \right]$ 17 $\bar{b} = [b_{e m b}; b_{b i l s t m}], \tilde{b} = [b_{e m b}; b_{s e l f a t t}]$ $\bar b = \left[ {{b_{emb}};{b_{bilstm}}} \right],\tilde b = \left[ {{b_{emb}};{b_{selfatt}}} \right]$ where $\bar{a}$ ${\bar a}$ and $\bar{b}$ ${\bar b}$ are BiLSTM codes after residual concatenation, and $\bar{a}$ ${\bar a}$ and $\bar{b}$ ${\bar b}$ are self-attentive codes after residual concatenation. a_emb and b_emb are the vectors of the two inputs after embedding operation, a_bilstm and b_bilstm are the features after BiLSTM coding, and a_selfatt and b_selfatt are the feature vectors after self-attentive coding.

c)

Interaction layer

In order to be able to distinguish the critical and irrelevant parts between two sentences, an interaction layer is added to the model. The interaction layer can fully explore the interaction information between sentences and represent each sentence accordingly while considering the interaction effect with another sentence.

The model selects dot product as the realization method of attention weight. After dot product, the interaction layer gets the attention weight e_ij, as shown in Equation (18). This similarity matrix is then used to generate sentence representations weighted by the similarity to each other by combining the feature vectors of the two inputs, as shown in Eqs. (19) and (20): 18 $e_{i j} = {\bar{a}}_{i}^{T} {\bar{b}}_{j}$ ${e_{ij}} = \bar a_i^T{{\bar b}_j}$ 19 $\hat{a} = \sum_{j = 1}^{l_{b}} \frac{\exp (e_{i j})}{\sum_{k = 1}^{l_{b}} e x p (e_{i k})} {\bar{b}}_{j}, \forall i \in [1, \dots, l_{a}]$ $\hat a = \sum\limits_{j = 1}^{{l_b}} {{{\exp ({e_{ij}})} \over {\sum\limits_{k = 1}^{{l_b}} e xp({e_{ik}})}}} {{\bar b}_j},\>\forall i \in [1, \cdots ,{l_a}]$ 20 $\hat{b} = \sum_{i = 1}^{l_{a}} \frac{\exp (e_{i j})}{\sum_{k = 1}^{l_{a}} e x p (e_{k j})} {\bar{a}}_{i}, \forall j \in [1, \dots, l_{b}]$ $\hat b = \sum\limits_{i = 1}^{{l_a}} {{{\exp ({e_{ij}})} \over {\sum\limits_{k = 1}^{{l_a}} e xp({e_{kj}})}}} {{\bar a}_i},\>\forall j \in [1, \cdots ,{l_b}]$

Where e_ij is the attention weight generated by dot product computation and $\hat{a}$ ${\hat a}$, $\hat{b}$ ${\hat b}$ is the feature representation of the two inputs after bidirectional attention. $\bar{a}$ ${\bar a}$, $\bar{b}$ ${\bar b}$ is the BiLSTM encoding of the input to the interaction layer after residual concatenation.

Finally, using an approach similar to residual concatenation, the inputs, outputs, and results after computation of this attention layer are subjected to a splicing operation that preserves the encoding features of the sentence, the interaction features, and the difference and product between these two as in Eqs. (21) and (22) to obtain a new feature representation of the two sentences. Namely: 21 $v_{a} = [\bar{a}; \hat{a}; \bar{a} - \hat{a}; \bar{a} ▯ \hat{a}]$ ${v_a} = [\bar a;\hat a;\bar a - \hat a;\bar a \unicode {x25AF}\hat a]$ 22 $v_{b} = [\bar{b}; \hat{b}; \bar{b} - \hat{b}; \bar{b} ▯ \hat{b}]$ ${v_b} = [\bar b;\hat b;\bar b - \hat b;\bar b \unicode{x25AF} \hat b]$

Where v_a and v_b are the final feature representations of the interaction layer. $\bar{a}$ ${\bar a}$, $\bar{b}$ ${\bar b}$ are the feature vectors of the input text after BiLSTM coding, respectively, and $\hat{a}$ ${\hat a}$ and $\hat{b}$ ${\hat b}$ are the feature vectors obtained from $\bar{a}$ ${\bar a}$ and $\bar{b}$ ${\bar b}$ after bi-directional attention.

d)

Fusion prediction layer

The model uses both average pooling and maximum pooling operations to stitch together all the pooled features of the text into a final fixed-length vector, as shown in Eqs. (23) and (24): 23 $o_{a} = [v_{a, a v e}; {\tilde{a}}_{a v e}; {\tilde{a}}_{\max}; v_{a, \max}]$ ${o_a} = \left[ {{v_{a,ave}};{{\tilde a}_{ave}};{{\tilde a}_{\max }};{v_{a,\max }}} \right]$ 24 $o_{b} = [v_{b, a v e}; {\tilde{b}}_{a v e}; {\tilde{b}}_{\max}; v_{b, \max}]$ ${o_b} = \left[ {{v_{b,ave}};{{\tilde b}_{ave}};{{\tilde b}_{\max }};{v_{b,\max }}} \right]$ where v_a,ave, v_b,ave is the feature representation of the interaction layer features after average pooling and v_a,max,v_b,max is the feature representation of the interaction layer features after maximum pooling. ${\tilde{a}}_{a v e}$ ${{\tilde a}_{ave}}$, ${\tilde{b}}_{a v e}$ ${{\tilde b}_{ave}}$, ${\tilde{a}}_{\max}$ ${{\tilde a}_{\max }}$, ${\tilde{b}}_{\max}$ ${{\tilde b}_{\max }}$ is then the result of the average and maximum pooling operations performed by f-aon coding. o_a, o_b is the final feature of the two inputs after splicing the two pooling results.

Finally, the spliced two sentence vectors are interacted again, and elemental subtraction and elemental multiplication are used to obtain the similarities and differences between the two sentences. And the two sentence representations and the results obtained from the elemental subtraction and multiplication computation are spliced to obtain the final feature representation for similarity prediction as shown in Eq. (25): 25 $O u t p u t = [o_{a}; o_{b}; o_{a} - o_{b}; o_{a} ▯ o_{b}]$ $Output = [{o_a};{o_b};{o_a} - {o_b};{o_a} \unicode {x25AF} {o_b}]$

The final spliced vectors are fed into a multilayer perceptual machine and feature classification is performed using the Sigmoid function. The Sigmoid function normalizes the output and maps it to the (0, 1) interval to get the final prediction of whether the two input short texts are similar or not.

(2)

Optimization of fast query technology based on cloud computing

In order to improve the translator’s experience and reduce the real-time translation time. The optimization of query algorithm also considers other methods to shorten the query time.

In order to minimize the differences brought by the different computers of users, most of the calculation work is put on the server to complete in the software design, and the user side only completes the transmission and display of data information. This “cloud computing” design architecture maximizes the performance of the computer-aided translation software and at the same time allows each user to enjoy a relatively close user experience.

Parallel computing is also designed to improve performance. Due to the large database table on the reason after the above optimization of the query speed is still not ideal, so the need to split the database into a number of small databases for parallel query and then merge the query results, thus minimizing the query time. That is::Parallel computing is also designed to improve performance. Due to the large database table on the reason after the above optimization of the query speed is still not ideal, so the need to split the database into a number of small databases for parallel query and then merge the query results, thus minimizing the query time. That is: 26 $t = t_{s 0} + \sum_{0}^{n} t_{i} + t_{s 1}$ $t = {t_{s0}} + \sum\limits_0^n {{t_i}} + {t_{s1}}$

Where t is the entire computation process time t_s0 is the time spent on splitting the parallel computation task, t_i is the time spent on calculating a single parallel task, and t_s1 is the time spent on merging the computation results. t_s0,t_s1 increases with the number of task splits n, so it is not better to have a larger number of parallels. In this scheme, n=12 is chosen as the optimal number of parallel queries.

(3)

Workflow of computer-aided translation software

The workflow of computer-aided translation software is shown in Figure 1. The main workflow is:

In the first step, the translator inputs the working document to be translated (the original text) into the computer-aided translation software, which reads the original text sentence by sentence.

In the second step, the software automatically carries out fuzzy query according to the whole sentence read, returns the translation with similar matching degree from the corpus, and the original text and the translation are displayed in the working area at the same time.

In the third step, the software performs word classification on the current original text and returns the translation word by word according to the result of word classification.

In the fourth step, the translator proofreads the translation, word translation and original text against the translation returned from the corpus.

In the fifth step, after the proofreading of the translation is completed, the software inserts it into the original text position in the working document and updates the corpus at the same time.

Step 6: Return to the first step.

3

Analysis of the foreign translation and dissemination strategies of the culture of prefecture-level cities in Jiangsu Province

3.1

Utility of computer-assisted translation technology

In order to validate the effectiveness of the method designed in this paper from multiple aspects and to compare it with some efficient previous work, the classic SemEval dataset is used. The SemEval dataset is derived from the dataset used in the International Open Competition on Natural Language Technology. In it, a corpus of short texts is collected from several different aspects, such as forum replies, Q&A services, social news, and other domains. The evaluation metric is the Pearson’s coefficient, which is distributed as (0, 5), the closer to 5 the more it indicates a linear correlation between the two. 0 indicates that there is little necessary connection between the two short text sentences. SemEval is a collection of huge databases designed for different domains.

In order to explore the enhancement of the constructed network on the dataset comes from the feature fusion mechanism. Experiments were carried out on the named entity dataset for three of the metrics, i.e., the accuracy of Person Entity Name (PER), the accuracy of Location Entity Name (LOC), and the accuracy of Organization Entity Name (ORG). The performance of different feature combinations on NER data is shown in Figure 2.

In this experiment, variables such as the type of features are used to unfold the experiment. The figure shows the results of different feature combinations after extracting features and calculating short text similarity. The best results are obtained by the group of such features based on location, glove, and dependency information, where all the features are complete. The experimental group which has no additional features and just processed the data using random coding is set as baseline, as can be seen from the data in the figure it has achieved 84.996%, 83.969% and 82.657% accuracy for person entity, place entity and organization entity respectively.

In the case of multiple features, the multi-feature data yields the best results, with huge improvements over baseline in all three metrics. Compared to baseline, for the single-feature experimental control group, Glove information showed the greatest amount of improvement in the results, followed by dependent features. The lowest improvement in the results was for location.

And for the three control groups with dual features, most of their results show results beyond the single feature group. However, for the experimental data set based on location and dependency information compared to the experimental group based on Glove vectors, the Glove vectors still showed better results than the experimental group based on dependency vectors and location vectors, which suggests that there is a difference in the weighting between the features themselves on the results.

To summarize and analyze the above data, the best results achieved for the experimental group with full features (Glove, Location, Dependency) indicate that the data feature fusion and screening mechanism are effective. Without considering the influence of the weight of the features on the results, the performance of the experimental group with multiple features is better than that of a single feature, and the above performance proves the effectiveness of the data feature screening and fusion mechanism.

In order to make the described method objective and efficient, it is compared with some previous classical algorithms on the SemEval dataset. The experimental results are shown in Fig. 3, where the performance on 6 out of 7 datasets performs as the best. The performance on the remaining SMTnews dataset is not far from the performance of the best results.

There is no lack of classical methods in the figure, which at that time achieved very good results on these datasets. The table displays the Long-Short Memory Network method, an improved version of RNN.ST stands for the Skip-thought vector, while PSL stands for the PARAGRAM-SL999 vector. The two methods are used as Baseline and the other methods are used as reference values. Dan network achieved a Pearson coefficient value of 0.652 on SMTnews, which is a high result in comparison with the other methods. w2v-EDT is based on word2vec word vectors, and combines with the task’s needs to design the corresponding formulas for distance editing. It achieved a Pearson coefficient of 0.787 on Tweet, and the method constructed in this paper scored higher than the two baselines of ST and PSL on all the datasets. Compared with IRNN, LSTM, RNN and other methods, the method built in this paper also has greater advantages. It also shows the effectiveness of the method described in this paper, which can improve text translation accuracy when applied to foreign translation and communication of Jiangsu prefecture-level city culture. Combined with Internet websites such as News.com and other social platforms such as Weibo and WeChat, it ensures the limited output of culture.

3.2

Multimodal Translation Communication Effects

For this research, three versions of the questionnaire were designed according to the complexity, specificity, and differences of the respondents. The English version of the questionnaire, with English as the international lingua franca, was able to reach the widest range of people. The questionnaire mainly targets foreigners in Jiangsu, and nearly all of the respondents are native speakers of non-Chinese languages. Nanjing, Suzhou, Wuxi and Lianyungang in Jiangsu Province were selected for the field survey, and these people either stayed in Jiangsu for a long time or for a short time. These people have direct contact with Jiangsu, and therefore have a more intuitive understanding of the translation and dissemination of Jiangsu culture to the outside world.

The Chinese version of the questionnaire, the questionnaire for the Chinese abroad, this part of the population, the vast majority of the mother tongue is Chinese, influenced by China, more or less have a certain understanding of China. Chinese people from Jiangsu, comparing the development status quo of different regions, can have more opinions and suggestions on translating and disseminating the culture of Jiangsu prefecture-level cities to the outside world. Therefore, in the Chinese version of the questionnaire, although they are all placed in the online form, the questions are still set to be extensive and detailed.

English simplified version of the questionnaire. This questionnaire for foreigners living abroad found that most of the survey respondents have never been to Jiangsu, and rarely receive attention from the national media or networks due to their lack of travel. They know very little about Jiangsu, but this group of people is also the audience for the translation and dissemination of the culture of Jiangsu prefecture-level cities to the outside world. The simplified English version of the questionnaire, which grasps the degree of understanding of Jiangsu prefecture-level city culture by local people living abroad, is a reflection of the completeness of this research and an indispensable part of guiding Jiangsu’s foreign communication practice in the context of digitization.

3.2.1

Content of Interpretation

The cultural belt of prefecture-level cities in Jiangsu Province is rich in resources and extensive in content, and when it is really translated and disseminated to the outside world, it is necessary to clarify the cultural content of foreign dissemination. At the present stage of foreign dissemination, it is necessary to seize the core features of the non-heritage culture of the cultural belt of prefecture-level cities in Jiangsu Province, and disseminate from the core outward, gradually expanding the scope and breadth of dissemination. At the same time, the opinions of staff and students from foreign countries coming to China are combined. Living in a foreign language context for a long time, they will have higher sensitivity to foreign languages and better understand foreign language characteristics and expressions, listening to their opinions can improve the quality of foreign output content. While the translated content conforms to the characteristics of foreign languages and cultures, the essence of traditional Chinese culture is not lost. While trying to find common cultural symbols, we can create cultural resonance and enhance the acceptability of the output content.

Respondents’ keyword evaluation of the cultural characteristics of Jiangsu prefecture-level cities is shown in Table 1. According to the respondents’ understanding of the culture of Jiangsu prefecture-level cities, seven keywords of “openness, harmony, freedom, ecology, civilization, inclusiveness, and culturality” are summarized. Among them, 34% of the respondents believe that the keyword of Jiangsu prefecture-level city culture is openness. Thus, it can reflect the focus of the foreign translation of Jiangsu prefecture-level city culture, highlighting the openness of Jiangsu prefecture-level city culture, its rich spiritual civilization, and its inclusive cultural labels.

Table 1.

The key words evaluation of the cultural characteristics of Jiangsu city

	Key words	Frequency	Percentage	Effective percentage	Cumulative percentage
In effect	Open	186	0.34	0.34	0.34
	Harmony	84	0.15	0.15	0.49
	Freedom	65	0.12	0.12	0.61
	Ecology	31	0.06	0.06	0.67
	Civilization	67	0.12	0.12	0.79
	Inclusiveness	15	0.03	0.03	0.82
	Culture	98	0.18	0.18	1.00
Total		546	1.00	1.00	-

3.2.2

Translation and dissemination channels

The survey statistics of the ways to know about Jiangsu prefecture-level cities are shown in Table 2. Respondents’ channels of understanding Jiangsu prefecture-level cities are mainly through the Internet (37%), followed by visiting tourism (15%), and then cultural activities (12%). The Internet communication method combined with digital technology accounts for the main way of understanding.

Table 2.

To understand the way of survey in Jiangsu

	Pathway	Frequency	Percentage	Effective percentage	Cumulative percentage
In effect	The Internet	203	0.37	0.37	0.37
	Broadcast	12	0.02	0.02	0.39
	Newspapers	9	0.02	0.02	0.41
	Film and television	34	0.06	0.06	0.47
	Tour	81	0.15	0.15	0.62
	Cultural activity	65	0.12	0.12	0.74
	Interpersonal communication	18	0.03	0.03	0.77
	Official visit	20	0.04	0.04	0.81
	Outdoor advertising	16	0.03	0.03	0.84
	School class	25	0.05	0.05	0.88
	Other	63	0.12	0.12	1.00
Total		546	1.00	1.00	-

The statistics of ways to understand the culture of Jiangsu prefecture-level cities are shown in Table 3. Nowadays, audiences mostly learn about Jiangsu prefecture-level city culture from social platforms such as WeChat and Weibo. News websites and other emerging media also play a significant role in the dissemination of Jiangsu prefecture-level city culture. The proportion of the two is 47% and 21%, respectively. Social platforms such as WeChat and Weibo and emerging media such as news websites reflect the multimodal translation strategy for the dissemination of Jiangsu prefecture-level culture, which integrates linguistic and non-linguistic symbols such as movies, cartoons, advertisements, websites and videos, and is commonly used by respondents. After multimodal translation, Jiangsu contemporary literature is presented as hypertexts, movies, operas, picture books and other cultural products, which enable foreign audiences to receive literature in a more intuitive way, trigger emotional resonance, and further improve the effectiveness of communication.

Table 3.

To understand the method of culture in the local city of Jiangsu

	Pathway	Frequency	Percentage	Effective percentage	Cumulative percentage
In effect	Traditional media such as newspapers and television	84	0.15	0.15	0.15
	News stations and other emerging media	117	0.21	0.21	0.37
	Wechat, weibo and other social platforms	254	0.47	0.47	0.83
	Word-of-mouth of interpersonal communication	65	0.12	0.12	0.95
	Other	26	0.05	0.05	1.00
Total		546	1.00	1.00	-

In addition to the existing mainstream media, government departments, institutional media, platform media, and self-media can all be involved in the dissemination of culture in Jiangsu prefecture-level cities. It is important to fully mobilize all parties and fully exploit their respective advantages under the guidance of the government. The government provides general direction, and other platforms carry out a series of activities based on this general direction. For example, schools and communities can collaborate to promote canal culture within the community and schools, thus bringing closer together students and residents and experiencing its charm in real life. Network publicity media can produce bilingual videos under the guidance of professionals, making video communication more intuitive and more capable of revealing the stories, ideas, and spirit behind the culture.

4

Conclusion

This paper is oriented to the way of translation and dissemination of the culture of prefecture-level cities in Jiangsu Province, and puts forward the three major elements of translation and dissemination, books, magazines and newspapers, and the Internet respectively. It proposes a multimodal translation strategy that combines computer-assisted translation (CAT) technology with books, newspapers, and the Internet. Respondents are invited to share their impressions of the culture of Jiangsu prefecture-level cities, and the channels through which they learn about the culture of Jiangsu prefecture-level cities are counted. Taking into account the existing development, innovative ways of translating and disseminating prefecture-level city culture are proposed. 1)

The short text similarity computation model with multi-feature fusion is able to achieve high accuracy rates for character entity names, location entity names, and organization entity names in the presence of multiple features. In the classic SemEval dataset, 95.512%, 93.536%, and 91.944% accuracy rates are obtained for characters, locations, and organizations, respectively. Combined with the optimized computer-aided translation technology can accelerate the speed of external translation of Jiangsu prefecture-level culture and ensure the correctness of cultural communication to the outside world. Combined with the Internet and other emerging media, it can provide users with better cultural knowledge.

2)

Invite the public with different degrees of understanding of Jiangsu prefecture-level city culture to conduct a questionnaire survey, count the ways of understanding, and propose that in addition to the mainstream media, government departments, institutional media, platform media, and self-media should be involved in the dissemination of the culture of Jiangsu prefecture-level cities, so as to expand the influence of the translation and dissemination of the culture of Jiangsu prefecture-level cities to the outside world.

Language:: English

Publication timeframe:: 1 times per year
Journal Subjects:: Life Sciences, Life Sciences, other, Mathematics, Applied Mathematics, General Mathematics, Physics, Physics, other

Journal RSS Feed

Study on the Path of Foreign Cultural Translation and Dissemination of Jiangsu Prefecture-level Cities under Digital Technology Support

Jinhua Qi

Xueheng Zhou

Published Online: Mar 19, 2025

Received: Nov 09, 2024

Accepted: Feb 15, 2025

DOI: https://doi.org/10.2478/amns-2025-0425

KeywordsComputer translation technology, Text similarity, Multi-feature fusion, Cultural translation, Communication paths

© 2025 Jinhua Qi et al., published by Sciendo

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

Keywords
Computer translation technology, Text similarity, Multi-feature fusion, Cultural translation, Communication paths