Cross-culture marketing semantic field reconstruction and emotion migration translation method and system

Abstract

The invention relates to the field of computer language translation, in particular to a cross-culture marketing semantic field reconstruction and emotion migration translation method and system.The method comprises the steps that a culture semantic field of a target brand is established, value invariants of the target brand are defined based on the culture semantic field, and then input of source content of the target brand is responded; the method comprises the following steps of: mapping source contents by adopting a preset emotion migration operator to obtain a target emotion vector, then, carrying out decoupling and function reconstruction on source culture symbols in the source contents based on a preset metaphor decoupling encoder and a function alignment reconstructor to obtain optimal replacement symbols of the source culture symbols, and finally, carrying out function reconstruction on the optimal replacement symbols of the source culture symbols. And on the basis of the value invariant, the target emotion vector and the optimal replacement symbol, generating an interpretation text under the target cultural context through semantic splicing. Compared with the prior art, the method improves the expression cross-border translation effect of cross-border brands for overseas promotion.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computer language translation. More particularly, the present application relates to a semantic field reconstruction and emotional transfer translation method and system for cross-cultural marketing. BACKGROUND

[0002] In the era of digital marketing with globalization and localization, brands in different regions / countries need to achieve the dual goals of value consistency and emotional resonance in a multicultural context to meet the real needs and cultural identity of the target group. With the development of large language models and cross-modal representation learning, cross-cultural marketing has evolved from coarse-grained machine translation adaptation to fine-grained cultural semantic decoding and reconstruction. In the prior art, there are mainly the following ways: (1) Direct translation of source language advertisements into target language through neural machine translation (NMT) and supplemented by keyword replacement. (2) According to the preset content conversion rules in Hofstede's Cultural Dimensions Theory model, the source language is interpreted into the target language. (3) Use the LIWC (Linguistic Inquiry and Word Count) model to adjust the emotional expression polarity in cross-language scenarios. (4) Use the multilingual CLIP (Contrastive Language-Image Pre-training) model to generate localized languages for its graphic and semantic.

[0003] The above methods have the following defects: first, the translation results have cultural representation flattening. The above two methods mainly simplify cultural differences into word or emotional mapping labels, and cannot reconstruct the topological structure of cultural semantic fields (for example, "free" is associated with different networks in the East and West), which will lead to distortion of the brand core value of the advertising brand promotion language. Second, only the emotional expression intensity is adjusted, and the unique expression differences of different cultures are not adapted, which will cause cognitive conflicts and weaken the affinity of brand overseas marketing. Third, there is a loss of cultural metaphors. The deep symbols are stripped of their social functions in the translation process, for example, "red envelope" is translated as "red envelope", which is superficial and even culturally offensive. Fourth, the rule base of the above methods is static and lagging, and the preset mapping cannot respond to semantic evolution caused by Z generation language and network popular words, resulting in a disconnect between brand information and user cognition.

[0004] In summary, the prior art mainly has the problem of poor translation effect. SUMMARY

[0005] To solve the above technical problem of poor translation effect, the present application discloses a semantic field reconstruction and emotional transfer translation method and system for cross-cultural marketing.

[0006] In a first aspect, the present application discloses a semantic field reconstruction and emotional transfer translation method for cross-cultural marketing, comprising: establishing a cultural semantic field of a target brand and defining a value invariant of the target brand based on the cultural semantic field; in response to input of source content of the target brand, mapping the source content using a preset emotional transfer operator to obtain a target emotional vector; decoupling and functionally reconstructing source cultural symbols in the source content based on a preset metaphor decoupling encoder and a function alignment reconstructor to obtain optimal replacement symbols of the source cultural symbols; based on the value invariant, the target emotional vector and the optimal replacement symbols, generating an interpretive text in a target cultural context through semantic splicing.

[0007] Beneficial effects: The method constructs a cultural semantic field to refine the cultural modeling granularity, and defines a value invariant of the target brand using the cultural semantic field to accurately find "semantic anchor points" supporting subsequent semantic splicing, ensuring that the generated interpretive text does not deviate from the core culture of the target brand. On this basis, the method maps the source content using an emotional transfer operator to obtain a target emotional vector as a "style controller" to determine the tone and emotion of the generated text. As for the optimal replacement symbol, it is similar to a mandatory keyword, which is used to require the text to generate corresponding interpretive text around the symbol. Compared with the prior art, the method can reconstruct the topological structure of the cultural semantic field, adapt to different cultural expression differences through emotional transfer operator mapping, and decouple and reconstruct cultural metaphors through a metaphor decoupling encoder and a function alignment reconstructor, which can eliminate the defects in the prior art, thereby improving the cross-border translation effect of the language used by the cross-border brand for overseas promotion.

[0008] Preferably, the cultural semantic field of multiple languages is established, comprising: constructing a cross-cultural corpus covering multiple regional languages; extracting cultural semantic fields of the target brand in different cultures from the cross-cultural corpus using a contrastive learning method.

[0009] Preferably, the value invariant of the target brand is defined based on the cultural semantic field, comprising: determining a public intersection area of multiple languages in the cultural semantic field; calculating the Euclidean distance between each semantic vector in the public intersection area and the original brand core value vector; selecting a semantic vector with the smallest Euclidean distance as the value invariant.

[0010] Preferably, the specific algorithm for selecting a semantic vector with the smallest Euclidean distance as the value invariant is:

[0011] In the formula, represents the value invariance, represents the variable value that minimizes the target function, represents each semantic vector in the public intersection area, represents the original brand core value vector, represents the cultural semantic field of the language, represents the public intersection area, represents the Euclidean norm.

[0012] Preferably, the preset emotional transfer operator is used to map the source content, including: Based on the emotional expression base of the source culture and the target culture, a cross-cultural emotional transfer matrix is constructed as an emotional transfer operator.

[0013] Preferably, the metaphor decoupling encoder is configured to perform the following steps: Establish a vector representation mapping of the source cultural symbol; Decompose the source cultural symbol into surface form feature components and deep function feature components.

[0014] Preferably, the function alignment reconstructor is configured to perform the following steps: Receive the deep function feature component; In the symbol library of the target culture, calculate the similarity between the function feature component and the deep function feature component of the candidate symbol; Select the candidate symbol with the highest similarity as the best replacement symbol.

[0015] Preferably, after generating the literal translation text under the target cultural context, the method of the present application further comprises: Map the literal translation text back to the semantic space; Calculate the distance sum of the literal translation texts in different target languages in the shared semantic space as the consistency loss; If the consistency loss exceeds the preset threshold, return to the step of establishing the cultural semantic field of the target brand.

[0016] Beneficial effects: Through the above technical solution, if the consistency loss exceeds the preset threshold, return to the step of establishing the cultural semantic field of the target brand. This way can iteratively update the literal translation text to improve the cultural fit of the literal translation.

[0017] Preferably, based on the value invariance, the target emotional vector and the best replacement symbol, the literal translation text under the target cultural context is generated by semantic splicing, specifically: The value invariable, the target emotion vector and the best replacement symbol are input into a deep learning-based natural language processing model for semantic splicing to obtain the sense translation text.

[0018] In a second aspect, the present application discloses a semantic field reconstruction and emotion transfer translation system for cross-cultural marketing, comprising a processor and a memory, the memory storing computer program instructions, when the computer program instructions are executed by the processor, realizing the cross-cultural marketing semantic field reconstruction and emotion transfer translation method of the first aspect.

[0019] The present application has the following advantages: Compared with the prior art, the method of the present application can reconstruct the topological structure of the cultural semantic field, map the expression differences specific to different cultures through an emotion transfer operator, and decouple and reconstruct cultural metaphors through a metaphor decoupling encoder and a function alignment reconstructor, thereby eliminating the defects in the prior art and improving the cross-border translation effect of the language used by cross-border brands for overseas promotion. BRIEF DESCRIPTION OF DRAWINGS

[0020] The above and other objects, features and advantages of the exemplary embodiments of the present application will be more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which several embodiments of the present application are shown by way of example, and wherein like or corresponding elements refer to like or corresponding parts throughout several views, and in which: Figure 1 is a flowchart of the cross-cultural marketing semantic field reconstruction and emotion transfer translation method in the first embodiment of the present application; Figure 2 is a structural schematic diagram of the cross-cultural marketing semantic field reconstruction and emotion transfer translation system in the second embodiment of the present application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0022] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0023] Embodiment One As shown in the figure, the present embodiment discloses a cross-cultural marketing semantic field reconstruction and emotion transfer translation method, comprising: Figure 1 S10: establishing a cultural semantic field of a target brand and defining a value invariable of the target brand based on the cultural semantic field. S10: establishing a cultural semantic field of a target brand and defining a value invariable of the target brand based on the cultural semantic field.

[0024] wherein the step S10 comprises: S11: constructing a cross-cultural corpus covering multiple regional languages.

[0025] In the present embodiment, the multiple regional languages include more than 50 different regional or national languages, and the data source of the cross-cultural corpus can be Ethnologue or Glottolog.

[0026] S12: extracting a cultural semantic field for representing the target brand in different cultures from the cross-cultural corpus by using a contrastive learning method.

[0027] It should be explained that the contrastive learning method refers to similarity comparison query in the cross-cultural corpus based on a rule engine. Specifically, the target brand extracts the cultural semantic field by inputting the contrast conditions into the rule engine. It can be a market that the target brand intends to enter, such as “Southeast Asia”, “Middle East” or “Western Europe”, or a product / service that the target platform intends to promote, or a combination of the above two or a combination containing other contrast conditions.

[0028] Specifically, the step S12 extracts the cultural semantic field embedding by contrastive learning , wherein each word item is represented as a context-aware vector. Wherein, represents the cultural semantic field of the th language, represents a d-dimensional real vector space, and the number of dimensions depends on the number of conditions set in the rule engine. The information returned by the query is generated in the form of encoding , mapping the vocabulary of different languages into a high-dimensional geometric space, wherein each point represents the semantics of a word.

[0029] Further, the step S10 further comprises: S13: determining a public intersection area of multiple languages in the cultural semantic field.

[0030] In the present embodiment, the public intersection area represents the cross-cultural human commonality existing in all target cultures, which is a “consensus” in the psychological and cognitive level, and in the physical basic language description level, it includes commodity attributes such as softness, color or taste, etc.; it also includes commodity concepts such as cheap, expensive or new product, etc. In the emotional basic language description level, it includes basic emotions such as joy, anger, sadness and happiness, and it also includes family and love, such as maternal love, protective desire for young children or friendship, etc.

[0031] S14: calculating the Euclidean distance between each semantic vector in the public intersection area and the original brand core value vector.

[0032] S15: Select the semantic vector that minimizes the Euclidean distance as the value invariant.

[0033] The specific calculation expressions for steps S14-S15 above are as follows:

[0034] In the formula, Representing value invariants, This represents taking the value of the variable that minimizes the objective function. Represents the semantic vectors within the common intersection region. This represents the original brand's core value vector. Indicates the first The cultural semantic field of a language Indicates a common intersection area. This represents the Euclidean norm.

[0035] Through steps S10-S15 above, the method of this embodiment refines the granularity of cultural modeling by constructing the above-mentioned cultural semantic field, and uses the cultural semantic field to calculate the value invariant of the target brand, thereby accurately finding the "semantic anchor point" that supports subsequent semantic splicing, thus ensuring that the brand value remains consistent in each cultural semantic field after migration, so that the final generated paraphrased text will not deviate from the core culture of the target brand.

[0036] S20: In response to the input of source content from the target brand, the source content is mapped using a preset sentiment transfer operator to obtain the target sentiment vector.

[0037] Furthermore, step S20 above includes: Based on the emotional expression bases of the source culture and the target culture, a cross-cultural emotional transfer matrix is ​​constructed as an emotional transfer operator.

[0038] Specifically, regarding the construction of emotion expression bases: First, define the emotional expression basis for each culture. In the formula, culture The basis of emotional expression This indicates the basis for the expression of emotion. There are 3 basis vectors, each representing a unique archetype of emotional expression in that culture, such as semantic expressions of “subtlety,” “irony,” or “directness.”

[0039] Then, based on the above-mentioned emotional expression basis, an emotional transfer operator is designed, specifically as follows: Utilizing pre-built linear mappings Transforming the source culture's emotional expression into the target culture's expression:

[0040] In the formula, Indicate target culture The target sentiment vector in Representing the source culture The emotional expression vector in This represents a cross-cultural emotional transfer matrix.

[0041] Among them, the aforementioned cross-cultural emotion transfer matrix is ​​a transfer matrix learned through adversarial training, which is used to rotate / scale the expression of the source culture into the space of the target culture.

[0042] Through the above technical solution, the method of this embodiment obtains the target sentiment vector of the source content by mapping the sentiment transfer operator, which serves as a "style controller" to determine the tone and emotion of the generated text.

[0043] S30: Based on the preset metaphor decoupling encoder and functional alignment reconstructor, the source cultural symbols in the source content are decoupled and functionally reconstructed to obtain the best replacement symbol for the source cultural symbols.

[0044] The metaphor decoupling encoder is configured to perform the following steps: S31: Establish a vector representation mapping of source cultural symbols.

[0045] S32: Decompose the source cultural symbols into surface formal feature components and deep functional feature components.

[0046] Specifically, the expressions for steps S31-S32 at the algorithm layer are as follows:

[0047] The above formula represents the source cultural symbols Decomposed into surface-level feature components and deep functional characteristic components For example, the source cultural symbol is "red envelope," which, through the aforementioned metaphorical decoupling encoder, is decoupled into the surface formal feature component "red envelope" and the deep functional feature component "social reciprocity ritual." The source cultural symbol is "dragon," which, through the aforementioned metaphorical decoupling encoder, is decoupled into the surface formal feature component "flying snake-like creature" and the deep functional feature component "totem, auspiciousness, or power."

[0048] Furthermore, the aforementioned function alignment refactorer is configured to perform the following steps: S33: Receive deep functional feature components.

[0049] S34: In the symbol library of the target culture, calculate the similarity between the functional feature components and the deep functional feature components of the candidate symbols.

[0050] S35: selecting the candidate symbol with the highest similarity as the best replacement symbol.

[0051] Specifically, the specific algorithm expression for implementing the above steps S33-S35 is as follows:

[0052] In the formula, represents the best replacement symbol, represents the symbol that maximizes the similarity, represents the cosine similarity function, represents the functional feature component of the candidate symbol, represents that the best replacement symbol belongs to the target culture candidate symbol set in the target culture.

[0053] Through the technical solutions described in the above steps S33-S35, the metaphorical transfer of "change in shape but not in spirit" can be realized, and cognitive errors and stereotypes in culture can be reduced.

[0054] S40: based on the value invariants, the target emotional vector, and the best replacement symbol, generating an interpretive text in the target cultural context through semantic splicing.

[0055] Specifically, by inputting the value invariants, the target emotional vector, and the best replacement symbol into a deep learning-based natural language processing model for semantic splicing, an interpretive text is obtained. In the above natural language processing model, the value invariants serve as "semantic anchors", the target emotional vector serves as a "style controller", and the best replacement symbol serves as a "mandatory keyword", all of which guide the natural language processing model to generate an interpretive text that incorporates the best replacement symbol.

[0056] Through the above steps S10-S40, the method of the present embodiment can eliminate the defects existing in the prior art, thereby improving the cross-border translation effect of the language used by cross-border brands for overseas promotion.

[0057] Furthermore, in order to improve the translation effect, after the above step S40, the method of the present embodiment further comprises: S50: dynamically updating the cultural semantic field.

[0058] Specifically, the cultural semantic field embedding is updated in a sliding window through real-time interaction data of user-generated content (UGC) and social media:

[0059] In the formula, represents the cultural semantic field updated at the moment; represents Cultural semantic field of time; denotes a forgetting factor for controlling the retention ratio of historical information; denotes an encoding function for converting real-time interaction data collected at the time into a semantic vector.

[0060] Through the above step S50, the method of the embodiment has a dynamic updated cultural semantic field, and compared with the prior art based on a static rule base, the method of the embodiment has stronger adaptability to network traffic hotspots.

[0061] Furthermore, in order to further improve the accuracy of the interpretation of the method of the embodiment, after the above step S40, the method of the embodiment further comprises: S60: mapping the interpreted text back to the semantic space.

[0062] S70: calculating the distance sum of the interpreted texts in different target language versions in the shared semantic space as a consistency loss.

[0063] wherein the algorithm expression for calculating the consistency loss is:

[0064] wherein, denotes the consistency loss; denotes a summation symbol; denotes the cultural index serial number of two different languages; denotes a cross-language projection function for mapping vectors of different cultures to the same shared alignment space, ensuring that the brand value remains consistent during evolution; denotes a target brand the semantic vector of the target brand in the first cultural context; denotes a target brand the semantic vector of the target brand in the second cultural context.

[0065] For example, taking Nike as an example, in the culture of the United States, may be [Action, Individualism, Achievement], and the semantic vector (i.e., the vector expression form of the interpreted text) generated by the above steps S10-S40 can be [Struggle, Progress, Win] in the culture of China.

[0066] S80: if the consistency loss exceeds a preset threshold, return to step S10.

[0067] Compared with the prior art, the technical advantages of the method of the embodiment are as follows:

[0068] Further, after the above-mentioned interpretation text generation, when the target brand generates a corresponding commercial, the video segment, audio, and the above-mentioned interpretation text can be multi-modal fragmented encoded, and then the multi-modal highlight synthesis technology with the publication number CN120321474B Chinese patent technology is used to output a highlight mixed commercial short film with the interpretation text.

[0069] Embodiment two As Figure 2 shown, the embodiment discloses a semantic field reconstruction and emotional migration translation system for cross-cultural marketing, which comprises a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, the semantic field reconstruction and emotional migration translation method for cross-cultural marketing recorded in embodiment one is realized.

[0070] The embodiment system also includes a communication interface and other components well known to those skilled in the art, the settings and functions of which are known in the art, and therefore will not be described here.

[0071] In the present application, the aforementioned memory can be any tangible medium containing or storing a program, which can be used or combined with an instruction execution system, device or instrument. For example, the computer readable storage medium can be any appropriate magnetic storage medium or magneto-optical storage medium, such as resistive random access memory RRAM (Resistive Random Access Memory), dynamic random access memory DRAM (Dynamic Random Access Memory), static random access memory SRAM (Static Random-Access Memory), enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), high bandwidth memory HBM (High-Bandwidth Memory), hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, module or both. Any such computer storage medium can be part of the device or accessible or connectable to the device. Any application or module described in the present application can be implemented using computer readable / executable instructions that can be stored or otherwise held by such computer readable medium.

[0072] In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three or more, etc., unless otherwise explicitly specified.

[0073] While the specification has illustrated and described various embodiments of the application, it will be clear to those of ordinary skill in the art that various changes, modifications, and substitutions can be made thereto without departing from the spirit and scope of the application. It is understood that in the process of practicing the application, various alternatives, modifications, and equivalents can be employed.

Claims

1. A semantic field reconstruction and emotional transfer translation method for cross-cultural marketing, characterized in that, The method comprises the following steps: establishing a cultural semantic field of a target brand and defining a value invariant of the target brand based on the cultural semantic field; in response to input of source content of the target brand, mapping the source content by using a preset emotional transfer operator to obtain a target emotional vector; decoupling and functionally reconstructing a source cultural symbol in the source content based on a preset metaphor decoupling encoder and a function alignment reconstructor to obtain an optimal replacement symbol of the source cultural symbol; generating an interpretative text under a target cultural context by semantic splicing based on the value invariant, the target emotional vector and the optimal replacement symbol. 2.The semantic field reconstruction and emotional transfer translation method of cross-cultural marketing according to claim 1, wherein, The method for establishing a cultural semantic field of multiple languages comprises the following steps: constructing a cross-cultural corpus covering multiple regional languages; extracting a cultural semantic field of the target brand in different cultures from the cross-cultural corpus by using a contrastive learning method. 3.The semantic field reconstruction and emotional transfer translation method of cross-cultural marketing according to claim 1, wherein, Defining a value invariant of the target brand based on the cultural semantic field comprises the following steps: determining a public intersection area of multiple languages in the cultural semantic field; calculating the Euclidean distance between each semantic vector in the public intersection area and an original brand core value vector; selecting a semantic vector with the minimum Euclidean distance as the value invariant. 4.The semantic field reconstruction and emotional transfer translation method of cross-cultural marketing according to claim 3, characterized in that, The specific algorithm for selecting a semantic vector with the minimum Euclidean distance as the value invariant is as follows: In the formula, denotes the value invariance, denotes the variable value at which the target function is minimized, denotes each semantic vector within the common intersection region, denotes the original brand core value vector, denotes the cultural semantic field of the th language, denotes the common intersection region, denotes the Euclidean norm. 5.The semantic field reconstruction and emotional transfer translation method of cross-cultural marketing according to claim 1, wherein, mapping the source content by using a preset emotional transfer operator comprises the following steps: constructing a cross-cultural emotional transfer matrix as the emotional transfer operator based on emotional expression bases of source culture and target culture. 6.The semantic field reconstruction and emotional transfer translation method of cross-cultural marketing according to claim 1, wherein, The metaphor decoupling encoder is configured to perform the following steps: establishing a vector representation mapping of the source cultural symbol; decomposing the source cultural symbol into a surface form feature component and a deep function feature component. 7.The semantic field reconstruction and emotional transfer translation method of cross-cultural marketing according to claim 6, wherein, The function alignment reconstructor is configured to perform the following steps: receiving the deep function feature component; calculating the similarity between the function feature component of a candidate symbol and the deep function feature component in a symbol library of the target culture; selecting the candidate symbol with the highest similarity as the optimal replacement symbol. 8.The semantic field reconstruction and emotional transfer translation method of cross-cultural marketing according to claim 1, wherein, After generating the interpretative text under the target cultural context, the method further comprises the following steps: mapping the interpretative text back to a semantic space; calculating the distance sum of the interpretative texts in different target language versions in the shared semantic space as a consistency loss; if the consistency loss exceeds a preset threshold, returning to the step of establishing the cultural semantic field of the target brand. 9.The semantic field reconstruction and emotional transfer translation method of cross-cultural marketing according to claim 1, wherein, Generating the interpretative text under the target cultural context by semantic splicing based on the value invariant, the target emotional vector and the optimal replacement symbol specifically comprises the following steps: inputting the value invariant, the target emotional vector and the optimal replacement symbol into a natural language processing model based on deep learning for semantic splicing to obtain the interpretative text.

10. A semantic field reconstruction and emotional transfer translation system for cross-cultural marketing, characterized in that, The device comprises a processor and a memory, and the memory stores computer program instructions which, when executed by the processor, implement the semantic field reconstruction and emotional transfer translation method for cross-cultural marketing according to any one of claims 1-9.

Images