Word meaning sentence construction method and system based on speech thinking

Abstract

The invention provides a word meaning sentence construction method and system based on speech thinking, and relates to the technical field of natural language processing, and the method comprises the steps: 1, obtaining an initial statement input by a user, a preset rhetorical manipulation type and literature style preference, and analyzing a semantic core and an emotional tendency in the initial statement, based on the semantic core, calling a rhetorical knowledge base of a Uygur's speech thinking theory, including a metaphor mapping matrix, a symbol association network and a grammar deformation rule set, and matching an adaptive composite rhetorical structure; step 2, constructing a three-dimensional situation semantic field based on the semantic core and the emotional tendency, and extracting and acquiring initial weight parameters of each dimension of an author cognition dimension, a reader acceptance dimension and a text function dimension through multi-modal features; according to the method, word meaning sentence construction is performed in combination with a speech thinking theory, so that image coordination and language adaptation of candidate sentence patterns are realized, and language expression is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular to a method and system for constructing word meaning sentences based on speech thinking. Background Art

[0002] When students write essays, traditional word-meaning sentence-building techniques have some limitations. Traditional techniques lack flexibility when processing word-meaning sentence-building, making it difficult to fully adapt to diverse verbal thinking needs. For example, when a student wants to create a paragraph describing a mountaineering experience based on the three words "courage," "challenge," and "growth," the sentences generated by traditional techniques mostly follow fixed grammar and common collocations, producing relatively routine and uncreative sentences like "During the mountaineering process, I mustered the courage to challenge difficulties and ultimately achieved growth." Traditional techniques struggle to flexibly construct sentences that are rich in visual imagery and emotional tension based on verbal thinking.

[0003] Furthermore, traditional writing techniques lack a deep understanding and application of semantics, making them prone to monotonous or inappropriate sentence construction. When asked to form a sentence around the words "technology," "life," and "convenience," traditional writing techniques might produce sentences like "Technology makes our lives more convenient." However, in actual writing, students may want to incorporate specific life scenarios, highlighting the convenience brought by technology while also expressing their reflections on its development. Traditional writing techniques, lacking a deep understanding of semantics and a dynamic grasp of verbal thinking, struggle to construct sentences that are both relevant to the topic and possess depth and layering. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for constructing word meaning sentences based on language thinking, realize word meaning sentence construction based on language thinking, and improve the flexibility, appropriateness and expression depth of sentence formation.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0006] In a first aspect, a method for constructing word meaning sentences based on verbal thinking is provided, the method comprising:

[0007] Step 1: Obtain the user's initial sentence, preset rhetorical device types, and literary style preferences, and analyze the semantic core and emotional tendency of the initial sentence. Based on the semantic core, call on the rhetorical knowledge base of Vygotsky's theory of verbal thinking, including the metaphor mapping matrix, symbolic association network, and grammatical inflection rule set, to match the appropriate compound rhetorical structure;

[0008] Step 2: Based on the semantic core and emotional tendency, a three-dimensional contextual semantic field is constructed, and the initial weight parameters of each dimension of the author's cognitive dimension, the reader's acceptance dimension, and the text function dimension are obtained through multimodal feature extraction;

[0009] Step 3: Based on the initial weight parameters of each dimension, three detection points are determined in the 3D context semantic field. Dynamically changing polygons are generated according to the spatial coordinates. The dynamically changing polygons are meshed and analyzed. The vertex displacement, area change, and boundary curvature characteristics of each mesh unit are analyzed to generate adjustment values ​​reflecting the polygon morphological changes.

[0010] Step 4: Use the adjustment value to adjust the initial weight parameters of each dimension to obtain the adjusted weight parameters of each dimension;

[0011] Step 5: Based on the compound rhetorical structure and the adjusted weight parameters of each dimension, a set of candidate sentences including two rhetorical devices is generated, and the abstract level of the verbal thinking path is marked for each candidate sentence;

[0012] Step 6: Based on literary style preferences, the candidate sentence set is tested for rhythm and rhyme, image coordination is evaluated, and language adaptability is verified to generate a corresponding thinking path presentation.

[0013] Furthermore, the initial sentence input by the user, the preset rhetorical device type and literary style preference are obtained, and the semantic core and emotional tendency in the initial sentence are analyzed. Based on the semantic core, the rhetorical knowledge base of Vygotsky's theory of verbal thinking is called, including the metaphor mapping matrix, symbolic association network and grammatical inflection rule set, to match the adaptive compound rhetorical structure, including:

[0014] Perform dependency syntax analysis on the initial sentence, extract the predicate verb and the governing argument as the semantic core, and calculate the sentiment tendency intensity value through sentiment dictionary matching;

[0015] The semantic core is input into the metaphor mapping matrix of the Vygotsky rhetorical knowledge base, and the source domain-target domain mapping pairs that meet the threshold conditions are determined according to the emotional tendency intensity value to generate the initial metaphor set;

[0016] Each source domain symbol in the initial metaphor set is input into the symbolic association network, and the three-layer association nodes are recursively retrieved. The candidate symbol set is formed by combining cultural prototype symbols, historical allusion symbols and natural image symbols.

[0017] The grammatical inflection rule set is called to reorganize the candidate symbolic symbol set, including applying metonymy compression rules to the target domain and metaphor expansion rules to the source domain, to obtain a composite rhetorical structure including two rhetorical techniques, namely the core metaphor, auxiliary symbol and grammatical inflection marker.

[0018] Furthermore, based on the semantic core and emotional tendency, a three-dimensional contextual semantic field is constructed, and the initial weight parameters of each dimension of the author's cognitive dimension, the reader's acceptance dimension, and the text function dimension are obtained through multimodal feature extraction, including:

[0019] With the semantic core as the origin, the emotional tendency as the vertical axis, and the text theme relevance as the horizontal axis, a three-dimensional situational semantic field is constructed, including author intention, contextual information, and semantic association.

[0020] Based on the three-dimensional contextual semantic field, the author's writing history data, the pre-built reader portrait library and the functional labels of the complex rhetorical structure are extracted to generate feature vectors representing the author's cognitive dimension, the reader's acceptance dimension and the text's functional dimension;

[0021] The author cognition dimension feature vector, reader acceptance dimension feature vector and text function dimension feature vector are normalized respectively to obtain the corresponding author cognition dimension initial weight parameters, reader acceptance dimension initial weight parameters and text function dimension initial weight parameters.

[0022] Furthermore, based on the initial weight parameters of each dimension, three detection points are determined in the three-dimensional context semantic field. Dynamically changing polygons are generated according to the spatial coordinates. The dynamically changing polygons are meshed and analyzed. The vertex displacement, area change, and boundary curvature characteristics of each mesh unit are analyzed to generate adjustment values ​​reflecting the polygon morphological changes, including:

[0023] In the three-dimensional situational semantic field, based on the initial weight parameters of each dimension, the intersection of the author's cognitive dimension and emotional tendency is determined as the first detection point, the intersection of the reader's acceptance dimension and the semantic core is determined as the second detection point, and the intersection of the text function dimension and metaphor mapping is determined as the third detection point.

[0024] Based on the spatial coordinates of the three detection points, dynamically changing polygons are generated according to preset time series parameters, including sentence advancement speed and rhetorical conversion frequency, and the polygon vertices move in three-dimensional space according to the time parameters;

[0025] For dynamically changing polygons, a uniform meshing method is used to segment them into several small mesh units, and the coordinates of each mesh unit vertex at each time point are recorded;

[0026] Based on the recorded grid cell vertex coordinates, the difference between vertex coordinates at adjacent time points is calculated to obtain vertex displacement, the rate of change of grid cell area per unit time is counted, and the curvature value of the polygon boundary is calculated;

[0027] The vertex displacement, area change rate and polygon boundary curvature value are normalized to generate an adjustment value reflecting the polygon morphology change.

[0028] Furthermore, the initial weight parameters of each dimension are adjusted using the adjustment value to obtain the adjusted weight parameters of each dimension, including:

[0029] Input the vertex displacement in the geometric feature into the weight adjustment rule of the author cognitive dimension to generate a first intermediate parameter, and perform incremental and decrement corrections on the initial weight parameter of the author cognitive dimension according to the direction and magnitude of the displacement vector in the first intermediate parameter to obtain the author cognitive dimension correction weight;

[0030] Input the area change rate into the weight adjustment rule of the reader acceptance dimension, generate a second intermediate parameter according to the monotonic increase and decrease trend, and dynamically adjust the proportional coefficient of the reader acceptance dimension weight parameter based on the change trend indicator in the second intermediate parameter to obtain the reader acceptance dimension correction weight;

[0031] Input the boundary curvature value into the weight adjustment rule of the text function dimension, generate a third intermediate parameter according to the straight-curvature characteristic, and use the curvature quantization value in the third intermediate parameter to implement a smoothness strength constraint on the text function dimension weight parameter to obtain the text function dimension correction weight;

[0032] The weighted fusion of the author cognition dimension correction weight, the reader acceptance dimension correction weight and the text function dimension correction weight is performed to generate a set of adjusted dimension weight parameters that reflects the dynamic changes of polygon morphology.

[0033] Furthermore, based on the compound rhetorical structure and the adjusted weight parameters of each dimension, a set of candidate sentences including two rhetorical devices is generated, and the abstract level of the verbal thinking path is marked for each candidate sentence, including:

[0034] The adjusted weight parameters of the author's cognitive dimension, reader's acceptance dimension, and text function dimension are integrated with the core metaphors, auxiliary symbols, and grammatical inflection markers in the complex rhetorical structure, and a combination set of rhetorical elements that meets the multi-dimensional requirements is generated through sorting.

[0035] Based on the combination set of rhetorical elements, two candidate sentence sets of rhetorical devices, namely, the combination of metaphor and symbol, and the combination of metaphor and grammatical deformation, are generated;

[0036] For each sentence in the candidate sentence set, the nested logic and semantic advancement path of the rhetorical devices are analyzed, and according to the hierarchical division standard of Vygotsky's verbal thinking theory, the abstract level of the verbal thinking path from single rhetoric mapping to multi-rhetoric collaboration is marked for each sentence.

[0037] Furthermore, based on literary style preferences, the candidate sentence sets are tested for rhythm and rhyme, image coordination, and language adaptability, generating corresponding thought path presentations, including:

[0038] Based on the rhyme characteristic parameters in literary style preferences, including level and oblique tone distribution, syllable density, and stress period, the syntactic structure of the candidate sentence is divided into syllables, and the standard deviation of the syllable drop value and pause length is calculated to generate the rhythm adaptability score of the candidate sentence.

[0039] Based on the rhythmic fit score, the cultural archetype symbol library in Vygotsky's symbolic association network is used to calculate the semantic relevance of the source and target domain images of the metaphorical mapping in the sentence. The image conflict index is then constructed based on the emotional tendency intensity value to determine the symbolic symbol combination scheme that conforms to the style preference and has a harmonious rhythm.

[0040] Based on the symbolic symbol combination scheme and the weight parameters of the text function dimension, the candidate sentences are matched with the register features to generate a comprehensive evaluation result reflecting the adaptability of the language style;

[0041] The comprehensive evaluation results of rhythm adaptation score, symbol combination scheme and style adaptation are integrated to generate a thinking path presentation including rhetorical coordination, style deviation and cognitive load value.

[0042] The second aspect is a system for constructing word meaning sentences based on verbal thinking, including:

[0043] The semantic analysis module is used to analyze the semantic core and emotional tendency of the initial sentence input by the user, and match the appropriate complex rhetorical structure based on Vygotsky's theory of verbal thinking;

[0044] The semantic construction module is used to construct a three-dimensional contextual semantic field including author, reader and text functions based on the semantic core and emotional tendency, and extract the initial weight parameters of each dimension;

[0045] The morphological analysis module is used to determine the detection points in the three-dimensional context semantic field and generate dynamic polygons. It analyzes the morphological change characteristics through mesh cutting and generates adjustment values.

[0046] A parameter adjustment module is used to dynamically adjust the initial weight parameters of each dimension using the adjustment values ​​generated by the dynamic polygon;

[0047] The sentence generation module is used to generate a set of candidate sentences including multiple rhetorical techniques based on the compound rhetorical structure and the adjusted weight parameters, and annotate the corresponding abstract thinking path level;

[0048] The adaptation evaluation module is used to conduct a comprehensive evaluation of the rhythm, image coordination and language adaptability of candidate sentences based on the user's literary style preferences, and obtain the final thinking path presentation.

[0049] According to a third aspect, a computing device includes:

[0050] one or more processors;

[0051] The storage device is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0052] In a fourth aspect, a computer-readable storage medium stores a program, which implements the method when executed by a processor.

[0053] The above solution of the present invention includes at least the following beneficial effects:

[0054] Breaking through the limitations of traditional sentence formation that relies on grammatical rules, this system deeply understands the semantic connotations and logical connections of word meanings based on verbal thinking patterns. Based on fragmented word meanings entered by the user, it can quickly analyze the semantic and logical relationships between each word meaning and automatically construct complete sentences that conform to human language expression habits. This significantly improves sentence formation efficiency and avoids incoherent sentences and semantic contradictions. By simulating the human verbal thinking process, it not only accurately captures the surface meaning of word meanings but also explores deeper semantic information and contextual connections, making the generated sentences more semantically coherent and reasonable. When dealing with complex semantic situations involving polysemous and synonymous words, it accurately selects appropriate word meanings based on the context, improving the semantic accuracy and fluency of sentences and enhancing its ability to understand and process natural language. It is widely applicable to various language scenarios and fields, from daily communication to professional terminology to literary works, dynamically adjusting sentence formation strategies based on different contexts and needs. It supports user-defined parameters such as sentence formation style, tone, and sentence structure to meet diverse language generation needs. Its excellent scalability and versatility provide powerful technical support for intelligent writing, machine translation, human-computer interaction, and other fields. This method simulates the real human speech thinking process. For language learners, by observing and learning the sentence formation process and results generated based on this method, they can better understand how word meanings are used in different contexts, master the sentence formation rules and expression habits of natural language, assist in language learning and training, and effectively improve language learning effects and language application ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of a method for constructing word meaning sentences based on verbal thinking provided by an embodiment of the present invention.

[0056] Figure 2 It is a schematic diagram of a word meaning sentence construction system based on language thinking provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0058] like Figure 1 As shown, an embodiment of the present invention proposes a method for constructing word meaning sentences based on verbal thinking, the method comprising the following steps:

[0059] Step 1: Obtain the user's initial sentence, preset rhetorical device types, and literary style preferences, and analyze the semantic core and emotional tendency of the initial sentence. Based on the semantic core, call on the rhetorical knowledge base of Vygotsky's theory of verbal thinking, including the metaphor mapping matrix, symbolic association network, and grammatical inflection rule set, to match the appropriate compound rhetorical structure;

[0060] Step 2: Based on the semantic core and emotional tendency, a three-dimensional contextual semantic field is constructed, and the initial weight parameters of each dimension of the author's cognitive dimension, the reader's acceptance dimension, and the text function dimension are obtained through multimodal feature extraction;

[0061] Step 3: Based on the initial weight parameters of each dimension, three detection points are determined in the 3D context semantic field. Dynamically changing polygons are generated according to the spatial coordinates. The dynamically changing polygons are meshed and analyzed. The vertex displacement, area change, and boundary curvature characteristics of each mesh unit are analyzed to generate adjustment values ​​reflecting the polygon morphological changes.

[0062] Step 4: Use the adjustment value to adjust the initial weight parameters of each dimension to obtain the adjusted weight parameters of each dimension;

[0063] Step 5: Based on the compound rhetorical structure and the adjusted weight parameters of each dimension, a set of candidate sentences including two rhetorical devices is generated, and the abstract level of the verbal thinking path is marked for each candidate sentence;

[0064] Step 6: Based on literary style preferences, the candidate sentence set is tested for rhythm and rhyme, image coordination is evaluated, and language adaptability is verified to generate a corresponding thinking path presentation.

[0065] In this embodiment of the present invention, by obtaining the user's initial sentence input, presetting the type of rhetorical device, and literary style preferences, the user's personalized creative needs can be precisely tailored to avoid blindly generating content that does not meet expectations. Analysis of the semantic core and emotional tendencies of the initial sentence, combined with the rhetorical knowledge base of Vygotsky's theory of verbal thinking, can deeply explore semantic connotations and quickly and accurately match complex rhetorical structures. This not only significantly improves the accuracy of rhetorical device adaptation, but also closely integrates rhetoric with semantics and emotion, enhancing the expressiveness and appeal of sentences, making the generated text more artistic and unique in terms of linguistic expression. Constructing a three-dimensional contextual semantic field and extracting multimodal features to determine the initial weight parameters for each dimension fully considers the multiple factors of the author, reader, and the function of the text itself during the text creation process. Starting from the author's cognitive dimension, the author's intention and logical thinking can be reflected; the reader's acceptance dimension ensures that the text meets the audience's understanding and preferences; and the text's functional dimension ensures that the content meets specific expression needs. This multi-dimensional consideration breaks the limitations of a single perspective, making the generated text more comprehensive and three-dimensional in semantic expression, better adaptable to different application scenarios, and effectively improving the text's applicability and dissemination effectiveness.

[0066] By identifying detection points within a three-dimensional contextual semantic field and generating dynamically changing polygons, and then analyzing their morphological variations through mesh cutting, a dynamic adjustment mechanism is introduced for text creation. This spatial geometry-based analysis method concretizes abstract semantic relationships. By quantitatively analyzing polygonal morphological variations, it can keenly capture the dynamic changes and interactions of various factors within the semantic field. The generated adjustment values ​​accurately reflect the changing trends of the semantic field, allowing the text to be flexibly adjusted during the creation process based on the dynamic changes in the semantic environment, maintaining semantic coherence and logic, and enhancing the text's dynamic adaptability.

[0067] Adjustment values ​​are used to adjust the initial weight parameters of each dimension, enabling fine-grained control of factors influencing text creation. The importance of the three dimensions of author cognition, reader acceptance, and text function varies across different text requirements and contexts. Dynamically adjusting the weight parameters can highlight the role of key dimensions and balance the relationships between them, ensuring that the generated text achieves the desired effect across different emphases. Based on the compound rhetorical structure and the adjusted weight parameters, a set of candidate sentences is generated, and the abstraction level of the verbal thought process is annotated, providing a rich and diverse selection for text creation. The combination of the compound rhetorical structure and weight parameters ensures the high quality of the candidate sentences in terms of rhetorical expression and semantic adaptability. Furthermore, the annotation of the abstraction level of the verbal thought process allows users to clearly understand the creative thinking and logical flow of each candidate sentence, facilitating selection based on their needs. Comprehensive testing and evaluation of the candidate sentence set based on literary style preferences ensures strict control of the text's stylistic consistency and quality. Rhythm detection makes the text more rhythmic and musical when read aloud, improving the fluency and appeal of the language; image coordination assessment ensures that the images in the text echo each other and are harmonious and unified, enhancing the text's ability to create artistic conception; style adaptability verification ensures that the text conforms to specific literary styles and application scenario requirements.

[0068] In a preferred embodiment of the present invention, the above step 1, obtaining the initial sentence input by the user, the preset rhetorical device type and the literary style preference, and analyzing the semantic core and emotional tendency in the initial sentence, and based on the semantic core, calling the rhetorical knowledge base of Vygotsky's theory of verbal thinking, including the metaphor mapping matrix, the symbolic association network and the grammatical deformation rule set, to match the adapted compound rhetorical structure, may include:

[0069] Step 110: performing dependency syntax analysis on the initial sentence, extracting the predicate verb and the dominant argument as the semantic core, and calculating the sentiment tendency intensity value through sentiment dictionary matching;

[0070] Step 111: input the semantic core into the metaphor mapping matrix of the Vygotsky rhetoric knowledge base, determine the source domain-target domain mapping pairs that meet the threshold conditions based on the sentiment tendency intensity value, and generate an initial metaphor set;

[0071] Step 112: Input each source domain symbol in the initial metaphor set into the symbolic association network, recursively search the three-layer association nodes, and combine cultural prototype symbols, historical allusion symbols, and natural image symbols to form a candidate symbol set;

[0072] Step 113, calling the grammatical inflection rule set to reorganize the candidate symbolic symbol set, including applying the metonymy compression rule to the target domain and the metaphor expansion rule to the source domain, to obtain a composite rhetorical structure including two rhetorical devices, namely the core metaphor, the auxiliary symbol and the grammatical inflection identifier.

[0073] In an embodiment of the present invention, the initial sentence is structurally decomposed using dependency parsing techniques in natural language processing. Taking the sentence "Warm sunlight caresses the sleeping earth" as an example, the analysis identifies the dependencies between sentence components, determining that "caress" is the predicate verb, "sunlight" is its subject (agent argument), and "earth" is its object (patient argument). These three elements are then extracted as the semantic core. Using a pre-built sentiment lexicon, each word in the initial sentence is matched with sentiment words in the lexicon. Each word in the sentiment lexicon is labeled as positive, negative, or neutral and assigned a corresponding sentiment intensity value (e.g., positive words range from 0.5 to 1, negative words range from -1 to -0.5, and neutral is 0). A weighted sum is then taken of the words in the sentence, for example, "warm" (positive, intensity 0.7), "caress" (neutral, intensity 0), "sleeping" (neutral, intensity 0), and "earth" (neutral, intensity 0). The resulting sentiment intensity value is 0.7 × 1 = 0.7, confirming that the sentence as a whole has a positive sentiment tendency.

[0074] The extracted semantic cores (e.g., "sunshine," "caress," and "earth" in "sunshine gently caresses the earth") are input into the metaphor mapping matrix of the Vygotsky rhetorical knowledge base. The metaphor mapping matrix stores a large number of semantic concepts and their corresponding metaphorical mappings. Taking "sunshine" as an example, the matrix may contain mappings such as "sunshine-hope" and "sunshine-warm emotion." Based on the calculated sentiment intensity value of 0.7, a threshold condition is set (e.g., positive mappings with an intensity greater than 0.5 are included in the screening). This screens out matching source-target domain mapping pairs, such as "sunshine-hope," to generate the initial metaphor set.

[0075] The source domain symbols (e.g., "sunshine") from the initial metaphor set are input into a symbolic association network. This network uses semantic symbols as nodes, and the nodes form a network structure through semantic associations and connections across cultural, historical, and natural dimensions. Starting with the "sunshine" node, a recursive search is performed on three levels of associated nodes. The first level searches for symbols directly related to "sunshine," such as "sunflower," which grows toward the sun. The second level searches for symbols related to "sunflower," such as "loyalty" (sunflowers often symbolize loyalty). The third level searches for symbols related to "loyalty," such as "guardian knight" (chivalry includes loyalty). By combining cultural archetypal symbols (e.g., the mythical sun god symbolizes power), historical allusions (e.g., "working at sunrise" symbolizes diligence), and natural imagery (e.g., the temporal connection between "sunset glow" and "sunshine"), a candidate symbol set containing a variety of symbols is formed.

[0076] The grammatical inflection rule set is applied to the candidate symbolic symbol set. For the target domain (e.g., "hope"), metonymy compression is applied, simplifying complex concepts through metonymy. For example, "hope" is metonymically transformed into "lighthouse" (a lighthouse often points the way and symbolizes hope). For the source domain (e.g., "sunshine"), metaphorical expansion is applied, expanding the concept by adding modifiers or associated imagery, as in "golden sunshine." Through this reorganization, metaphors (e.g., "sunshine-hope") and symbols (e.g., "lighthouse" symbolizes "hope") are combined and annotated with grammatical inflection markers (e.g., the specific operational methods of metonymy compression and metaphorical expansion). This results in a composite rhetorical structure, such as "golden sunshine, like a guiding lighthouse, illuminates our path forward." The core metaphor is "sunshine-hope," and the auxiliary symbol is "lighthouse." The grammatical inflection markers reflect the adjustment process between the source and target domains.

[0077] Dependency parsing accurately identifies the semantic core, ensuring rhetorical manipulation focuses on the key content of a sentence and avoiding rhetoric deviation from the main theme. Calculating sentiment intensity values ​​can detect the emotional tone of the text, providing emotional guidance for the selection of metaphors and symbols, ensuring that rhetorical expression aligns with emotion and enhances the text's appeal. For example, selecting positive metaphors and symbols in sentences expressing positive emotions can enhance and accurately convey the text's emotional expression. Filtering metaphor mapping pairs based on semantic core and sentiment orientation can quickly match the most appropriate metaphorical relationships from a vast pool of metaphor resources, improving the accuracy and efficiency of metaphor selection. Filtering based on sentiment thresholds ensures that metaphorical expression aligns with the sentiment of the text, making metaphors a powerful tool for enhancing emotional expression and enhancing the emotional resonance of the text while conveying both semantic meaning. Recursive search of the symbolic association network and the integration of multi-dimensional symbols enrich the symbolic connotations of the text. Exploring symbolic symbols from multiple levels ensures that the text not only reflects superficial semantics but also contains profound cultural, historical, and natural implications, enhancing the text's cultural heritage and artistic depth. Furthermore, this rich set of symbolic symbols provides diverse options for rhetorical creation, making the text more flexible and vivid. The application of grammatical inflection rules optimizes and reorganizes candidate symbolic symbols, integrating rhetorical devices more naturally into the text through metonymy compression and metaphor expansion. The formation of compound rhetorical structures skillfully combines metaphor and symbolism to create novel and unique forms of expression, enhancing the text's creativity and expressiveness.

[0078] In a preferred embodiment of the present invention, the above step 2, based on the semantic core and emotional tendency, constructs a three-dimensional contextual semantic field, and obtains the initial weight parameters of each dimension of the author's cognitive dimension, the reader's acceptance dimension, and the text function dimension through multimodal feature extraction, which may include:

[0079] Step 220 , with the semantic core as the origin, the emotional tendency as the vertical axis, and the text topic relevance as the horizontal axis, construct a three-dimensional contextual semantic field including author intention, contextual information, and semantic association;

[0080] Step 221: Based on the three-dimensional contextual semantic field, extract the author's writing history data, the pre-built reader portrait library, and the functional labels of the complex rhetorical structure to generate a feature vector representing the author's cognitive dimension, the reader's acceptance dimension, and the text's functional dimension;

[0081] In step 222, the author cognition dimension feature vector, the reader acceptance dimension feature vector, and the text function dimension feature vector are normalized respectively to obtain the corresponding author cognition dimension initial weight parameters, reader acceptance dimension initial weight parameters, and text function dimension initial weight parameters.

[0082] In an embodiment of the present invention, the semantic core extracted in step 1 is used as the origin of the three-dimensional space. For example, for the sentence "The dewdrops in the morning sparkle on the flower petals", "the dewdrops sparkle" is the semantic core, and the space is constructed with this as the starting point. Taking the emotional tendency as the vertical axis direction, if the emotional tendency strength value of the sentence is calculated to be 0.6 (positive emotion) through step 110, the upward vertical axis represents the enhancement of positive emotion, and the downward vertical axis represents the enhancement of negative emotion. The text topic relevance is used as the horizontal axis direction, and the relevance value is determined by calculating the degree of matching between the vocabulary in the sentence and the preset topic vocabulary library. Assuming that the theme is "the beauty of nature", the frequency and weight of the words such as "morning", "dewdrops" and "petals" in the sentence are counted in the topic vocabulary library, and the topic relevance value of the sentence is assigned to 0.7.

[0083] Within the defined three-dimensional space, the author's intent, contextual information, and semantic associations are populated. Author's intent can be inferred from the user's input descriptions and historical creative habits. For example, if a user has repeatedly created sentences praising nature, it can be inferred that the current creation's intent is also related to praising nature. Contextual information includes background information such as the time and setting of the sentence creation. If the sentence was created in the early morning, then the time information "early morning" is part of the contextual information. Semantic associations are the connections between words in the sentence and with external semantic concepts. For example, "dewdrops" are associated with semantic concepts such as "fresh" and "pure." Integrating this information into the three-dimensional space constructs a complete three-dimensional contextual semantic field.

[0084] Extract information from the author's writing history, such as the author's past writing style (e.g., ornate, simple), commonly used vocabulary, and rhetorical preferences. Count the number of metaphors used in the author's past 10 works, as well as the number of frequently used words and their frequency of occurrence. This data is then quantified and converted into numerical features. For example, the frequency of metaphor use is coded on a scale of 1-5 (1 indicating rare use, 5 indicating frequent use). Commonly used words are assigned weights based on their frequency of occurrence. This ultimately creates a multidimensional numerical vector, which serves as a feature vector representing the author's cognitive dimension. Based on a pre-built reader profile database, analyze information such as the target readership's age, gender, educational level, and reading preferences. For example, assume the target readership is college students aged 18-25 who prefer poetic and imaginative texts. These features are quantified, for example, encoding the age range as a specific numerical value and assigning corresponding weights to reading preferences such as "poetic" and "imaginative." This creates a multidimensional vector representing the feature vector representing the reader's perception dimension.

[0085] According to the functional labels of compound rhetorical structures, such as metaphors for figurative expression and symbols for deepening connotations, the functions of the text are classified and quantified. The rhetorical devices used in the sentences and their corresponding functional labels are counted, and each functional label is given a certain weight score. For example, the weight of the metaphor functional label is 0.4, and the weight of the symbol functional label is 0.3. These scores are combined into a multidimensional vector as the feature vector of the text functional dimension. For each value in the feature vector of the author's cognitive dimension, the normalization formula is used for processing. Assume that the feature vector of the author's cognitive dimension is , find the maximum value in this vector and minimum value , through the formula For each element Perform calculations and map all values ​​to the interval [0, 1] to obtain a normalized vector. The values ​​of each element of this vector are the initial weight parameters of the author's cognitive dimension. Similarly, process the reader's acceptance dimension feature vector, find the maximum and minimum values ​​in the reader's acceptance dimension feature vector, and use the above normalization formula to map each element in the vector to the interval [0, 1]. The obtained element value is the initial weight parameter of the reader's acceptance dimension. Following the same normalization process, operate on the text function dimension feature vector, normalize its elements to the interval [0, 1], and the obtained value is the initial weight parameter of the text function dimension.

[0086] Constructing a three-dimensional contextual semantic field with the semantic core as the origin provides a clear semantic framework for text creation, ensuring that creation remains centered on the core content and avoiding semantic deviation. Using sentiment and thematic relevance as axes can intuitively reflect the text's emotional direction and thematic fit, facilitating a precise grasp of the text's overall tone and direction. By incorporating author intent, contextual information, and semantic associations, the semantic field is enriched with semantic information, laying the foundation for more accurate understanding and text generation, facilitating the production of high-quality text that meets the user's creative intent and contextual needs. Feature vectors are extracted from three dimensions: author cognition, reader acceptance, and text function, comprehensively considering key influencing factors in the text creation process. Extracting feature vectors from the author cognition dimension by analyzing the author's writing history data provides a deeper understanding of the author's writing style and preferences, ensuring that the generated text better meets the author's personalized needs. Extracting feature vectors from the reader acceptance dimension based on a reader profile database facilitates the creation of text that caters to the target readership, enhancing its readability and appeal. Extracting feature vectors from the text function dimension based on compound rhetorical structure functional labels clarifies the text's functional positioning, making the use of rhetorical techniques more targeted and improving the text's expressive effectiveness and practicality. Normalizing the feature vectors of each dimension brings feature data of varying magnitudes and ranges onto the same scale, eliminating dimensional differences between the data. This makes the initial weight parameters of each dimension comparable and computable, facilitating flexible adjustment of the weights of each dimension based on specific needs during the text creation process.

[0087] In a preferred embodiment of the present invention, step 3 above, based on the initial weight parameters of each dimension, determines three detection points in the three-dimensional context semantic field, generates dynamically changing polygons according to the spatial coordinates, performs mesh cutting on the dynamically changing polygons, analyzes the vertex displacement, area change, and boundary curvature characteristics of each mesh unit, and generates adjustment values ​​reflecting the polygon morphological changes, which may include:

[0088] Step 330: In the three-dimensional contextual semantic field, based on the initial weight parameters of each dimension, determine the intersection of the author's cognitive dimension and the emotional tendency as the first detection point, the intersection of the reader's acceptance dimension and the semantic core as the second detection point, and the intersection of the text function dimension and the metaphorical mapping as the third detection point.

[0089] Step 331 : Based on the spatial coordinates of the three detection points and according to preset time series parameters, including sentence advancement speed and rhetorical conversion frequency, a dynamically changing polygon is generated, and the polygon vertices move in the three-dimensional space according to the time parameters;

[0090] Step 332: segment the dynamically changing polygon using a uniform meshing method to obtain a number of small mesh units, and record the coordinates of each mesh unit vertex at each time point;

[0091] Step 333: Based on the recorded grid cell vertex coordinates, the difference between the vertex coordinates at adjacent time points is calculated to obtain the vertex displacement, the rate of change of the grid cell area per unit time is calculated, and the curvature value of the polygon boundary is calculated;

[0092] In step 334 , the vertex displacement, area change rate, and polygon boundary curvature value are normalized to generate an adjustment value reflecting the change in polygon morphology.

[0093] In this embodiment of the present invention, it is assumed that the author's cognitive dimension vector is represented as ,in 、 、 These correspond to the quantitative values ​​of sub-dimensions such as cognitive depth, uniqueness of perspective, and knowledge density (the value range is [0, 1]).

[0094] The sentiment tendency dimension vector is expressed as ,in 、 、 Quantitative values ​​corresponding to emotion intensity, polarity, and complexity.

[0095] Reader accepts dimension vector , including the dimensions of difficulty of understanding, resonance, and memorization; semantic core vector , including topic clarity, concept relevance, and semantic cohesion; text function vector , including information transmission efficiency, persuasiveness, and aesthetic value; metaphor mapping vector , including metaphor richness, mapping consistency, and abstraction level.

[0096] The first detection point (the intersection of the author's cognition and emotional tendency): Set the initial weight parameter to and , intersection coordinates , that is, weighted sum of two dimension vectors, where the weight reflects the importance of the initial dimension. For example, if =0.6, =0.4, then .

[0097] The second detection point (the intersection of reader acceptance and semantic core): Similarly, the weight is and , intersection .

[0098] The third detection point (the intersection of text function and metaphor mapping): weight and , intersection .

[0099] In traditional detection point calculation, weight parameters are usually fixed values. Particle swarm optimization algorithm is now introduced to dynamically adjust weights to adapt to the dynamic changes in text semantics. Each particle represents a set of weight parameters. , must meet 、 、 Taking "semantic field energy entropy" as the goal, the smaller the energy entropy, the more orderly the dimensional interaction, and the more the detection point can reflect the semantic core. The particle updates the weight according to "position = position + speed", and the speed formula is ,in is the inertia weight, , is the learning factor, , is a random number, particle In the The current position in the dimension, is the final position of the individual particle, is the final global position. After each iteration, the weights are updated through the particle swarm optimization algorithm and the coordinates of the detection points are recalculated so that the detection points are dynamically adjusted according to the text semantics.

[0100] Statement advancement speed : The unit is "dimension / time step", which indicates the change amplitude of the semantic dimension within each time step, for example =0.1 means that each dimension vector increases by 0.1 times the initial value at each time step.

[0101] Rhetorical Switch Frequency : The unit is "times / time step", which indicates the number of times the rhetorical mode is switched in each time step, and is used to modulate the movement frequency of polygon vertices.

[0102] Let the initial polygon vertices be , No. Vertex coordinates at the time step The calculation is as follows:

[0103] ,in, is the time step, For the The semantic change direction vector of each detection point (determined by the weight gradient adjusted by the particle swarm optimization algorithm), is a random normal vector perpendicular to the semantic field plane (simulating the randomness of rhetorical transformation). The three detection points are used as initial vertices to form a triangle. The vertex coordinates are updated according to the above formula every time step to form a dynamic polygon. If the text contains multimodal information (such as metaphor switching), it triggers Modulation causes periodic fluctuations in the vertex movement trajectory, for example, when the frequency of rhetorical switching increases, The amplitude increases, the vertex movement increases, and the polygon shape changes more dramatically.

[0104] Assume that the minimum bounding box of the dynamic polygon in three-dimensional space is , define the grid resolution as (like =10), then the size of each grid cell is:

[0105] , , .

[0106] For each grid cell, its vertex coordinates satisfy:

[0107] );

[0108] );

[0109] );

[0110] At each time step, the coordinates of all grid cell vertices inside the polygon are recorded. If the vertex is outside the polygon, it is marked as an invalid coordinate (the ray method can be used to determine whether a point is inside the polygon: emit rays from the point in any direction and count the number of intersections with the polygon edge. Odd numbers are inside and even numbers are outside).

[0111] Set up the first The coordinates of the grid cell vertices at the time step are , No. The time step is , then the vertex displacement vector is:

[0112] , displacement size The average displacement per unit time is .

[0113] For each grid cell, the Area of ​​time step The area of ​​a three-dimensional polygon can be calculated by projecting the grid cell vertices onto the nearest semantic field plane (e.g., Determined plane), calculate the area of ​​the projected polygon (using the vector cross product method: if the projected vertex is , then the area is ).

[0114] Area change rate , reflects the degree of expansion or contraction of the grid unit in unit time. , its curvature It can be calculated by the position vectors of adjacent vertices: Three points, construct the tangent vector , after normalization, we get , then the curvature ,The larger the curvature value, the higher the degree of boundary curvature, and the higher the local complexity of the corresponding semantic field.

[0115] Assume that the original values ​​of vertex displacement, area change rate, and curvature are , whose maximum and minimum values ​​are , , , then the normalized value is:

[0116] , , introduce the particle swarm optimization algorithm to adjust the weight ), adjust the value , where the weights are dynamically adjusted by the particle swarm optimization algorithm according to the text semantics. For example, when the text sentiment intensity is high, PSO automatically increases (vertex displacement weights) to make the adjustment values ​​more sensitive to dynamic changes in the semantic field.

[0117] Using a particle swarm optimization algorithm to dynamically optimize detection point weights, the detection points in the three-dimensional semantic field can adaptively capture the interactive relationship between authorial cognition, reader reception, and textual function. For example, when analyzing poetry, the weights of sentiment and metaphorical mapping can be automatically adjusted based on the frequency of rhetorical transitions, avoiding the semantic drift caused by fixed weights in traditional methods. Dynamic polygons generated based on time series parameters can reflect the semantic morphological changes during sentence progression in real time. Combined with mesh cutting and feature parameter analysis, the vertex displacement (reflecting semantic flow), area change rate (reflecting semantic expansion / contraction), and boundary curvature (reflecting semantic complexity) of the semantic field can be quantified. For example, when analyzing chapter transitions in novels, the area change rate of the dynamic polygons can intuitively indicate the closeness of the chapter semantic cohesion, while the curvature mutation points can locate the transitions in narrative perspective. A high proportion of curvature in the adjusted value (reflected by the PSO weight) indicates that the local semantic complexity of the text exceeds the standard and requires simplification. A negative area change rate with a large absolute value indicates excessive semantic contraction and requires additional transition information. For example, in automated proofreading systems, this mechanism can help identify paragraphs with logical jumps in academic papers and recommend the addition of connectives or case examples. The combination of the construction of a three-dimensional contextual semantic field and the PSO algorithm provides a unified mathematical framework for semantic alignment of multimodal data such as text, images, and audio. For example, in multimedia document analysis, PSO can simultaneously optimize the weights of both the text semantic dimension and the image visual feature dimension, enabling dynamic polygon morphological changes to simultaneously reflect the association between text sentiment and image tonality, thus achieving cross-modal semantic consistency analysis.

[0118] In a preferred embodiment of the present invention, the above step 4, using the adjustment value to adjust the initial weight parameters of each dimension to obtain the adjusted weight parameters of each dimension, may include:

[0119] Step 440: Input the vertex displacement in the geometric feature into the weight adjustment rule of the author cognitive dimension to generate a first intermediate parameter, and perform increment and decrement corrections on the initial weight parameter of the author cognitive dimension based on the direction and magnitude of the displacement vector in the first intermediate parameter to obtain the author cognitive dimension correction weight;

[0120] Step 441: Input the area change rate into the weight adjustment rule of the reader acceptance dimension, generate a second intermediate parameter based on the monotonic increase and decrease trend, and dynamically adjust the proportional coefficient of the reader acceptance dimension weight parameter based on the change trend indicator in the second intermediate parameter to obtain the reader acceptance dimension correction weight;

[0121] Step 442: Input the boundary curvature value into the weight adjustment rule of the text function dimension, generate a third intermediate parameter based on the straight-curvature characteristic, and use the curvature quantization value in the third intermediate parameter to implement a smoothness strength constraint on the text function dimension weight parameter to obtain a text function dimension correction weight.

[0122] Step 443 , weighted fusion is performed on the author cognition dimension correction weight, the reader acceptance dimension correction weight, and the text function dimension correction weight to generate a set of adjusted dimension weight parameters that reflects the dynamic changes in polygon morphology.

[0123] In the embodiment of the present invention, the vertex displacement in the geometric feature is used as input, and the weight adjustment rule of the author's cognitive dimension is a function relationship based on mathematical mapping. Assume that the function is ,in Represents the vertex displacement. The vertex displacement is a vector data, containing direction and amplitude information. Let the vertex displacement be , through the function Perform calculations, functions It can be a multivariate function, for example ( 、 、 is a pre-set coefficient), the calculation result is the first intermediate parameter .

[0124] The first intermediate parameter The direction and magnitude of the displacement vector in determine the correction method for the initial weight parameter of the author’s cognitive dimension. Suppose the initial weight parameter of the author’s cognitive dimension is , if the direction of the displacement vector meets a certain preset positive rule (for example, the angle with the positive direction of the coordinate axis is within a certain range), and the amplitude is greater than the threshold , then perform incremental correction, the correction formula is ( is the incremental correction coefficient); if the displacement vector direction meets the preset negative rule and the amplitude is greater than the threshold , then the reduction correction is performed, and the correction formula is ( is the reduction correction coefficient), and finally the correction weight of the author's cognitive dimension is obtained .

[0125] Input the area change rate into the weight adjustment rule of the reader acceptance dimension, and let the function corresponding to this rule be ,in Represents the area change rate. The area change rate is a scalar value, which is expressed by the function Perform calculations, assuming ( 、 、 is a pre-set coefficient), the calculation result is the second intermediate parameter According to the monotonic increase and decrease trend of the area change rate (which can be determined by derivation, if The derivative of is greater than 0, which is monotonically increasing; less than 0, which is monotonically decreasing). Generate corresponding change trend identification.

[0126] Assume that the initial proportional coefficient of the reader acceptance dimension weight parameter is , based on the second intermediate parameter Dynamic adjustment is performed on the change trend indicator in the . If the change trend is increasing, the adjustment formula is ( is the increasing adjustment coefficient); if the change trend is decreasing, the adjustment formula is ( is the decreasing adjustment coefficient), thereby obtaining the modified weight of the reader acceptance dimension ( Accepts dimension initial weight parameters for readers).

[0127] Input the boundary curvature value into the weight adjustment rule of the text function dimension, and let the function corresponding to this rule be ,in Represents the boundary curvature value. The boundary curvature value is a scalar, which is obtained by the function Perform calculations, assuming ( 、 is a pre-set coefficient), the calculation result is the third intermediate parameter , which contains the curvature quantization value. Let the text feature dimension weight parameter be , using the third intermediate parameter The curvature quantization value in imposes a smoothness strength constraint on it. If the curvature quantization value is greater than the threshold , indicating that the boundary is tortuous, and the weight needs to be reduced to enhance smoothness. The adjustment formula is ( is the constraint coefficient); if the curvature quantization value is less than or equal to the threshold , the weight remains unchanged or is slightly adjusted, and the final result is the text function dimension correction weight . Modify the weight of the author's cognitive dimension , Reader acceptance dimension correction weight and text function dimension correction weight Perform weighted fusion. Assume that the fusion coefficients are ,and + + =1, the weight parameter set of each dimension after adjustment The calculation formula is , this set reflects the dynamic changes of polygon morphology.

[0128] By combining geometric features such as vertex displacement, area change rate, and boundary curvature with weight adjustment rules for different dimensions, we can accurately capture the impact of polygon morphological changes on author cognition, reader acceptance, and text function dimensions, making the adjustment of weight parameters more in line with actual needs and avoiding the blindness and subjectivity of weight setting. In each step, the weights are dynamically corrected and adjusted based on intermediate parameters and different conditions, allowing the weight parameters to be adaptively adjusted according to the real-time changes in polygon morphology. In practical application scenarios, such as graphic design and text typesetting, this method can timely optimize weights in the face of constantly changing input data, ensuring the effectiveness and rationality of the processing results. Finally, the corrected weights of the three dimensions are weighted and fused, fully considering the interrelationships and importance of different dimensions, and optimizing the weight parameters overall. The resulting weight parameter set can more comprehensively and accurately reflect the dynamic changes of polygon morphology.

[0129] In a preferred embodiment of the present invention, the above step 5 generates a set of candidate sentences including two rhetorical devices based on the composite rhetorical structure and the adjusted weight parameters of each dimension, and labels each candidate sentence with the abstract level of the verbal thinking path, which may include:

[0130] Step 550: The adjusted author cognition dimension, reader acceptance dimension, and text function dimension weight parameters are integrated with the core metaphor, auxiliary symbol, and grammatical inflection markers in the complex rhetorical structure, and a rhetorical element combination set that meets the multi-dimensional requirements is generated by sorting.

[0131] Step 551: Based on the rhetorical element combination set, a candidate sentence set of two rhetorical devices, a combination of metaphor and symbol, and a combination of metaphor and grammatical inflection, is generated;

[0132] Step 552: For each sentence in the candidate sentence set, analyze the nested logic and semantic advancement path of the rhetorical device, and mark each sentence with an abstract level of speech thinking path from single rhetoric mapping to multi-rhetoric collaboration according to the hierarchical division standard of Vygotsky's speech thinking theory.

[0133] In this embodiment of the present invention, the adjusted weight parameters for the author's cognitive dimension, the reader's acceptance dimension, and the text function dimension are each considered important indicators for measuring the adaptability of rhetorical elements. For a core metaphor in a complex rhetorical structure, the author's cognitive dimension assesses its alignment with the author's intended meaning and assigns an initial adaptability score of 0-10. The reader's acceptance dimension assesses the metaphor's ease of understanding and assigns a further score of 0-10. The text function dimension considers the metaphor's support for the main theme and also assigns a score of 0-10. The scores for each of these three dimensions are multiplied by the corresponding weight parameters and summed to obtain a comprehensive score for each core metaphor. For example, core metaphor A scores 8 in the author's cognitive dimension, 7 in the reader's acceptance dimension, and 6 in the text function dimension. If the weight parameters for these three dimensions are 0.3, 0.4, and 0.3, respectively, its comprehensive score = 8 × 0.3 + 7 × 0.4 + 6 × 0.3 = 7.

[0134] Auxiliary symbols and grammatical inflection markers are processed in the same manner. For auxiliary symbols, their relevance to the core metaphor is assessed from the perspective of author cognition; their aiding effect on reader comprehension is assessed from the perspective of reader acceptance; and their contribution to the text's atmosphere is assessed from the perspective of text function. Scores are assigned to each dimension and multiplied by the corresponding weight parameter, resulting in a composite score. For grammatical inflection markers, their conformity to the author's expression habits is assessed from the perspective of author cognition; their impact on reading fluency is assessed from the perspective of reader acceptance; and their contribution to sentence structure optimization is assessed from the perspective of text function. Scores are assigned and a composite score is calculated. All core metaphors, auxiliary symbols, and grammatical inflection markers are sorted from high to low based on their composite scores. Combination rules are set, such that each combination must contain a core metaphor ranked in the top five, two auxiliary symbols ranked in the top ten, and one grammatical inflection marker ranked in the top eight. Following this rule, elements are selected from the sorted list to generate multiple different rhetorical element combinations, which together constitute a rhetorical element combination set.

[0135] Select a combination from the set of rhetorical element combinations, with the core metaphor as the main body. Assume the core metaphor is "Life is a journey," and the selected auxiliary symbols are "Signposts guide the direction" and "Luggage is filled with memories." Based on a pre-set sentence template, such as "[Core metaphor], [Auxiliary symbol 1], [Auxiliary symbol 2]," combine these into a candidate sentence: "On this journey of life, signposts guide the direction, and the luggage is filled with memories." By varying the sentence template and auxiliary symbol combinations, a series of candidate sentences combining metaphors and symbols are generated. Similarly, select a combination from the set of rhetorical element combinations, using the core metaphor as the basis. For example, if the core metaphor is "Dreams are lighthouses," the selected grammatical variation identifier is "Inverted sentence structure." First, determine the normal sentence structure: "The lighthouse is a dream." Then, according to the rules of the inverted sentence structure, adjust it to "It is the lighthouse, the dream." Add appropriate conjunctions and modifiers to form a candidate sentence structure such as "On the confusing sea, it is the lighthouse, the dream that illuminates the way forward." By using different grammatical inflection markers and adjustment methods, candidate sentences combining metaphor and grammatical inflection are generated, which together with the former constitute a complete set of candidate sentences.

[0136] For each sentence in the candidate set, we first analyze the rhetorical devices it contains. For example, "The flower, brilliant as a flame, dances in the wind, announcing its beauty to the world" contains the metaphor "the flower is like a flame" and the personification "the flower proclaims its beauty." Analyzing the nested relationship between these two rhetorical devices reveals that the metaphor first imbues the flower with the qualities of a flame, while the personification further builds on this by imbuing the flower with the human-like behavior of announcing it, forming a semantic progression from attribute description to behavioral expression. This approach allows us to identify the order and interrelationships of the rhetorical devices within each sentence, constructing a semantic progression path. According to the hierarchical division standard of Vygotsky's theory of verbal thinking, if a sentence contains only one rhetorical device, such as "the sun is like a disk", it is marked as a single rhetorical mapping level; if a sentence contains two rhetorical devices and they are independent of each other and simply superimposed, such as "the moon gently looks at the earth, and the stars wink mischievously", it is marked as a simple rhetorical combination level; if multiple rhetorical devices in a sentence are intertwined with each other and jointly construct complex semantics, such as the above example of "that flower as gorgeous as a flame...", it is marked as a multi-rhetorical collaborative level, thus completing the labeling of the abstract level of the verbal thinking path of each candidate sentence.

[0137] By integrating multi-dimensional weighting parameters with complex rhetorical structures, candidate sentence structures are generated, fully considering the author's creative intent, the reader's comprehension difficulty, and the functional requirements of the text. This allows the generated candidate sentence structures to precisely match the intended expression in various application scenarios, providing more appropriate sentence structures for both emotional expression in literary creation and information transmission in business copywriting, significantly enhancing the candidate sentence structures' relevance and practicality. The generated candidate sentence structures combine metaphor and symbol, as well as metaphor and grammatical inflection, transcending the limitations of traditional single rhetoric. By combining different rhetorical methods, novel and unique forms of expression are created, injecting new vitality into language expression. In fields such as advertising creativity and poetry, this approach can help creators create more engaging and impactful language works, promoting rhetorical innovation. Based on Vygotsky's theory of verbal thought, the abstract levels of verbal thought paths are labeled, establishing a clear analytical system for candidate sentence structures. Educators can use this hierarchical division to carry out step-by-step language teaching, helping students master the use of rhetorical devices from simple to complex; language researchers can use this framework to deeply explore the relationship between language expression and thinking development, providing new perspectives and methods for linguistic research.

[0138] In a preferred embodiment of the present invention, the above step 6, based on the literary style preference, performs rhythm detection, image coordination evaluation, and language adaptability verification on the candidate sentence set to generate the corresponding thinking path presentation, which may include:

[0139] Step 660 , based on the rhyme characteristic parameters in the literary style preference, including level and oblique tone distribution, syllable density, and stress period, the syntactic structure of the candidate sentence is divided into syllables, the standard deviation of the syllable drop value and the pause length is calculated, and a rhythm adaptability score of the candidate sentence is generated;

[0140] Step 661: Based on the rhythm adaptation score, the cultural prototype symbol library in the Vygotsky symbol association network is called to calculate the semantic association between the source domain and target domain images of the metaphor mapping in the sentence pattern, and the image conflict index is constructed in combination with the emotional tendency intensity value to determine a symbol combination scheme that meets the style preference and has a harmonious rhythm;

[0141] Step 662 , based on the symbolic symbol combination scheme and the weight parameters of the text function dimension, register feature matching is performed on the candidate sentence patterns to generate a comprehensive evaluation result reflecting the register adaptability;

[0142] Step 663 , integrate the comprehensive evaluation results of rhythm adaptability score, symbol combination scheme and style adaptability to generate a thinking path presentation including rhetorical coordination, style deviation and cognitive load value.

[0143] In an embodiment of the present invention, the syntactic structure of a candidate sentence is analyzed based on the phonetic and rhyme characteristic parameters specified in the literary style preference, such as the level and oblique tone distribution rules (for example, the level and oblique tone alternation requirements of classical poetry), the syllable density reference value (which specifies the appropriate range of syllables per sentence), and the stress period pattern (which determines the frequency and location of stressed syllables). The number of syllables in each foot is counted, and the difference in the number of syllables between adjacent feet is calculated to obtain a series of syllable drop values. For example, if two adjacent feet in a sentence have 3 and 4 syllables, respectively, the syllable drop value is 1. Simultaneously, the actual pause duration between each foot is measured according to preset pause duration standards (for example, a comma pause of 0.5 seconds, a period pause of 1 second, etc.), the average of these pause durations is calculated, and the standard deviation of the pause durations is then calculated using a standard deviation formula. The syllable drop value and the standard deviation of the pause duration are used as evaluation indicators, and a scoring formula is established in conjunction with the requirements for rhythmic fluency specified in the literary style preference. For example, if the syllable drop value and the pause length standard deviation are smaller, the rhythm is smoother and more fluent, and a higher score is given; otherwise, a lower score is given. Assuming the score range is 0-10, the above two indicators are converted into a rhythm adaptation score through linear or nonlinear functions.

[0144] Based on the rhythmic adaptability scores obtained in step 660, candidate sentences with high scores and good rhythmic adaptability are selected. The cultural prototype symbol library in the Vygotsky symbolic association network is used to analyze the source and target domain images of the metaphorical mapping in the sentence. For example, for the metaphor "time is like flowing water," "time" is the target domain, and "flowing water" is the source domain. Semantic features and cultural association information related to "time" and "flowing water" are extracted from the symbol library. The semantic association value between the two is calculated by methods such as calculating the cosine similarity of the semantic vectors. In addition to the semantic association, the emotional tendency intensity value of the imagery in the sentence is considered. Using sentiment analysis tools or a preset sentiment vocabulary, the emotion expressed by each image is determined to be positive, negative, or neutral, and a corresponding emotional intensity value is assigned (e.g., 1 for positive, -1 for negative, and 0 for neutral). Calculate the absolute difference between the source and target domain image sentiment intensity values. Combined with semantic relevance, construct an image conflict index formula. For example, the image conflict index = semantic relevance - |source domain sentiment intensity - target domain sentiment intensity|. The lower the index, the better the image harmony. Based on the image conflict index, screen the symbol combinations in the candidate sentences. Select symbol combinations with a low image conflict index, consistent with literary style preferences, and harmonious rhythm to form the final symbol combination plan.

[0145] Based on the symbolic symbol combination scheme determined in step 661 and combined with the weight parameters of the text function dimension, candidate sentences are matched with their register characteristics. First, the different register types (e.g., formal written language, spoken language, literary language, etc.) and their characteristic indicators (e.g., lexical complexity, rigor of sentence structure, etc.) involved in the text function dimension are identified. Then, the characteristics of the candidate sentences in terms of vocabulary usage, grammatical structure, and expression are analyzed and compared with the characteristic indicators of each register. The matching score of each candidate sentence with different registers is calculated. Based on the weight parameters of the text function dimension, the matching scores of different registers are weighted and summed to obtain a comprehensive evaluation result reflecting the adaptability of the language style. For example, if the text function focuses primarily on literary expression, the literary language register will be given a higher weight, and its matching score will account for a larger proportion in the comprehensive evaluation.

[0146] The rhythmic adaptability score obtained in step 660, the symbolic symbol combination scheme determined in step 661, and the comprehensive assessment results of register adaptability generated in step 662 are integrated. Based on the degree of coordination between different rhetorical devices in the symbolic symbol combination scheme, the rhythmic adaptability score, and the comprehensive assessment results of register adaptability, a comprehensive evaluation function is set to calculate rhetorical coordination. For example, if the multiple rhetorical devices in a sentence echo each other, the rhythm is smooth, and the register is appropriate, the rhetorical coordination is high. The various characteristics of the candidate sentence (such as rhythm, imagery, and register) are compared with the standard characteristics of literary style preferences, and the degree of difference is calculated to obtain the style deviation. The smaller the difference, the lower the style deviation. Factors such as sentence complexity (such as the number and degree of nesting of rhetorical devices, the unfamiliarity of vocabulary), rhythmic fluency, and imagery clarity are taken into account. The more complex the sentence, the greater the difficulty of understanding, and the higher the cognitive load value. The calculated rhetorical coordination, style deviation, and cognitive load values ​​are summarized to generate an intuitive thinking path presentation, clearly showing the performance of each candidate sentence in different dimensions.

[0147] Rhythm testing ensures that candidate sentences meet the rhyme requirements of a specific literary style, making the language easy to read and enhancing the work's musical beauty. Imagery coordination assessment avoids image conflicts, making metaphorical expressions more natural and reasonable, and enriching the work's connotations. Verification of register adaptability ensures that sentences use appropriate language styles in different contexts, enhancing the standardization of language expression and thus comprehensively improving the artistry and quality of the work. Evaluating and optimizing candidate sentences based on literary style preferences can precisely match the author's desired stylistic characteristics. Whether it's the elegant rhythm of classical poetry or the free and dynamic style of modern prose, the generated sentences can better serve the overall style, avoid stylistic confusion, and make the work more unified and unique. The generated thought paths present a quantitative and intuitive representation of the candidate sentences' performance across various dimensions. Authors can quickly judge the strengths and weaknesses of sentences based on rhetorical coordination, stylistic deviation, and cognitive load, and selectively select or modify them to improve creative efficiency. This also provides an objective analytical basis for literary research and language teaching, contributing to a deeper understanding of the laws of language use in literary creation.

[0148] like Figure 2 As shown, an embodiment of the present invention further provides a system for constructing word meaning sentences based on verbal thinking, including:

[0149] The semantic analysis module is used to analyze the semantic core and emotional tendency of the initial sentence input by the user, and match the appropriate complex rhetorical structure based on Vygotsky's theory of verbal thinking;

[0150] The semantic construction module is used to construct a three-dimensional contextual semantic field including author, reader and text functions based on the semantic core and emotional tendency, and extract the initial weight parameters of each dimension;

[0151] The morphological analysis module is used to determine the detection points in the three-dimensional context semantic field and generate dynamic polygons. It analyzes the morphological change characteristics through mesh cutting and generates adjustment values.

[0152] A parameter adjustment module is used to dynamically adjust the initial weight parameters of each dimension using the adjustment values ​​generated by the dynamic polygon;

[0153] The sentence generation module is used to generate a set of candidate sentences including multiple rhetorical techniques based on the compound rhetorical structure and the adjusted weight parameters, and annotate the corresponding abstract thinking path level;

[0154] The adaptation evaluation module is used to conduct a comprehensive evaluation of the rhythm, image coordination and language adaptability of candidate sentences based on the user's literary style preferences, and obtain the final thinking path presentation.

[0155] It should be noted that this system is a system corresponding to the above method, and all implementation methods in the above method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0156] An embodiment of the present invention further provides a computing device comprising: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the above-described method. All implementations in the above-described method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0157] The embodiment of the present invention further provides a computer-readable storage medium storing instructions, which, when executed on a computer, causes the computer to execute the above-described method. All implementations in the above-described method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0158] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for constructing word meaning sentences based on verbal thinking, characterized in that: The method comprises: Step 1: Obtain the user's initial sentence, preset rhetorical device types, and literary style preferences, and analyze the semantic core and emotional tendency of the initial sentence. Based on the semantic core, call on the rhetorical knowledge base of Vygotsky's theory of verbal thinking, including the metaphor mapping matrix, symbolic association network, and grammatical inflection rule set, to match the appropriate compound rhetorical structure; Step 2: Based on the semantic core and emotional tendency, a three-dimensional contextual semantic field is constructed, and the initial weight parameters of each dimension of the author's cognitive dimension, the reader's acceptance dimension, and the text function dimension are obtained through multimodal feature extraction; Step 3: Based on the initial weight parameters of each dimension, three detection points are determined in the 3D context semantic field. Dynamically changing polygons are generated according to the spatial coordinates. The dynamically changing polygons are meshed and analyzed. The vertex displacement, area change, and boundary curvature characteristics of each mesh unit are analyzed to generate adjustment values ​​reflecting the polygon morphological changes. Step 4: Use the adjustment value to adjust the initial weight parameters of each dimension to obtain the adjusted weight parameters of each dimension; Step 5: Based on the compound rhetorical structure and the adjusted weight parameters of each dimension, a set of candidate sentences including two rhetorical devices is generated, and the abstract level of the verbal thinking path is marked for each candidate sentence; Step 6: Based on literary style preferences, the candidate sentence set is tested for rhythm and rhyme, image coordination is evaluated, and language adaptability is verified to generate a corresponding thinking path presentation.

2. The method for constructing word meaning sentences based on verbal thinking according to claim 1 is characterized in that: The system obtains the user's initial sentence input, the preset rhetorical device type, and the literary style preference, and analyzes the semantic core and emotional tendency of the initial sentence. Based on the semantic core, it calls the rhetorical knowledge base of Vygotsky's theory of verbal thinking, including the metaphor mapping matrix, symbolic association network, and grammatical inflection rule set, to match the appropriate compound rhetorical structure, including: Perform dependency syntax analysis on the initial sentence, extract the predicate verb and the governing argument as the semantic core, and calculate the sentiment tendency intensity value through sentiment dictionary matching; The semantic core is input into the metaphor mapping matrix of the Vygotsky rhetorical knowledge base, and the source domain-target domain mapping pairs that meet the threshold conditions are determined according to the emotional tendency intensity value to generate the initial metaphor set; Each source domain symbol in the initial metaphor set is input into the symbolic association network, and the three-layer association nodes are recursively retrieved. The candidate symbol set is formed by combining cultural prototype symbols, historical allusion symbols and natural image symbols. The grammatical inflection rule set is called to reorganize the candidate symbolic symbol set, including applying metonymy compression rules to the target domain and metaphor expansion rules to the source domain, to obtain a composite rhetorical structure including two rhetorical techniques, namely the core metaphor, auxiliary symbol and grammatical inflection marker.

3. The method for constructing word meaning sentences based on verbal thinking according to claim 2 is characterized in that: Based on the semantic core and emotional tendency, a three-dimensional contextual semantic field is constructed, and the initial weight parameters of each dimension of the author's cognitive dimension, the reader's acceptance dimension, and the text function dimension are obtained through multimodal feature extraction, including: With the semantic core as the origin, the emotional tendency as the vertical axis, and the text theme relevance as the horizontal axis, a three-dimensional situational semantic field is constructed, including author intention, contextual information, and semantic association. Based on the three-dimensional contextual semantic field, the author's writing history data, the pre-built reader portrait library and the functional labels of the complex rhetorical structure are extracted to generate feature vectors representing the author's cognitive dimension, the reader's acceptance dimension and the text's functional dimension; The author cognition dimension feature vector, reader acceptance dimension feature vector and text function dimension feature vector are normalized respectively to obtain the corresponding author cognition dimension initial weight parameters, reader acceptance dimension initial weight parameters and text function dimension initial weight parameters.

4. The method for constructing word meaning sentences based on verbal thinking according to claim 3 is characterized in that: Based on the initial weight parameters of each dimension, three detection points are determined in the 3D context semantic field. Dynamically changing polygons are generated according to the spatial coordinates. The dynamically changing polygons are meshed and analyzed. The vertex displacement, area change, and boundary curvature characteristics of each mesh unit are analyzed to generate adjustment values ​​reflecting the polygon morphological changes, including: In the three-dimensional situational semantic field, based on the initial weight parameters of each dimension, the intersection of the author's cognitive dimension and emotional tendency is determined as the first detection point, the intersection of the reader's acceptance dimension and the semantic core is determined as the second detection point, and the intersection of the text function dimension and metaphor mapping is determined as the third detection point. Based on the spatial coordinates of the three detection points, dynamically changing polygons are generated according to preset time series parameters, including sentence advancement speed and rhetorical conversion frequency, and the polygon vertices move in three-dimensional space according to the time parameters; For dynamically changing polygons, a uniform meshing method is used to segment them into several small mesh units, and the coordinates of each mesh unit vertex at each time point are recorded; Based on the recorded grid cell vertex coordinates, the difference between vertex coordinates at adjacent time points is calculated to obtain vertex displacement, the rate of change of grid cell area per unit time is counted, and the curvature value of the polygon boundary is calculated; The vertex displacement, area change rate and polygon boundary curvature value are normalized to generate an adjustment value reflecting the polygon morphology change.

5. The method for constructing word meaning sentences based on verbal thinking according to claim 4 is characterized in that: The initial weight parameters of each dimension are adjusted using the adjustment value to obtain the adjusted weight parameters of each dimension, including: Input the vertex displacement in the geometric feature into the weight adjustment rule of the author cognitive dimension to generate a first intermediate parameter, and perform incremental and decrement corrections on the initial weight parameter of the author cognitive dimension according to the direction and magnitude of the displacement vector in the first intermediate parameter to obtain the author cognitive dimension correction weight; Input the area change rate into the weight adjustment rule of the reader acceptance dimension, generate a second intermediate parameter according to the monotonic increase and decrease trend, and dynamically adjust the proportional coefficient of the reader acceptance dimension weight parameter based on the change trend indicator in the second intermediate parameter to obtain the reader acceptance dimension correction weight; Input the boundary curvature value into the weight adjustment rule of the text function dimension, generate a third intermediate parameter according to the straight-curvature characteristic, and use the curvature quantization value in the third intermediate parameter to implement a smoothness strength constraint on the text function dimension weight parameter to obtain the text function dimension correction weight; The weighted fusion of the author cognition dimension correction weight, the reader acceptance dimension correction weight and the text function dimension correction weight is performed to generate a set of adjusted dimension weight parameters that reflects the dynamic changes of polygon morphology.

6. The method for constructing word meaning sentences based on verbal thinking according to claim 5 is characterized in that: Based on the compound rhetorical structure and the adjusted weight parameters of each dimension, a set of candidate sentences including two rhetorical devices is generated, and the abstract level of the verbal thinking path is annotated for each candidate sentence, including: The adjusted weight parameters of the author's cognitive dimension, reader's acceptance dimension, and text function dimension are integrated with the core metaphors, auxiliary symbols, and grammatical inflection markers in the complex rhetorical structure, and a combination set of rhetorical elements that meets the multi-dimensional requirements is generated through sorting. Based on the combination set of rhetorical elements, two candidate sentence sets of rhetorical devices, namely, the combination of metaphor and symbol, and the combination of metaphor and grammatical deformation, are generated; For each sentence in the candidate sentence set, the nested logic and semantic advancement path of the rhetorical devices are analyzed, and according to the hierarchical division standard of Vygotsky's verbal thinking theory, the abstract level of the verbal thinking path from single rhetoric mapping to multi-rhetoric collaboration is marked for each sentence.

7. The method for constructing word meaning sentences based on verbal thinking according to claim 6 is characterized in that: Based on literary style preferences, the candidate sentence set is tested for rhythm, image coordination, and language adaptability, generating corresponding thought path presentations, including: Based on the rhyme characteristic parameters in literary style preferences, including level and oblique tone distribution, syllable density, and stress period, the syntactic structure of the candidate sentence is divided into syllables, and the standard deviation of the syllable drop value and pause length is calculated to generate the rhythm adaptability score of the candidate sentence. Based on the rhythmic fit score, the cultural archetype symbol library in Vygotsky's symbolic association network is used to calculate the semantic relevance of the source and target domain images of the metaphorical mapping in the sentence. The image conflict index is then constructed based on the emotional tendency intensity value to determine the symbolic symbol combination scheme that conforms to the style preference and has a harmonious rhythm. Based on the symbolic symbol combination scheme and the weight parameters of the text function dimension, the candidate sentences are matched with the register features to generate a comprehensive evaluation result reflecting the adaptability of the language style; The comprehensive evaluation results of rhythm adaptation score, symbol combination scheme and style adaptation are integrated to generate a thinking path presentation including rhetorical coordination, style deviation and cognitive load value.

8. A system for constructing word meaning sentences based on verbal thinking, the system implementing the method according to any one of claims 1 to 7, characterized in that: include: The semantic analysis module is used to analyze the semantic core and emotional tendency of the initial sentence input by the user, and match the appropriate complex rhetorical structure based on Vygotsky's theory of verbal thinking; The semantic construction module is used to construct a three-dimensional contextual semantic field including author, reader and text functions based on the semantic core and emotional tendency, and extract the initial weight parameters of each dimension; The morphological analysis module is used to determine the detection points in the three-dimensional context semantic field and generate dynamic polygons. It analyzes the morphological change characteristics through mesh cutting and generates adjustment values. A parameter adjustment module is used to dynamically adjust the initial weight parameters of each dimension using the adjustment values ​​generated by the dynamic polygon; The sentence generation module is used to generate a set of candidate sentences including multiple rhetorical techniques based on the compound rhetorical structure and the adjusted weight parameters, and annotate the corresponding abstract thinking path level; The adaptation evaluation module is used to conduct a comprehensive evaluation of the rhythm, image coordination and language adaptability of candidate sentences based on the user's literary style preferences, and obtain the final thinking path presentation.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images