Self-adaptive poem creation method fusing traditional culture elements and modern technology

Abstract

The invention relates to the technical field of poem creation, and discloses a self-adaptive poem creation method fusing traditional culture elements and a modern technology. The method comprises the following steps: S1, collecting multi-modal culture data; s2, processing the multi-modal culture data; s3, establishing a dynamic culture tensor according to the triple information; s4, generating a quantum state according to user input; s5, according to the dynamic culture tensor, constructing a sparse culture Hamiltonian; s6, the quantum state is evolved through a variable component sub-circuit; s7, carrying out cross-modal semantic correction; and S8, outputting and generating a final poem and image description. By adopting the technical scheme of combining quantum calculation and multi-modal culture tensor, the technical effect of more accurately capturing culture information in the creation process is achieved, the diversity of quantum states and the depth of culture data are fully utilized, and the defect that a traditional method lacks flexibility in expression of culture details and styles is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of poetry creation, and specifically to an adaptive poetry creation method that integrates traditional cultural elements with modern technology. Background Art

[0002] Currently, traditional poetry creation methods often rely on fixed rules or templates, which limits the flexibility of artistic creation. Existing text generation models often lack a deep understanding of culture and style, resulting in shallow cultural connotations in generated works. Poetry generated using traditional algorithms often fails to express rich emotions and diverse cultural elements. This model fails to meet the modern user's demand for personalized and in-depth expression. Therefore, how to effectively integrate cultural information into creative creation has become an imperative issue.

[0003] Furthermore, existing technologies often rely on a single evaluation criterion for content generation. This approach is insufficient for multi-level evaluation of user experience. The system is unable to adapt in real time to user feedback, creating a kind of information silo. This directly impacts user satisfaction and results in works that fail to reflect the aesthetic and emotional needs of modern readers. Consequently, artistic creation loses its interactivity with the audience and fails to establish an effective feedback mechanism.

[0004] Furthermore, while some current style transfer algorithms have made technological progress, most remain at the surface level of processing images or language. They often fail to delve deeply into cultural connotations, making it difficult to create artistic expressions with cultural depth. Traditional algorithms in these methods struggle to adapt to complex cultural contexts, resulting in a lack of innovation and artistry in the generated content. The resulting artworks using these techniques often lack impressive cultural depth and artistic tension.

[0005] Finally, existing generative models are often static and lack the ability to adapt to user feedback. They cannot dynamically adjust generation strategies during the creation process and can only produce fixed outputs based on preset parameters. Consequently, users' personalized needs are difficult to fully meet. These limitations hinder the development of artistic creation, resulting in insufficiently rich generated content and a failure to resonate deeply with users. Summary of the Invention

[0006] In response to the shortcomings of existing technologies, the present invention provides an adaptive poetry creation method that integrates traditional cultural elements with modern technologies, which solves the problems of existing poetry creation methods in capturing cultural information, meeting users' personalized needs and achieving artistic innovation.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: an adaptive poetry creation method that integrates traditional cultural elements and modern technology, comprising the following steps;

[0008] S1. Collect multimodal cultural data, including classical poetry texts, related images, and user style preference data;

[0009] S2, processing multimodal cultural data to extract imagery, emotion, and style to form structured triplets;

[0010] S3. Build a dynamic cultural tensor based on the triple information to reflect the relationship between imagery, emotion, style, and modality in the multimodal data. The tensor is subsequently used to initialize the quantum state.

[0011] S4. Generate quantum states based on user input to represent the desired style and imagery. The quantum states form the basis for subsequent style transfer.

[0012] S5. Construct a sparse cultural Hamiltonian based on the dynamic cultural tensor for style transfer.

[0013] S6. Evolving the quantum state through a variational quantum circuit to achieve style transfer and optimize style fidelity;

[0014] S7. Cross-modal semantic correction: During the generation phase, cross-modal semantic correction is performed on the style transfer results to ensure cultural and semantic consistency of the generated poems.

[0015] S8. Output the generated final poem and image description to form a work with multimodal fusion features, and feed back the generated results to the user.

[0016] Preferably, the multimodal cultural data in step S1 further includes historical text records uploaded by the user interaction platform, so that the system can adjust the creative style in a targeted manner.

[0017] Preferably, the extraction of imagery, emotion and style in step S2 is achieved through a deep learning model, and the model is a bidirectional long short-term memory network or a convolutional neural network.

[0018] Preferably, the construction process of the dynamic cultural tensor in step S3 adopts a tensor decomposition algorithm, and the optimization criterion is to select the rank of the tensor by minimizing the following loss function:

[0019]

[0020] Among them, R represents the optimization result, C represents the dynamic culture tensor, and λ r represents the weight of the solution, a r ,b r ,c r ,d r represents the factor components of the tensor, represents element-wise product, ||·||F represents the Frobenius norm, argmin R′ It represents the process of finding R′ that minimizes the expression, i.e., the optimization process.

[0021] Preferably, the process of generating the quantum state in step S4 includes the following steps:

[0022] After receiving user input, analyzing the user input text to identify target style and imagery;

[0023] Based on the recognition results, feature information related to the target style and imagery is extracted from the dynamic cultural tensor;

[0024] The quantum state is constructed using the selected feature information so that the quantum state accurately represents the artistic style and emotional expression required by the user.

[0025] Preferably, the sparse cultural Hamiltonian in step S5 is constructed by the following formula:

[0026]

[0027] in, represents the Hamiltonian, represents the summation of r from 1 to R, R is the upper limit of the summation, λ r represents the weight coefficient, Denotes the three different components of the Paul i matrix, σ z , σ x , σ y is the standard notation for the Paul i matrix, a r , b r , c r represents the label or index associated with each Paul i matrix, Represents the Kronecker product, that is, the tensor product.

[0028] Preferably, in step S6, the quantum state is evolved and the style fidelity is optimized by a variational quantum circuit, and the loss function includes at least the following parts:

[0029]

[0030] in, represents the loss function, 1 represents the constant term, represents the inner product norm between quantum states, ψ f represents the quantum state after style transfer, represents the quantum state of the target style, Represents the quantum state |ψ f > and The inner product between , μ represents the regularization parameter, Represents the Hamiltonian The trace of the square of represents the Hamiltonian of the quantum system, and Tr represents the trace of the matrix.

[0031] Preferably, the cross-modal semantic correction step in step S7 adopts a meta-learning algorithm to optimize the cultural adaptability of the generated results to achieve high consistency between the generated poems and the image descriptions.

[0032] Preferably, the user style preference data in step S1 is updated through a history generation result feedback system.

[0033] Preferably, the output generation in step S8 is combined with a cultural tuning mechanism, which can adapt to the needs of users with different cultural backgrounds and automatically adjust the parameter settings of the generation model.

[0034] The present invention provides an adaptive poetry creation method that integrates traditional cultural elements with modern technology.

[0035] It has the following beneficial effects:

[0036] 1. This invention combines quantum computing with multimodal cultural tensors to more accurately capture cultural information during the creative process. Compared to traditional rule-based approaches, this approach leverages the diversity of quantum states and the depth of cultural data, addressing the traditional approach's lack of flexibility in expressing cultural details and styles.

[0037] 2. This invention effectively enhances the personalization and user satisfaction of generated works by combining user feedback with a generation algorithm. This technical effect ensures that works better reflect users' emotions and aesthetic needs. Compared with the existing single evaluation criteria approach, it addresses the difficulty of meeting the diverse needs of users.

[0038] 3. By incorporating a dynamic cultural Hamiltonian for quantum state evolution, this invention achieves the technical effect of achieving both complexity and innovation in artistic creation. Compared to the simpler style transfer methods currently available, this approach enhances the cultural depth and artistic quality of a work, addressing the inability of traditional methods to deeply explore cultural connotations.

[0039] 4. This invention, through the use of variational quantum circuits for quantum state evolution, achieves the technical effect of enhancing creative flexibility and expressive richness in the generation of artworks. Compared to the limitations of static generative models in existing technologies, this solution can dynamically adjust the creative direction based on real-time feedback, addressing the shortcomings of traditional methods in terms of stylistic adaptability and interactivity, making the creative process more personalized to the user's needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0042] Please see the attached Figure 1 , an embodiment of the present invention provides an adaptive poetry creation method that integrates traditional cultural elements and modern technology, characterized by comprising the following steps;

[0043] S1. Collect multimodal cultural data, including classical poetry texts, related images, and user style preference data;

[0044] Specifically, in this embodiment, the collection of multimodal cultural data in step S1 includes the acquisition of classical poetry texts, related images, and user style preference data. The logical framework and specific implementation of this process are described below.

[0045] First, multiple data sources were identified and integrated, including historical literature databases, online poetry platforms, user-uploaded content, and cultural and art websites. These data sources were then integrated into the data collection module.

[0046] The classical poetry texts are then extracted using natural language processing (NLP) technology. This technology employs word segmentation, part-of-speech tagging, and named entity recognition algorithms. The primary basis for feature extraction is a context-based word embedding model. Using pre-trained models (such as BERT), each piece of poetry is converted into several word vectors, forming a multi-dimensional feature matrix based on the text.

[0047] Based on this foundation, relevant images are collected immediately. Using image recognition and processing algorithms, images relevant to classical poetry are captured from specific data sources (such as artwork and images of traditional cultural elements). The image data processing module uses a convolutional neural network (CNN) to extract features from images, converting each image into a feature vector, forming multiple image feature datasets.

[0048] Next, the acquisition of user style preference data relies on a feedback system based on historical generation results. Users submit their evaluations of system-generated poems through an interactive platform. Based on these feedback, the system creates a personalized style library. This library uses a clustering analysis algorithm to categorize user preferences and generate targeted style tags.

[0049] This step also constructs a data integration module to integrate the data collected from the above sub-modules to form a fused dataset containing text information, image features and user preferences.

[0050] On this basis, the constructed triple information provides the necessary feature input for the subsequent construction of the dynamic cultural tensor. Specifically, the interaction between the three elements of imagery, emotion, and style will be further explored in subsequent steps.

[0051] The above process is mathematically modeled using the following formula to describe the relationship between each variable:

[0052] D = f(T, I, U);

[0053] Here, D represents the collected multimodal dataset, T represents the classical poetry text feature matrix, I represents the related image feature matrix, and U represents the user style preference data matrix. Function f represents the mapping relationship between data collection and integration.

[0054] In this embodiment, step S1 can achieve comprehensive and accurate collection of multimodal cultural data. The algorithms and models used in this collection process fully reflect the integrity of the technical features and ensure the foundation for data processing in subsequent steps.

[0055] S2, processing multimodal cultural data to extract imagery, emotion, and style to form structured triplets;

[0056] Specifically, in step S2 of this embodiment, the collected multimodal cultural data is deeply processed to extract the imagery, emotion, and style therein to form structured triples, laying the foundation for the subsequent construction of a dynamic cultural tensor.

[0057] First, the acquired classical poetry and related image data are processed. The text data undergoes word segmentation and part-of-speech tagging. This word segmentation process uses a dictionary-based approach, with custom dictionaries and statistical language models used to improve accuracy. Next, the text is vectorized, using pre-trained models such as Word2Vec or BERT. Each word is mapped to a vector, forming a feature matrix for the text.

[0058] The sentiment analysis module then applies sentiment analysis tools to the segmented text to determine the sentiment of the sentence or phrase. Sentiment analysis involves a variety of algorithms, including, for example, sentiment lexicons, machine learning classification, or deep learning methods. This process generates a sentiment score for each text segment, expressed as follows:

[0059] E x =f(T x );

[0060] Among them, E x is the sentiment score of the xth text, T x is the feature vector of the x-th text, and the function f describes the calculation process of the sentiment score.

[0061] Next, the relevant images are analyzed using image analysis techniques to extract features. A convolutional neural network (CNN) is used to process the extracted images, generating corresponding image feature vectors and forming an image feature matrix. This process effectively identifies and encodes key image features, such as color, form, and texture, providing a foundation for subsequent image extraction.

[0062] After combining the text feature matrix and the image feature matrix, the features of both are integrated to generate a multimodal feature matrix. This matrix contains the image information in the text, the visual features in the image, and the relationship between the two, making subsequent data processing more efficient. This processing process is derived using the following formula:

[0063] M=g(T,I);

[0064] Among them, M is the multimodal feature matrix, T is the text feature matrix, I is the image feature matrix, and the function g represents the mapping of feature integration.

[0065] Finally, we use a deep learning model to classify and encode styles. By training the model to identify style features in historical data, we output a structured triple containing imagery, emotion, and style information. This triple is expressed as:

[0066] Q = (I, E, S);

[0067] Among them, Q is a structured triplet, I represents the image vector, E is the sentiment score, and S is the style label. This triplet will lay an important foundation for the subsequent construction of the dynamic cultural tensor.

[0068] Step S2 of this embodiment ensures the accurate extraction of imagery, emotion, and style, and the formed data structure has good operability, providing strong support for the generation of the target dynamic cultural tensor.

[0069] S3. Build a dynamic cultural tensor based on the triple information to reflect the relationship between imagery, emotion, style, and modality in the multimodal data. The tensor is subsequently used to initialize the quantum state.

[0070] Specifically, this embodiment has successfully completed the collection and processing of multimodal cultural data in the previous steps, laying the foundation for the subsequent construction of a dynamic cultural tensor. Specifically, the imagery, emotion, and style information extracted in step S2 will be used to construct a dynamic cultural tensor that comprehensively reflects the relationships between these elements. This process not only effectively integrates multimodal data but also provides a rich cultural context for subsequent poetry creation.

[0071] In this embodiment, the construction of the dynamic cultural tensor is primarily based on the structured triplet information obtained in step S2, which includes image vectors, sentiment scores, and style labels. First, the collected multidimensional data is optimized using a tensor decomposition algorithm, particularly CP decomposition or Tucker decomposition. Specifically, because cultural data is typically high-dimensional, effective dimensionality reduction is required to capture the core information.

[0072] In the process of constructing the dynamic culture tensor, we first define the relevant variables. Set the dimension of the tensor to R and express it through the following formula:

[0073]

[0074] Among them, C represents the constructed dynamic cultural tensor, represents the sum of r from 1 to R, R represents the dimension of the constructed tensor, indicating that there are R terms in the formula that need to be accumulated, λ r Represents the weight coefficient of the rth tensor component. Each λ r Used to control the influence of the corresponding factors on the results, a r ,b r ,c r Represents three factor vectors or matrices, which are usually used for tensor decomposition or to represent different aspects of data. r ,b r ,c r Represents a dimension or feature of the data, a r A factor that represents a certain aspect may represent a certain attribute or feature of the data. r and c r It also represents different characteristics or dimensions of data. Represents the Kronecker product, which is a product operation between two sets of vectors, matrices or tensors. Unlike traditional matrix multiplication, the Kronecker product is an element-by-element product, that is, a r ,b r ,c r Multiply corresponding elements

[0075] Alternatively, tensor construction can be achieved by minimizing the following loss function to obtain the optimal tensor rank:

[0076]

[0077] Among them, R represents the optimization result, C represents the dynamic culture tensor, and λ r represents the weight of the solution, a r ,b r ,c r ,d r represents the factor component of the tensor, ° represents the element-wise product, ||·|| F represents the Frobenius norm, argmin R′ It represents the process of finding R′ that minimizes the expression, i.e., the optimization process.

[0078] During implementation, alternating least squares (ALS) or gradient descent can be used to optimize tensor decomposition and obtain optimal parameters. This process involves a large number of matrix operations, which increases computational complexity. Therefore, in practical applications, technologies such as parallel computing can be used to improve efficiency.

[0079] Furthermore, the constructed dynamic cultural tensor will form the basis for subsequent quantum state generation and style transfer, making the generated content not only creative but also reflecting the profound cultural heritage and style characteristics expected by users.

[0080] S4. Generate quantum states based on user input to represent the desired style and imagery. The quantum states form the basis for subsequent style transfer.

[0081] Specifically, after completing the dynamic cultural tensor construction step in this embodiment, the next step is to generate quantum states. This process connects the multimodal cultural information extracted previously to prepare for subsequent poetry creation. Specifically, the generated quantum states will represent the user's desired style and imagery, thus providing a quantum computing foundation for adaptive poetry creation.

[0082] In this embodiment, the process of generating quantum states is based on user input and data from the dynamic cultural tensor. First, the system analyzes the text entered by the user to identify the imagery and sentiment contained therein. This process utilizes the imagery vectors and sentiment scores defined in the previous steps. By defining the text's feature vectors, the user's desired style and imagery can be converted into a quantum state representation.

[0083] In terms of mathematical modeling, the generation of quantum states can be represented as a complex vector in the form of:

[0084]

[0085] Among them, |ψ> represents the quantum state, represents the sum of i from 1 to N, where N is the total number of possible states in the quantum system, c i represents the complex coefficients associated with the ground state |i>. These coefficients reflect the contribution strength or probability amplitude of the quantum state on different ground states |i>, where |i> represents the ground state of the quantum state and is usually a standardized representation of the quantum state.

[0086] In this formula, the coefficient c i It can be obtained by analyzing the correlation between user input and dynamic culture tensor. It can be defined as:

[0087] c i = <v i |I·E·S>;

[0088] Among them, |v i > represents the quantum ground state, I, E, and S represent the image vector, sentiment score, and style label, respectively. This process ensures that the generated quantum state not only reflects the user's artistic intent but is also consistent with the cultural tensor features.

[0089] As an option, the generation of quantum states can also be combined with an adaptive gain algorithm by iteratively optimizing the complex coefficient c i , gradually approaching the quantum state that the user hopes to achieve. This adaptive algorithm not only improves the accuracy of quantum state generation, but also can adjust the generation strategy according to different user inputs.

[0090] During the optimization process, the system integrates the characteristics of the dynamic cultural tensor, the imagery entered by the user, and the sentiment score to ensure that the output quantum state has greater cultural adaptability and artistic expression. This optimization result ultimately helps the model generate poetry content that conveys the target imagery and emotion.

[0091] S5. Based on the dynamic cultural tensor, a sparse cultural Hamiltonian is constructed for style transfer. Specifically, in this embodiment, the generation of quantum states in the previous step lays the foundation for subsequent processes. Next, a cultural Hamiltonian is constructed to describe the interactions between the elements in the dynamic cultural tensor. This step is key to further achieving style transfer and lays the physical foundation for generating artistic, adaptive poetry.

[0092] In this embodiment, the construction of the Hamiltonian is mainly based on the dynamic cultural tensor information obtained in step S3. First, we define the sparse cultural Hamiltonian, which is expressed in mathematical form as follows:

[0093]

[0094] in, represents the Hamiltonian, represents the summation of r from 1 to R, R is the upper limit of the summation, λ r represents the weight coefficient, Denotes the three different components of the Paul i matrix, σ z , σ x , σ y is the standard notation for the Paul i matrix, a r , b r , c r represents the label or index associated with each Paul i matrix, Represents the Kronecker product, that is, the tensor product.

[0095] The use of Paul i matrices in the formula effectively captures the coherence and dynamic relationships between quantum states. By designing sparsity features, the Hamiltonian not only reflects the interactions between cultural elements but also reduces computational complexity, making the calculation process more efficient.

[0096] As an option, the process of constructing the Hamiltonian can be optimized using the alternating least squares (ALS) method by continuously adjusting the weight coefficient λ r , so that the generated Hamiltonian can better conform to the actual cultural data. Specifically, a loss function can be designed to evaluate the accuracy of the Hamiltonian, expressed as:

[0097]

[0098] Among them, L loss Represents the loss function, which is usually used to quantify the error between the output of the model and the desired target. Represents the current estimated Hamiltonian, H target represents the target Hamiltonian, ||·|| F Represents the Frobenius norm, which is the square root of the sum of the squares of all elements in the matrix. The Frobenius norm is often used to measure the difference between matrices and is often used as an error measure in optimization problems. Represents the square of the Frobenius norm between the Hamiltonian estimate and the target value.

[0099] In actual implementation, the generated Hamiltonian will be connected to the quantum state generated in the previous step, providing the physical context for subsequent style transfer. Through appropriate quantum operations, this Hamiltonian can be applied to quantum computing devices to achieve control and adjustment of the generated results.

[0100] Specifically, the Hamiltonian is not constructed in isolation, but rather continuously interacts with the information in the dynamic cultural tensor. This connection ensures that the generated poetry remains consistent in both artistic and cultural aspects, and also enables the smooth execution of the style transfer process.

[0101] S6. Evolving the quantum state through a variational quantum circuit to achieve style transfer and optimize style fidelity;

[0102] Specifically, after constructing the cultural Hamiltonian, this embodiment evolves the quantum state using a variational quantum circuit to achieve style transfer and generate adaptive artwork. This step connects the previously established quantum state with the Hamiltonian, providing practical support for the creation of culturally rich poetry.

[0103] In this embodiment, the evolution of the quantum state is based on a given Hamiltonian. First, we will use the basic law of quantum state evolution, the Schrödinger equation, which is mathematically expressed as:

[0104]

[0105] Where i represents the imaginary unit, satisfying i 2 =-1, which is very common in quantum mechanics and complex mathematics. represents the reduced Planck constant, which is a fundamental constant in quantum mechanics and is defined as where h is Planck's constant, represents the partial derivative with respect to time t, which is used to describe the evolution of the system over time, and |ψ(t)> represents the change of quantum state ψ over time t. The quantum state can describe the evolution of the state of the system over time through time evolution. represents the Hamiltonian, the energy operator for a quantum system, typically used to describe the system's total energy. The Hamiltonian includes the system's kinetic and potential energy terms. |ψ(t)> appears again, representing the state of the quantum system at time t. It is the result of the Hamiltonian, namely, the time evolution of the quantum state.

[0106] As an alternative, a variational method can be used to approximate the solution to the above equation. Specifically, a variational quantum circuit can be introduced to adjust the quantum state and achieve the transfer of poetic style. In the quantum circuit, basic quantum gates such as rotation gates and CNOT gates will be used to manipulate qubits. The specific operation can be expressed as:

[0107] |ψ new >=U(θ)|ψ old >

[0108] Among them, U(θ) is the quantum operation related to the parameter θ, which includes the composite effect of multiple quantum gates, |ψ old > is the quantum state before the quantum operation is applied, and |ψ new > is the new quantum state after the transformation.

[0109] During implementation, the design of quantum circuits will be based on the data in the dynamic cultural tensor to ensure that the generated content is consistent with user needs. Therefore, precise quantum operations will not only affect the output, but also directly influence the style and emotional characteristics of the output poem.

[0110] Specifically, the design of the loss function will reflect the goal of style generation and can be expressed as:

[0111]

[0112] in, represents the loss function, 1 represents the constant term, represents the inner product norm between quantum states, ψ f represents the quantum state after style transfer, represents the quantum state of the target style, Represents the quantum state |ψ f > and The inner product between , μ represents the regularization parameter, Represents the Hamiltonian The trace of the square of represents the Hamiltonian of the quantum system, and Tr represents the trace of the matrix.

[0113] By continuously optimizing the loss function, the quantum state is iteratively updated. In this way, the evolved quantum state will fully reflect the desired cultural imagery and emotional attributes based on the feedback provided by users and the target content.

[0114] During execution, parallel computing techniques can be incorporated to improve the computational efficiency of the evolution process, particularly when dealing with high-dimensional quantum states. By optimizing the quantum state evolution process, the generated cultural content can be guaranteed to retain its originality while possessing profound cultural connotations.

[0115] S7. Cross-modal semantic correction: During the generation phase, cross-modal semantic correction is performed on the style transfer results to ensure cultural and semantic consistency of the generated poems.

[0116] Specifically, after completing quantum state evolution and generating dynamic artistic content that carries cultural connotations, this embodiment evaluates and optimizes the generated artwork. This process is crucial for ensuring that the created poetry meets user needs and also provides a basis for subsequent adjustments and improvements.

[0117] In this embodiment, the evaluation and optimization of generated artwork primarily relies on user feedback and an automated scoring mechanism. First, to quantify the quality of generated content, the system will design a multi-level evaluation criteria. These criteria include semantic consistency, cultural relevance, emotional resonance, and stylistic adaptability.

[0118] In general, the quality assessment of generated content can be expressed as the following mathematical model:

[0119] Q=w1S1+w2S2+w3S3+w4S4;

[0120] Among them, Q represents the comprehensive evaluation score, S1, S2, S3 and S4 are the scores corresponding to the above four evaluation criteria respectively, and w1, w2, w3 and w4 are the corresponding weight coefficients, reflecting the importance of each evaluation criterion in the comprehensive score.

[0121] During the evaluation process, user feedback is collected in both explicit and implicit ways. Explicit feedback includes ratings and comments, while implicit feedback is collected by analyzing user interaction behaviors and preferences.

[0122] Alternatively, the system can utilize natural language processing (NLP) technology to analyze user reviews and ratings and quantify sentiment. This process involves performing sentiment analysis on text data, such as using a sentiment lexicon or deep learning model, to calculate sentiment scores for user feedback, thereby reflecting the emotional resonance of the generated work.

[0123] Specifically, sentiment analysis can be achieved through the following formula:

[0124] E user =f(T feedback );

[0125] Among them, E user is the sentiment score extracted from user feedback, T feedback is the feature vector of the user feedback text, and function f represents the sentiment analysis process. The sentiment score will be combined with other evaluation criteria to form a comprehensive work quality analysis report.

[0126] After completing the quality assessment, the system will optimize based on the feedback. Specifically, genetic algorithms or particle swarm optimization algorithms can be used to continuously adjust the parameters used in the generation algorithm. For example, the parameters in the quantum circuit can be optimized to generate poetry content that better meets the user's expectations.

[0127] In one possible implementation, the system applies the following optimization formula:

[0128] P new =Pold +α·ΔP;

[0129] Here, P new is the optimized parameter, P old represents the currently used parameters, α is the learning rate, and ΔP is the adjustment amount calculated based on the evaluation feedback.

[0130] Through this feedback and optimization mechanism, the system continuously improves the quality of generated content. This process not only increases satisfaction with the final work, but also makes the adaptive poetry creation project more targeted and feasible.

[0131] In summary, the implementation of step S7 provides a scientific and reasonable basis for optimizing the generated artistic content, while ensuring the effectiveness of the generated content in terms of artistry and cultural adaptability.

[0132] S8, output the generated final poem and image description, forming a work with multimodal fusion features, and feed back the generated results to the user;

[0133] Specifically, after completing the evaluation and optimization of the generated artwork, the next step in this embodiment is to publish and disseminate the resulting cultural content. This process is crucial, allowing users and the general public to experience the innovative art forms brought about by technology and achieving the continuous dissemination and sharing of culture.

[0134] In this embodiment, the published content will take into account user feedback and optimization results to ensure that the generated poetry achieves optimal results in terms of style, emotion, and cultural relevance. Specifically, the publishing process of the generated poetry includes multiple adaptation formats, such as text, audio, or video, to suit the needs and preferences of different users.

[0135] During the publishing process, the generated text content must first be converted into a format suitable for visual and auditory communication. Specifically, text-to-speech (TTS) technology can be used to convert the generated poetry content into audio output. This process typically uses a deep learning-based speech synthesis model, which can generate natural and fluent speech.

[0136] Alternatively, the text-to-speech process can be expressed as:

[0137] Audio = T(Text);

[0138] Here, Audio represents the generated audio content, T is the function used to perform the transformation, and Text is the generated poem text. By adjusting the parameters of the synthesis model, the generated audio quality can be optimized, thereby improving the user experience.

[0139] Furthermore, the generated poetry can be incorporated into multimedia works, such as videos, to provide rich audiovisual effects. In this case, the generated poetry will be combined with images, music, and animation to form expressive artistic videos. Such works will support wider dissemination and attract a more diverse audience.

[0140] Typically, to enhance the interactivity of published works, online platforms offer users feedback and commentary features. This design allows users to share their feelings, which not only helps the creative team collect more data for the next round of optimization, but also promotes interaction with the audience. User feedback can be quantified using the following formula:

[0141]

[0142] Among them, F represents the overall feedback score, u j represents the participation of user j (which may indicate user activity, weight or importance), r j represents the rating of a work by user j, and M represents the total number of users who participated in the feedback.

[0143] In one possible implementation, the system could analyze user feedback in real time and adjust its dissemination strategy. For example, if a user responds positively to a certain style or theme, the system would prioritize similar works to meet the user's preferences.

[0144] Finally, the generated poems and related content will be widely disseminated through various channels such as social media, cultural and artistic platforms, and mobile applications. This not only ensures the accessibility of cultural and artistic works, but also enhances the user experience and encourages more people to participate in the interaction of cultural creation.

[0145] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive poetry creation method that integrates traditional cultural elements with modern technology, characterized by: The following steps are included: S1. Collect multimodal cultural data, including classical poetry texts, related images, and user style preference data; S2, processing multimodal cultural data to extract imagery, emotion, and style to form structured triplets; S3. Build a dynamic cultural tensor based on the triple information to reflect the relationship between imagery, emotion, style, and modality in the multimodal data. The tensor is subsequently used to initialize the quantum state. S4. Generate quantum states based on user input to represent the desired style and imagery. The quantum states form the basis for subsequent style transfer. S5. Construct a sparse cultural Hamiltonian based on the dynamic cultural tensor for style transfer. S6. Evolving the quantum state through a variational quantum circuit to achieve style transfer and optimize style fidelity; S7. Cross-modal semantic correction: During the generation phase, cross-modal semantic correction is performed on the style transfer results to ensure cultural and semantic consistency of the generated poems. S8. Output the generated final poem and image description to form a work with multimodal fusion features, and feed back the generated results to the user.

2. The adaptive poetry creation method that integrates traditional cultural elements and modern technology according to claim 1 is characterized in that: The multimodal cultural data in step S1 further includes historical text records uploaded by the user interaction platform, so that the system can adjust the creative style in a targeted manner.

3. The adaptive poetry creation method that integrates traditional cultural elements and modern technology according to claim 1 is characterized in that: The extraction of imagery, emotion, and style in step S2 is achieved through a deep learning model, which is a bidirectional long short-term memory network or a convolutional neural network.

4. The adaptive poetry creation method that integrates traditional cultural elements and modern technology according to claim 1 is characterized in that: The construction process of the dynamic cultural tensor in step S3 adopts a tensor decomposition algorithm, and the optimization criterion is to select the rank of the tensor by minimizing the following loss function: Among them, R represents the optimization result, C represents the dynamic culture tensor, and λ r represents the weight of the solution, a r ,b r ,c r ,d r represents the factor components of the tensor, represents element-wise product, ||·|| F represents the Frobenius norm, argmin R′ It represents the process of finding R′ that minimizes the expression, i.e., the optimization process.

5. The adaptive poetry creation method integrating traditional cultural elements and modern technology according to claim 1 is characterized in that: The process of generating the quantum state in step S4 comprises the following steps: After receiving user input, analyzing the user input text to identify target style and imagery; Based on the recognition results, feature information related to the target style and imagery is extracted from the dynamic cultural tensor; The quantum state is constructed using the selected feature information so that the quantum state accurately represents the artistic style and emotional expression required by the user.

6. The adaptive poetry creation method integrating traditional cultural elements and modern technology according to claim 1 is characterized in that: The sparse cultural Hamiltonian in step S5 is constructed by the following formula: in, represents the Hamiltonian, represents the summation of r from 1 to R, R is the upper limit of the summation, λ r represents the weight coefficient, Denotes the three different components of the Paul i matrix, σ z , σ x , σ y is the standard notation for the Paul i matrix, a r , b r , c r represents the label or index associated with each Paul i matrix, Represents the Kronecker product, that is, the tensor product.

7. The adaptive poetry creation method integrating traditional cultural elements and modern technology according to claim 1 is characterized in that: In step S6, the quantum state is evolved and the style fidelity is optimized by using a variational quantum circuit, and the loss function includes at least the following parts: in, represents the loss function, 1 represents the constant term, represents the inner product norm between quantum states, ψ f represents the quantum state after style transfer, represents the quantum state of the target style, Represents the quantum state |ψ f > and The inner product between , μ represents the regularization parameter, Represents the Hamiltonian The trace of the square of represents the Hamiltonian of the quantum system, and Tr represents the trace of the matrix.

8. The adaptive poetry creation method integrating traditional cultural elements with modern technology according to claim 1 is characterized in that: The cross-modal semantic correction step in step S7 adopts a meta-learning algorithm to optimize the cultural adaptability of the generated results to achieve high consistency between the generated poems and the image descriptions.

9. The adaptive poetry creation method integrating traditional cultural elements with modern technology according to claim 1 is characterized in that: In step S1, the user style preference data is updated through the history generation result feedback system.

10. The adaptive poetry creation method integrating traditional cultural elements and modern technology according to claim 1 is characterized in that: The output generation in step S8 is combined with a cultural tuning mechanism, which can adapt to the needs of users with different cultural backgrounds and automatically adjust the parameter settings of the generation model.

Images