User experience attribute extraction analysis method and system for complex product research and development

Abstract

The application provides a user experience attribute extraction analysis method and system for complex product research and development, and relates to the technical field of data analysis. The user experience attribute extraction analysis method comprises the following steps: obtaining user generated content (UGC) related to a target new product and preprocessing to obtain a target UGC dataset; converting the target UGC dataset into a domain-specific corpus, constructing a BERT-based sentence embedding model, converting sentences into high-dimensional sentence embedding vectors, clustering the sentences, integrating the user experience attributes obtained by clustering, calculating the weight of each user experience attribute, generating an attribute system and performing iterative optimization. Through semantic analysis of the context, the application ensures that the extracted attributes are more in line with the actual needs of users. Then, the user experience attributes obtained by clustering are integrated, the weights are allocated, the attribute system is generated and iterative optimization is performed, and the optimized target attribute system is output to meet the needs of user experience orientation and complex product research and development.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, specifically to a method and system for extracting and analyzing user experience attributes for complex product development. Background Technology

[0002] With the rapid development of next-generation information technology, the decision-making environment is becoming increasingly dynamic and complex, especially in the research and development of complex products. Traditional user experience attribute design methods typically rely on predefined, fixed sets of attributes, which may be applicable in static environments, but their limitations become particularly apparent in rapidly changing market and technological environments. Traditional attribute systems are often based on standard documents or historical data, lacking sensitivity to real-time market dynamics and user needs, making it difficult to adapt to the rapid evolution of today's technological innovations and market demands.

[0003] In related technologies, traditional systems relying on static attribute sets struggle to capture and reflect real-time changes in consumer preferences and behavioral trends when facing the challenges of rapidly changing market and technological environments. For example, consumer purchasing behavior and product preferences may change significantly in the short term, and traditional methods struggle to adapt quickly to these changes. Data-driven analytics offers new approaches to addressing these challenges. Utilizing user-generated content (UGC), including online reviews, social media interactions, and direct user feedback, real-time data can directly capture consumer behavior and preferences, providing immediate feedback for the continuous optimization of complex product development and services.

[0004] However, many existing user experience analysis systems rely on pre-defined standard attribute sets, resulting in significant lag in responding to real-world user experiences. Furthermore, they are typically based on manually set attribute weighting schemes, which are not only highly subjective but also fail to fully reflect the actual needs of users. Faced with complex, diverse, and dynamic user experience data, existing models often lack flexibility in attribute extraction and weighting, making real-time response and dynamic adjustment difficult. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides a method and system for extracting and analyzing user experience attributes for complex product development. This solves the problems of inflexibility in attribute extraction and weight allocation, difficulty in real-time response and dynamic adjustment, and poor effectiveness and accuracy in current user experience attribute analysis.

[0006] To achieve the above objectives, this application provides the following technical solution:

[0007] In a first aspect, embodiments of this application provide a method for extracting and analyzing user experience attributes for complex product development. This method includes: acquiring user-generated content (UGC) related to the target new product and preprocessing it to obtain a target UGC dataset; converting the target UGC dataset into a domain-specific corpus for use in complex product development; constructing a BERT-based sentence embedding model to convert a large number of sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information; clustering sentences using the sentence embedding vectors to determine potential user experience attributes; integrating the clustered user experience attributes and calculating the weight of each user experience attribute for weight allocation; generating an attribute system based on the weight allocation results and iteratively optimizing it to output an optimized target attribute system that meets the needs of user experience orientation and complex product development.

[0008] According to a first aspect of the embodiments of this application, the aforementioned acquisition of user-generated content (UGC) related to the target new product and the preprocessing to obtain the target UGC dataset may specifically include the following steps: determining the target new product and keywords related to the target new product; defining a list of related keywords based on the characteristics of the new product; acquiring a UGC dataset containing each keyword in the keyword list through an API interface or web page parsing method; and performing data cleaning, deduplication, anonymization, and formatting on the raw data in the UGC dataset to obtain the target UGC dataset.

[0009] According to a first aspect of the embodiments of this application, the aforementioned transformation of the target UGC dataset into a domain-specific corpus for complex product development may specifically include the following steps: dividing the target UGC dataset into sentence-level datasets to facilitate fine-grained text analysis; performing lexical analysis and word segmentation, using natural language processing tools to decompose each sentence in the sentence-level dataset into word sequences; constructing a domain-specific vocabulary, summarizing the words of all sentences to form a domain corpus; defining corpus quality assessment indicators to ensure the relevance and accuracy of the corpus; evaluating the quality of the corpus by calculating the values ​​of the corpus quality assessment indicators; and, if the corpus quality assessment indicators do not reach a preset threshold, re-dividing the target UGC dataset into sentence-level datasets.

[0010] The corpus quality assessment index satisfies the following expression:

[0011]

[0012] Where Q is the corpus quality assessment index, and N relevant N represents the number of relevant sentences extracted. total N represents the total number of sentences in the corpus; errors This indicates the number of errors or inconsistencies found in the corpus.

[0013] According to a first aspect of the embodiments of this application, the aforementioned construction of a BERT-based sentence embedding model, which transforms a large number of sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information, may specifically include the following steps: loading a pre-trained BERT model and fine-tuning it according to the domain corpus to obtain a domain-specific target BERT model; for each sentence in the sentence-level dataset, generating a high-dimensional sentence embedding vector using the target BERT model; representing the set of multiple sentence embedding vectors in matrix form, and introducing a piecewise soft relative cosine similarity metric to capture the semantic relationships between sentences.

[0014] According to a first aspect of the embodiments of this application, the aforementioned clustering of sentences using sentence embedding vectors to determine potential user experience attributes may specifically include the following steps: using the UMAP algorithm to reduce the dimensionality of the sentence embedding vectors to a two-dimensional or three-dimensional space to obtain a set of vectors after dimensionality reduction; using the HDBSCAN algorithm to perform density clustering on the dimensionality-reduced data, automatically identifying cluster structures and noise points in the data to obtain clustering results; adjusting the hyperparameters of the UMAP algorithm and the HDBSCAN algorithm based on the clustering results; wherein, the hyperparameters of the UMAP algorithm include the number of neighbors and the minimum distance; the hyperparameters of the HDBSCAN algorithm include the minimum cluster size and the minimum number of samples; evaluating the clustering results using preset evaluation metrics and selecting the optimal clustering model; the preset evaluation metrics include: silhouette coefficient, Calinski-Harabasz index, and Davies-Bouldin index; and identifying the sentence set corresponding to each cluster based on the clustering results, determining that the identified sentence set is a potential user experience attribute.

[0015] According to a first aspect of the embodiments of this application, the aforementioned integration of user experience attributes obtained by clustering and the calculation of the weight of each user experience attribute for weight allocation may specifically include the following steps: extracting keywords from each attribute cluster corresponding to the user experience attributes to construct an attribute word set; using CountVectorizer to convert the attribute word set into a feature vector and counting the frequency of each word in the attribute cluster; calculating the attribute-level word frequency-inverse document frequency of each word to measure the importance of each word relative to the whole; calculating the weight of each attribute set and normalizing the weight of each attribute set to obtain a first relative weight; and calculating the second relative weight of each word in its respective attribute set to characterize the importance of the word in the attribute set.

[0016] According to a first aspect of the embodiments of this application, the aforementioned generation of an attribute system based on the weight allocation result and its iterative optimization, outputting an optimized target attribute system to meet the needs of user experience orientation and complex product development, may specifically include the following steps: selecting the top K attributes with higher weights based on the first relative weight of the attribute set and the second relative weight of the attribute word to construct an attribute system, where K is a positive integer; calculating the similarity between attribute sets using a similarity measurement method, merging attribute sets with similarity higher than a preset similarity threshold to reduce redundancy; wherein the similarity threshold is determined based on the average and standard deviation of multiple similarities; iteratively performing attribute merging and weight recalculation operations until the attribute system is stable, and outputting the optimized target attribute system and its corresponding target weights.

[0017] Secondly, embodiments of this application provide a user experience attribute extraction and analysis system for complex product development. This system includes: an acquisition module, a first conversion module, a second conversion module, a clustering module, an integration module, and an iterative optimization module. The acquisition module acquires user-generated content (UGC) related to the target new product and preprocesses it to obtain a target UGC dataset. The first conversion module converts the target UGC dataset into a domain-specific corpus for use in complex product development. The second conversion module constructs a BERT-based sentence embedding model, converting a large number of sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information. The clustering module uses the sentence embedding vectors to cluster sentences to determine potential user experience attributes. The integration module integrates the clustered user experience attributes and calculates the weight of each attribute for weight allocation. The iterative optimization module generates an attribute system based on the weight allocation results and performs iterative optimization, outputting an optimized target attribute system that meets the needs of user experience orientation and complex product development.

[0018] Thirdly, embodiments of this application provide an electronic device, which includes: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the user experience attribute extraction and analysis method for complex product development described in the first aspect above.

[0019] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program or instructions, which, when executed by a processor, implement the user experience attribute extraction and analysis method for complex product development described in the first aspect above.

[0020] This application provides a method and system for extracting and analyzing user experience attributes for complex product development. Compared with existing technologies, it has the following advantages:

[0021] This application proposes a method for extracting and analyzing user experience attributes. Based on user-generated content (UGC) and combined with natural language processing (NLP) technology for automated processing, this application transforms the target UGC dataset into a corpus and introduces a BERT-based sentence embedding model to enhance the semantic understanding capability of attribute extraction and ensure the accuracy of the analysis results. This sentence embedding model transforms a large number of sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information and clusters the sentences to determine potential user experience attributes. Through semantic analysis of the context, it ensures that the extracted attributes better match the actual needs of users. Then, the user experience attributes obtained from the clustering are integrated, weighted, and an attribute system is generated and iteratively optimized. The optimized target attribute system is output to meet the needs of user experience orientation and complex product development. This invention can effectively handle long texts and large-scale corpora, reduce computational complexity, and provide more reliable data support for complex product development. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a user experience attribute extraction and analysis method for complex product development provided in an embodiment of this application;

[0024] Figure 2 yes Figure 1 An exemplary process diagram of S120;

[0025] Figure 3 This is a schematic diagram of the structure of a user experience attribute extraction and analysis system for complex product development provided in an embodiment of this application;

[0026] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0029] This application provides a method and system for extracting and analyzing user experience attributes for complex product development. It solves the problems of insufficient flexibility in attribute extraction and weight allocation, difficulty in real-time response and dynamic adjustment, and poor effectiveness and accuracy in current user experience attribute analysis.

[0030] The technical solution in this application is to solve the above-mentioned technical problems, and the general idea is as follows:

[0031] With the rapid development of next-generation information technology, the decision-making environment is becoming increasingly dynamic and complex, especially in the research and development of complex products. Traditional user experience attribute design methods typically rely on predefined, fixed sets of attributes, which may be applicable in static environments, but their limitations become particularly apparent in rapidly changing market and technological environments. Traditional attribute systems are often based on standard documents or historical data, lacking sensitivity to real-time market dynamics and user needs, making it difficult to adapt to the rapid evolution of today's technological innovations and market demands.

[0032] When faced with the challenges of a rapidly changing market and technological environment, traditional systems relying on static attribute sets struggle to capture and reflect immediate changes in consumer preferences and behavioral trends. For example, consumer purchasing behavior and product preferences may change significantly in the short term, and traditional methods struggle to adapt quickly to these changes. Therefore, building a user experience attribute analysis system that can dynamically integrate and continuously update user-generated content (UGC) becomes crucial to solving this problem.

[0033] Data-driven analytics offers new avenues for addressing these challenges. Leveraging user-generated content (UGC), including online reviews, social media interactions, and direct user feedback, real-time data can directly capture consumer behavior and preferences, providing immediate feedback for the continuous optimization of complex product development and services. However, many existing user experience analytics systems do not fully utilize these real-time data sources, lacking effective dynamic update mechanisms and still relying on pre-defined standard attribute sets, resulting in significant lag in responding to actual user experiences.

[0034] Furthermore, traditional methods are typically based on manually set attribute weighting schemes, which are not only highly subjective but also fail to fully reflect the actual needs of users. Faced with complex, diverse, and dynamic user experience data, existing models often lack flexibility in attribute extraction and weighting, making real-time response and dynamic adjustment difficult. Therefore, how to automate attribute extraction using UGC data and optimize attribute weighting through data-driven approaches is a key issue in improving the effectiveness and accuracy of user experience attribute analysis systems. Specifically, existing technologies have the following shortcomings:

[0035] (1) High Subjectivity: Because user experience analysis is often highly subjective, attribute systems that rely on manual definitions by experts or users are often biased and tend to reflect individual experiences and understandings, rather than comprehensively and objectively reflecting the real experiences and needs of all users. For example, in product design or service optimization, attributes set by experts may rely too much on their knowledge of specific fields, while ignoring the real feelings and feedback of ordinary users in actual use. This bias can lead to decision-makers being unable to accurately grasp user needs and preferences when developing or optimizing complex products, thus affecting the effectiveness of decisions and market competitiveness.

[0036] (2) Difficulty in handling massive amounts of data: UGC data, including comments, feedback, social media posts, etc., is often unstructured and voluminous. Traditional manual processing methods are not only time-consuming and labor-intensive, but also fail to fully utilize the potential decision-making information contained within. When faced with millions or even billions of UGC data points, manual definition and analysis of this data proves inadequate, failing to efficiently extract valuable user experience information. Existing attribute systems also lack the ability to automate the processing of large-scale data, making it impossible for them to quickly respond to market changes and timely capture shifts in user needs.

[0037] (3) Insufficient Semantic Understanding: UGC data typically contains a large amount of unstructured data, such as long texts, images, and videos, while existing decision-making systems are mostly limited to the analysis and processing of short texts. For user-generated long text content, traditional keyword matching or simple text analysis methods struggle to accurately capture the contextual relationships and deeper semantic information. Especially for user comments and feedback containing complex semantic structures, these methods often only identify surface-level keywords and fail to deeply understand the user's true intentions in the comments. For example, a user may mention multiple positive and negative experiences when reviewing a mobile phone; traditional analysis methods may only focus on the positive or negative words, ignoring the specific scenarios and contextual details implied in the review. This deficiency in semantic analysis leads to the poor performance of existing user experience analysis systems when processing complex UGC data.

[0038] (4) Difficulty in Adjusting Dynamic Attribute Weights: Traditional attribute systems typically rely on predefined fixed weights. Attribute weights are set during system design and remain unchanged throughout the subsequent analysis. This fixed-weight approach fails to reflect the dynamic nature of user needs over time. In practical applications, user needs and the market environment are constantly changing, and the importance of certain attributes may vary significantly across different time periods or market scenarios. For example, in the smartphone market, early consumers may have focused more on hardware performance, such as processors and memory, while with technological advancements, users may place greater emphasis on camera performance or system smoothness. Traditional attribute systems cannot dynamically adjust weights based on these changes, resulting in analysis results that lag behind market dynamics and fail to accurately reflect the latest user needs.

[0039] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0040] The following section first introduces a user experience attribute extraction and analysis method for complex product development provided by the embodiments of this application.

[0041] This application provides a flowchart illustrating a user experience attribute extraction and analysis method for complex product development, as shown in the embodiments below. Figure 1 As shown, the user experience attribute extraction and analysis method may include the following steps S110-S160.

[0042] S110. Obtain user-generated content (UGC) related to the target new product and preprocess it to obtain the target UGC dataset.

[0043] S120. Transform the target UGC dataset into a domain-specific corpus for use in complex product development.

[0044] S130. Construct a BERT-based sentence embedding model to transform a large number of sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information.

[0045] S140. Use sentence embedding vectors to cluster sentences to determine potential user experience attributes.

[0046] S150. Integrate the user experience attributes obtained from clustering and calculate the weight of each user experience attribute for weight allocation.

[0047] S160. Based on the weight allocation results, generate an attribute system and perform iterative optimization, outputting the optimized target attribute system to meet the needs of user experience orientation and complex product development.

[0048] The above describes a specific implementation of a user experience attribute extraction and analysis method for complex product development provided in this application. It is understood that by utilizing user experience data from the target UGC dataset, an advanced text mining algorithm automatically generates a user experience attribute system, reducing human intervention and improving efficiency. This application also introduces an optimized algorithm for large-scale corpus processing to reduce computational resource consumption, improve the efficiency of the attribute system generation process, and accelerate the development cycle of complex products.

[0049] Specifically, this application proposes a user experience attribute extraction and analysis method. Based on user-generated content (UGC) and combined with natural language processing technology for automated processing, this application transforms the target UGC dataset into a corpus and introduces a BERT-based sentence embedding model to enhance the semantic understanding capability of attribute extraction and ensure the accuracy of the analysis results. This sentence embedding model transforms a large number of sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information and clusters the sentences to determine potential user experience attributes. Through semantic analysis of the context, it ensures that the extracted attributes are more in line with the actual needs of users.

[0050] Furthermore, this application integrates and weights the user experience attributes obtained from clustering, generates an attribute system, and iteratively optimizes it to output an optimized target attribute system that meets the needs of user experience orientation and complex product development. This invention can effectively process long texts and diverse corpora, reduce computational complexity, and provide more reliable data support for complex product development.

[0051] In one example, the aforementioned acquisition of user-generated content (UGC) related to the target new product and its preprocessing to obtain the target UGC dataset, specifically S110 may include the following steps:

[0052] S210. Identify the target new product and related keywords, and define a list of relevant keywords based on the characteristics of the new product;

[0053] S220. Obtain the UGC dataset containing each keyword in the keyword list through API interface or web page parsing method;

[0054] S230. Perform data cleaning, deduplication, anonymization, and formatting on the original data in the UGC dataset to obtain the target UGC dataset.

[0055] In some embodiments, such as Figure 2 As shown, the aforementioned transformation of the target UGC dataset into a domain-specific corpus for complex product development, specifically step S120, may include the following steps:

[0056] S310. Divide the target UGC dataset into sentence-level datasets to facilitate fine-grained text analysis;

[0057] S320 performs lexical analysis and word segmentation, using natural language processing tools to decompose each sentence in the sentence-level dataset into a sequence of words.

[0058] S330. Construct a domain-specific vocabulary, summarize the words in all sentences, and form a domain corpus;

[0059] S340. Define corpus quality assessment indicators to ensure the relevance and accuracy of the corpus;

[0060] S350. The quality of the corpus is evaluated by calculating the value of the corpus quality assessment index. If the corpus quality assessment index does not reach the preset threshold, the target UGC dataset is reclassified into sentence-level datasets.

[0061] For example, the corpus quality assessment index satisfies the expression:

[0062]

[0063] Where Q is the quality evaluation index of the corpus, and N relevant N represents the number of relevant sentences extracted. total N represents the total number of sentences in the corpus; errors This indicates the number of errors or inconsistencies found in the corpus.

[0064] It should be noted that the above quality assessment indicators are used to quantitatively evaluate the constructed domain corpus, ensuring the high quality and relevance of the corpus, which can lay a solid foundation for subsequent analysis.

[0065] In some embodiments, the aforementioned construction of a BERT-based sentence embedding model transforms a large number of sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information. Specifically, S130 may include the following steps:

[0066] S410. Load the pre-trained BERT model and fine-tune it according to the domain corpus to obtain a domain-specific target BERT model.

[0067] S420. For each sentence in the sentence-level dataset, generate a high-dimensional sentence embedding vector using the target BERT model;

[0068] S430. Represent the set of multiple sentence embedding vectors as a matrix and introduce a segmented soft relative cosine similarity metric to capture the semantic relationships between sentences.

[0069] It should be noted that the piecewise soft relative cosine similarity (SSRcos) introduced in this application is an improved similarity measurement method; the piecewise soft relative cosine similarity (SSRcos) satisfies the expression:

[0070]

[0071] in, This similarity measurement method can more accurately capture the semantic relationships between sentences, improving the accuracy of subsequent clustering.

[0072] It should also be noted that this similarity measurement method, through segmentation and soft similarity calculation, overcomes the shortcomings of traditional cosine similarity in high-dimensional space, more accurately capturing semantic differences between sentences and improving the discriminative power of the embedded vectors. Its features are as follows:

[0073] (1) Embedding segmentation: The sentence embedding vectors U and V are segmented into T′ sub-vectors. This segmentation allows similarity calculation to take into account the local context of the sub-regions in the embedding, and may capture the semantic importance changes of different parts of the sentence embedding.

[0074] (2) Mean centering: Before similarity calculation, each segment U... τ and V τ Mean centering is performed. This step reduces the impact of absolute differences in magnitude between segments, allowing similarity measurements to focus on the relative distribution of values ​​within each segment.

[0075] (3) Soft similarity measurement: similarity between segment components It is calculated using a soft similarity function based on Euclidean distance. This introduces a non-linear and more tolerant similarity measure compared to the traditional strictly linear cosine similarity.

[0076] (4) Cross-segment aggregation: The final similarity measure is the average of the similarity calculated for each segment. This aggregation assumes that each segment has equal importance in contributing to the overall similarity score.

[0077] In this embodiment, by introducing BERT-based sentence embedding technology and combining it with an improved similarity measurement method—Soft Relative Cosine Similarity (SSRcos)—semantic analysis capabilities are effectively enhanced. Traditional methods often struggle to understand the contextual relationships within long texts, leading to the omission of important information during attribute extraction. This invention uses the BERT model for deep embedding of user-generated content and combines it with SSRcos for semantic similarity measurement, enabling more accurate capture of contextual relationships within the text. By comparing embedded vectors segment by segment, SSRcos not only improves the understanding of the semantics of long texts but also effectively overcomes the limitations of traditional cosine similarity in processing high-dimensional vectors.

[0078] In some embodiments, the aforementioned use of sentence embedding vectors to cluster sentences to determine potential user experience attributes, i.e., the aforementioned S140 may specifically include the following steps:

[0079] S510. The UMAP algorithm is used to reduce the dimensionality of sentence embedding vectors to two or three dimensions, resulting in a set of vectors after dimensionality reduction. It can be understood that the goal of the UMAP algorithm is to preserve as much of the original high-dimensional dataset A = {α1, α2, ..., α...} as possible during the dimensionality reduction process. N The fuzzy topology in}, where The UMAP algorithm attempts to find a low-dimensional representation β = {β1, β2, ..., β}. N},in This goal is achieved by optimizing the following cross-entropy loss function:

[0080]

[0081] in, q ij =(1+a(β) i -β j ) 2b ) -1 p ij This indicates the selection of α in a higher-dimensional space. j As α i The probability of a neighbor, which is based on α. i The Gaussian distribution with adaptive bandwidth (local density adjustment) centered at the center is calculated.

[0082] Specifically, Where d(α) i ,α j ) is a distance metric, ρ i Is it to αi The nearest neighbor distance, σ i Adjust the decay rate of the Gaussian kernel in the local neighborhood. And q ij This indicates the selection of β in the low-dimensional representation. j As β i The probability of a neighbor is defined by a generalized Cauchy distribution: Here, a and b are fitting parameters used to reflect the topological structure of the high-dimensional data. This process helps ensure that the dimensionality-reduced data still reflects the structural characteristics of the original data.

[0083] S520. The HDBSCAN algorithm is used to perform density clustering on the dimensionality-reduced data, automatically identifying cluster structures and noise points to obtain the clustering results. Understandably, to adapt to cluster formation in multi-dimensional user experience data, the algorithm constructs a mutually reachable distance graph based on core distance, defined as follows:

[0084] d mreach-σ (β i ,β j )=

[0085] max(core σ (β i ),core σ (β j ),d(β i ,β j ))

[0086] Wherein d(β) i ,β j ) is point β i ,β j The Euclidean distance between them, core σ (β) is the core distance from point β to its σ-th nearest neighbor.

[0087] S530. Based on the clustering results, adjust the hyperparameters of the UMAP and HDBSCAN algorithms; the hyperparameters of the UMAP algorithm include the number of neighbors and the minimum distance; the hyperparameters of the HDBSCAN algorithm include the minimum cluster size and the minimum number of samples.

[0088] S540. Use preset evaluation indicators to evaluate the clustering results and select the optimal clustering model. The preset evaluation indicators include: Silhouette Coefficient, Calinski-Harabasz index and Davies Bouldin index.

[0089] For example, the formula for calculating the silhouette coefficient is as follows:

[0090]

[0091] Where a(i) is the average distance between point i and other points in its cluster, and b(i) is the average distance between point i and all the nearest points that are not in its cluster.

[0092] The Calinski-Harabasz index is calculated as follows:

[0093]

[0094] Where N is the total number of points, k is the number of clusters, and Tr(B k ) is the trace of the inter-cluster dispersion matrix, Tr(W) k ) is the trace of the cluster scatter matrix.

[0095] The formula for calculating the Davies Bouldin index is as follows:

[0096]

[0097] Where k is the number of clusters, σ i It is the distance from all elements in cluster i to the centroid c. i The average distance, c i and c j It is the centroid of clusters i and j, d(c i ,c j ) is the center of mass c i and c j The distance between them.

[0098] S550. Based on the clustering results, identify the set of sentences corresponding to each cluster, and determine the identified set of sentences as potential user experience attributes.

[0099] In this embodiment, the UMAP algorithm is used to embed high-dimensional vectors. Dimensional reduction to a lower-dimensional space Where d′ << d; This application also utilizes the HDBSCAN algorithm to perform density clustering on the dimensionality-reduced data, automatically identifying clusters of different densities and noise points. By combining the UMAP algorithm and the HDBSCAN algorithm, the complexity of high-dimensional data is effectively reduced, while avoiding the limitation of the preset number of clusters, and it can adaptively discover potential user experience attribute clusters.

[0100] In some embodiments, the user experience attributes obtained from clustering are integrated, and the weight of each user experience attribute is calculated for weight allocation. Specifically, S150 may include the following steps:

[0101] S610. Extract keywords from each attribute cluster corresponding to the user experience attributes and construct an attribute word set A. c={t1,t2,…,t k}

[0102] S620. Use CountVectorizer to convert the set of attribute words into feature vectors and count the frequency of each word in the attribute cluster.

[0103] S630. Calculate the attribute-level term frequency-inverse document frequency for each word to measure the importance of each word relative to the whole.

[0104] S640. Calculate the weight of each attribute set and normalize the weight of each attribute set to obtain the first relative weight.

[0105] S650. Calculate the second relative weight of each word in its respective attribute set to characterize the importance of the word in the attribute set.

[0106] In the embodiments of this application, it can be understood that the application accurately calculates the weight of attribute words, highlights important words in a specific attribute set, eliminates the unreasonable influence of high-frequency general words on weight allocation, ensures the scientificity and objectivity of the attribute system, and ensures that the generated attribute system can accurately reflect the user's actual preferences and needs by adjusting the weight of high-frequency and low-frequency attributes.

[0107] In some embodiments, based on the aforementioned weight allocation results, an attribute system is generated and iteratively optimized to output an optimized target attribute system that meets the needs of user experience orientation and complex product development. That is, the aforementioned S160 may specifically include the following steps:

[0108] S710. Based on the first relative weight of the attribute set and the second relative weight of the attribute words, select the top K attributes with higher weights to construct an attribute system, where K is a positive integer;

[0109] S720. Using a similarity measurement method, calculate the similarity between attribute sets, and merge attribute sets with similarity higher than a preset similarity threshold to reduce redundancy; wherein, the similarity threshold is determined based on the average and standard deviation of multiple similarities;

[0110] S730. Iteratively execute attribute merging and weight recalculation operations until the attribute system is stable, and output the optimized target attribute system and its corresponding target weights.

[0111] In the embodiments of this application, it can be understood that this application automatically identifies and merges similar or duplicate attribute sets by calculating the similarity and threshold between attribute sets, thereby reducing redundancy, optimizing the structure of the attribute system, and making it more refined and efficient.

[0112] In some embodiments, this application provides a user experience attribute extraction and analysis system 800 for complex product development, such as... Figure 3 As shown, the user experience attribute extraction and analysis system 800 may include the following modules:

[0113] The acquisition module 810 is used to acquire user-generated content (UGC) related to the target new product and preprocess it to obtain the target UGC dataset;

[0114] The first conversion module 820 is used to convert the target UGC dataset into a domain-specific corpus for use in complex product development.

[0115] The second transformation module 830 is used to construct a BERT-based sentence embedding model, which transforms a large number of sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information.

[0116] Clustering module 840 is used to cluster sentences using sentence embedding vectors to determine potential user experience attributes;

[0117] The integration module 850 is used to integrate the user experience attributes obtained from clustering and calculate the weight of each user experience attribute for weight allocation.

[0118] The iterative optimization module 860 is used to generate an attribute system based on the weight allocation results and perform iterative optimization, outputting the optimized target attribute system to meet the needs of user experience orientation and complex product development.

[0119] According to embodiments of this application, any and multiple modules among the acquisition module 810, the first transformation module 820, the second transformation module 830, the clustering module 840, the integration module 850, and the iterative optimization module 860 can be merged into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module.

[0120] In some embodiments, the acquisition module 810 can be specifically used for:

[0121] Identify the target new product and related keywords, and define a list of relevant keywords based on the characteristics of the new product;

[0122] Obtain a UGC dataset containing each keyword in the keyword list through API interface or web page parsing method;

[0123] The raw data in the UGC dataset is cleaned, deduplicated, anonymized, and formatted to obtain the target UGC dataset.

[0124] In some embodiments, the first conversion module 820 may specifically be used for:

[0125] The target UGC dataset is divided into sentence-level datasets to facilitate fine-grained text analysis;

[0126] Perform lexical analysis and word segmentation, and use natural language processing tools to decompose each sentence in the sentence-level dataset into a sequence of words;

[0127] Construct a domain-specific vocabulary, summarize the words in all sentences, and form a domain corpus;

[0128] Define corpus quality assessment metrics to ensure the relevance and accuracy of the corpus;

[0129] The quality of the corpus is evaluated by calculating the values ​​of the corpus quality assessment index. If the corpus quality assessment index does not reach the preset threshold, the target UGC dataset is reclassified into sentence-level datasets.

[0130] In some embodiments, the second conversion module 830 may specifically be used for:

[0131] Load the pre-trained BERT model and fine-tune it according to the domain corpus to obtain a domain-specific target BERT model;

[0132] For each sentence in the sentence-level dataset, a high-dimensional sentence embedding vector is generated using the target BERT model;

[0133] The set of multiple sentence embedding vectors is represented as a matrix, and a piecewise soft relative cosine similarity metric is introduced to capture the semantic relationships between sentences.

[0134] In some embodiments, the clustering module 840 may specifically be used for:

[0135] The UMAP algorithm is used to reduce the dimension of sentence embedding vectors to two-dimensional or three-dimensional space, resulting in a set of vectors after dimension reduction.

[0136] The HDBSCAN algorithm is used to perform density clustering on the dimensionality-reduced data, automatically identifying cluster structures and noise points in the data to obtain clustering results.

[0137] Based on the clustering results, the hyperparameters of the UMAP and HDBSCAN algorithms were adjusted. The hyperparameters of the UMAP algorithm include the number of neighbors and the minimum distance; the hyperparameters of the HDBSCAN algorithm include the minimum cluster size and the minimum number of samples.

[0138] The clustering results are evaluated using preset evaluation metrics to select the optimal clustering model. The preset evaluation metrics include: silhouette coefficient, Calinski-Harabasz index, and Davies-Bouldin index.

[0139] Based on the clustering results, the sentence set corresponding to each cluster is identified, and the identified sentence set is determined to be a potential user experience attribute.

[0140] In some embodiments, the integration module 850 may specifically be used for:

[0141] Extract keywords from each attribute cluster corresponding to the user experience attributes to construct an attribute keyword set;

[0142] The CountVectorizer is used to transform the set of attribute words into feature vectors, and the frequency of each word in the attribute cluster is counted.

[0143] Calculate the attribute-level term frequency-inverse document frequency for each word to measure the importance of each word relative to the whole;

[0144] Calculate the weight of each attribute set and normalize the weight of each attribute set to obtain the first relative weight;

[0145] Calculate the second relative weight of each word in its respective attribute set to characterize the importance of the word in the attribute set.

[0146] In some embodiments, the iterative optimization module 860 can be specifically used for:

[0147] Based on the first relative weight of the attribute set and the second relative weight of the attribute words, select the top K attributes with higher weights to construct an attribute system, where K is a positive integer;

[0148] The similarity measurement method is used to calculate the similarity between attribute sets and merge attribute sets with similarity higher than a preset similarity threshold to reduce redundancy; the similarity threshold is determined based on the average and standard deviation of multiple similarities.

[0149] Iteratively perform attribute merging and weight recalculation operations until the attribute system is stable, and output the optimized target attribute system and its corresponding target weights.

[0150] Figure 3 Each module in the system shown has the function of implementing each step in the aforementioned user experience attribute extraction and analysis method for complex product development, and can achieve the corresponding technical effect. For the sake of brevity, it will not be elaborated here.

[0151] In some embodiments, this application provides an electronic device, the structural schematic of which is shown below. Figure 4 As shown.

[0152] The electronic device may include a processor 910 and a memory 920 storing computer program instructions.

[0153] Specifically, the processor 910 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0154] Memory 920 may include mass storage for data or instructions. For example, and not limitingly, memory 920 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 920 may include removable or non-removable (or fixed) media. Where appropriate, memory 920 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 920 is non-volatile solid-state memory.

[0155] Memory 920 may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory 920 includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it can perform the operations described in any of the user experience attribute extraction and analysis methods for complex product development in the above embodiments.

[0156] The processor 910 reads and executes computer program instructions stored in the memory 920 to implement any of the user experience attribute extraction and analysis methods for complex product development in the above embodiments.

[0157] In one example, the electronic device may also include a communication interface 930 and a bus 900. Wherein, such as Figure 4 As shown, the processor 910, memory 920, and communication interface 930 are connected via bus 900 and communicate with each other.

[0158] The communication interface 930 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0159] Bus 900 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 900 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0160] Furthermore, in conjunction with the user experience attribute extraction and analysis method for complex product development described in the above embodiments, this application embodiment can provide a computer storage medium for implementation. This computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the user experience attribute extraction and analysis methods for complex product development described in the above embodiments.

[0161] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0162] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0163] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0164] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0165] In summary, compared with the prior art, this application has the following beneficial effects:

[0166] 1. This application significantly improves the efficiency and accuracy of attribute system generation through a data-driven algorithm. In traditional methods, attribute extraction relies on manual definitions by experts, which is not only time-consuming but also susceptible to subjective biases, resulting in an attribute system that lacks real-time performance and scientific rigor. This invention, by combining modern natural language processing technology, can automatically extract decision attributes from massive amounts of user-generated content (UGC), significantly reducing manual intervention.

[0167] 2. This application effectively reduces computational complexity by optimizing the computational architecture. Similarity calculation based on the traditional BERT model, especially for large-scale text embedding and comparison tasks, often consumes a large amount of computational resources, leading to increased time and cost. This invention significantly reduces the time and resource consumption of repeated calculations by pre-compiling sentence embeddings and combining them with an improved SSRcos algorithm.

[0168] 3. This application ensures that the extracted attributes better match the actual needs of users through semantic analysis of the context; then, it integrates and weights the user experience attributes obtained from clustering, generates an attribute system, and iteratively optimizes it, outputting an optimized target attribute system to meet the needs of user experience orientation and complex product development. This invention can effectively process long texts and diverse corpora, reduce computational complexity, and provide more reliable data support for complex product development.

[0169] 4. This application has broad application prospects. Besides its excellent performance in multi-attribute group decision-making problems, it can also be applied to various scenarios, such as user experience analysis, market research, and product evaluation. In market research, companies can use this invention to extract consumers' real needs and preferences to optimize product design. In user experience analysis, the platform can monitor user feedback in real time and adjust service content to improve user satisfaction. The broad applicability of this invention provides powerful data support in various complex decision-making scenarios, helping companies and organizations better cope with rapidly changing market environments and diverse user needs.

[0170] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting and analyzing user experience attributes for complex product development, characterized in that, include: Acquire user-generated content (UGC) related to the target new product and preprocess it to obtain the target UGC dataset; The target UGC dataset is transformed into a domain-specific corpus for use in complex product development; A BERT-based sentence embedding model is constructed to transform sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information; The sentences are clustered using the sentence embedding vectors to determine potential user experience attributes; The user experience attributes obtained from clustering are integrated, and the weight of each user experience attribute is calculated for weight allocation; Based on the weight allocation results, an attribute system is generated and iteratively optimized, and the optimized target attribute system is output to meet the needs of user experience orientation and complex product development. The process of converting the target UGC dataset into a domain-specific corpus for use in complex product development includes: The target UGC dataset is divided into sentence-level datasets to facilitate fine-grained text analysis; The construction of a BERT-based sentence embedding model, which transforms sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information, includes: Load the pre-trained BERT model and fine-tune it according to the domain corpus to obtain a domain-specific target BERT model; For each sentence in the sentence-level dataset, a high-dimensional sentence embedding vector is generated using the target BERT model; The set of multiple sentence embedding vectors is represented in matrix form, and a piecewise soft relative cosine similarity metric is introduced to capture the semantic relationships between sentences; The segmented soft relative cosine similarity measurement method is an improved similarity measurement method. Through segmentation and soft similarity calculation, it captures the semantic differences between sentences and improves the discriminativeness of the embedded vectors. The segmented soft relative cosine similarity measurement method includes: embedding segmentation, mean centering, soft similarity measurement, and cross-segment aggregation; the embedding segmentation includes: embedding sentences into vectors. and Divided into The sub-vectors allow similarity calculation to take into account the local context of the embedded sub-regions; the mean centering includes: performing mean centering on each segment before similarity calculation to reduce the influence of absolute magnitude differences between segments, so that the similarity measurement focuses on the relative distribution of values ​​within each segment; the soft similarity measurement includes: calculating the similarity between segment components using a soft similarity function based on Euclidean distance; the cross-segment aggregation includes averaging the similarity calculated for each segment.

2. The user experience attribute extraction and analysis method for complex product development as described in claim 1, characterized in that, The process of acquiring user-generated content (UGC) related to the target new product and preprocessing it to obtain the target UGC dataset includes: Identify the target new product and the keywords related to the target new product, and define a list of relevant keywords based on the characteristics of the new product; Obtain the UGC dataset containing each keyword in the keyword list through API interface or web page parsing method; The original data in the UGC dataset is cleaned, deduplicated, anonymized, and formatted to obtain the target UGC dataset.

3. The user experience attribute extraction and analysis method for complex product development as described in claim 1, characterized in that, The step of converting the target UGC dataset into a domain-specific corpus for use in complex product development also includes: Perform lexical analysis and word segmentation, and use natural language processing tools to decompose each sentence in the sentence-level dataset into a sequence of words; Construct a domain-specific vocabulary, summarize the words in all sentences, and form a domain corpus; Define corpus quality assessment metrics to ensure the relevance and accuracy of the corpus; The quality of the corpus is evaluated by calculating the value of the corpus quality evaluation index. If the corpus quality evaluation index does not reach the preset threshold, the target UGC dataset is reclassified into sentence-level datasets. The corpus quality evaluation index satisfies the following expression: in, The corpus is used as a quality assessment indicator. This indicates the number of relevant sentences extracted; This represents the total number of sentences in the corpus. This indicates the number of errors or inconsistencies found in the corpus.

4. The user experience attribute extraction and analysis method for complex product development as described in claim 1, characterized in that, The process of clustering sentences using the sentence embedding vectors to determine potential user experience attributes includes: The UMAP algorithm is used to reduce the dimension of the sentence embedding vector to a two- or three-dimensional space, resulting in a set of vectors after dimension reduction. The HDBSCAN algorithm is used to perform density clustering on the dimensionality-reduced data, automatically identifying cluster structures and noise points in the data to obtain clustering results. Based on the clustering results, the hyperparameters of the UMAP algorithm and the HDBSCAN algorithm are adjusted; wherein, the hyperparameters of the UMAP algorithm include the number of neighbors and the minimum distance; and the hyperparameters of the HDBSCAN algorithm include the minimum cluster size and the minimum number of samples. The clustering results are evaluated using preset evaluation metrics to select the optimal clustering model; the preset evaluation metrics include: silhouette coefficient, Calinski-Harabasz index and Davies-Bouldin index; Based on the clustering results, the sentence set corresponding to each cluster is identified, and the identified sentence set is determined to be a potential user experience attribute.

5. The user experience attribute extraction and analysis method for complex product development as described in claim 1, characterized in that, The process of integrating the user experience attributes obtained from clustering and calculating the weight of each user experience attribute for weight allocation includes: Extract keywords from each attribute cluster corresponding to the user experience attributes to construct an attribute word set; The set of attribute words is converted into a feature vector using CountVectorizer, and the frequency of each word in the attribute cluster is counted. Calculate the attribute-level term frequency-inverse document frequency for each word to measure the importance of each word relative to the whole; Calculate the weight of each attribute set and normalize the weight of each attribute set to obtain the first relative weight; Calculate the second relative weight of each word in its respective attribute set to characterize the importance of the word in the attribute set.

6. The user experience attribute extraction and analysis method for complex product development as described in claim 5, characterized in that, Based on the weight allocation results, an attribute system is generated and iteratively optimized to output an optimized target attribute system that meets the needs of user experience orientation and complex product development, including: Based on the first relative weight of the attribute set and the second relative weight of the attribute words, select the top K attributes with higher weights to construct an attribute system, where K is a positive integer; The similarity measurement method is used to calculate the similarity between attribute sets, and attribute sets with similarity higher than a preset similarity threshold are merged to reduce redundancy; wherein, the similarity threshold is determined based on the average and standard deviation of multiple similarities; Iteratively perform attribute merging and weight recalculation operations until the attribute system is stable, and output the optimized target attribute system and its corresponding target weights.

7. A user experience attribute extraction and analysis system for complex product development, characterized in that, include: The acquisition module is used to acquire user-generated content (UGC) related to the target new product and preprocess it to obtain the target UGC dataset; The first conversion module is used to convert the target UGC dataset into a domain-specific corpus for use in complex product development. The second transformation module is used to construct a BERT-based sentence embedding model to transform sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information. The clustering module is used to cluster sentences using the sentence embedding vectors to determine potential user experience attributes; An integration module is used to integrate the user experience attributes obtained from clustering and calculate the weight of each user experience attribute for weight allocation. The iterative optimization module is used to generate an attribute system based on the weight allocation results and perform iterative optimization, outputting the optimized target attribute system to meet the needs of user experience orientation and complex product development. The process of converting the target UGC dataset into a domain-specific corpus for use in complex product development includes: The target UGC dataset is divided into sentence-level datasets to facilitate fine-grained text analysis; The construction of a BERT-based sentence embedding model, which transforms sentences in the corpus into high-dimensional sentence embedding vectors to capture semantic information, includes: Load the pre-trained BERT model and fine-tune it according to the domain corpus to obtain a domain-specific target BERT model; For each sentence in the sentence-level dataset, a high-dimensional sentence embedding vector is generated using the target BERT model; The set of multiple sentence embedding vectors is represented in matrix form, and a piecewise soft relative cosine similarity metric is introduced to capture the semantic relationships between sentences; The segmented soft relative cosine similarity measurement method is an improved similarity measurement method. Through segmentation and soft similarity calculation, it captures the semantic differences between sentences and improves the discriminativeness of the embedded vectors. The segmented soft relative cosine similarity measurement method includes: embedding segmentation, mean centering, soft similarity measurement, and cross-segment aggregation; the embedding segmentation includes: embedding sentences into vectors. and Divided into The sub-vectors allow similarity calculation to take into account the local context of the embedded sub-regions; the mean centering includes: performing mean centering on each segment before similarity calculation to reduce the influence of absolute magnitude differences between segments, so that the similarity measurement focuses on the relative distribution of values ​​within each segment; the soft similarity measurement includes: calculating the similarity between segment components using a soft similarity function based on Euclidean distance; the cross-segment aggregation includes averaging the similarity calculated for each segment.

8. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the user experience attribute extraction and analysis method for complex product development as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program or instructions, which, when executed by a processor, implement the user experience attribute extraction and analysis method for complex product development as described in any one of claims 1 to 6.

Images