Dynamic construction method of service quality evaluation system fused with large language model

Abstract

The invention relates to a method for dynamically constructing a service quality evaluation system fused with a large language model, which comprises the following steps of: extracting themes, keywords and representative documents from a user evaluation text data set, constructing cue words and inputting the cue words into the large language model to generate candidate primary service dimensions divided by five service dimensions of an SERVQUAL model; the service quality candidate index item list is composed of candidate second-level topics and candidate third-level index items; constructing a multi-channel heterogeneous hypergraph set fusing sliding window co-occurrence and an external knowledge graph; graph structure information is extracted, joint prediction of emotion labels and emotion scores of user evaluation texts is carried out, and topic-level multi-class input factor spaces are constructed to be used for calculating final fusion weights; and performing comprehensive sorting and dynamic screening on the candidate three-level index items to form a multi-level service quality evaluation system index scale. According to the method, accurate modeling of the user perception quality is realized, and the emotional discrimination of the indexes of the service quality evaluation system can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of natural language processing and large language model application, and specifically relates to a method for dynamically constructing a service quality evaluation system integrating a large language model. Background Art

[0002] As a core component of modern service management, service quality evaluation directly impacts the efficiency of service providers' responses to user needs and the degree of resource allocation optimization. Traditional service quality evaluation methods, exemplified by the SERVQUAL model, construct an evaluation framework based on five service dimensions: reliability, responsiveness, assurance, tangibility, and responsiveness. In recent years, with significant breakthroughs in natural language processing, semantic modeling, and sentiment analysis using big language models, integrating big language model technology to enhance traditional service quality evaluation systems has become an emerging technology. For example, these technologies can effectively mine implicit service dimensions in user feedback, identify multi-level indicators, and leverage contextual semantic modeling to systematically model service attributes and user experience elements. This provides a technical foundation for intelligent and personalized dimension expansion and dynamic calibration of weighting factors within the SERVQUAL model in complex multi-source data environments.

[0003] However, current service quality evaluation methods based on the SERVQUAL model and its extended application systems face numerous challenges. First, traditional static questionnaire-based evaluation systems often struggle to adapt to data-driven, intelligently collaborative user service needs due to issues such as insufficient indicator adaptability, limited assessment accuracy, and delayed timeliness. There is an urgent need to break through the limitations of static frameworks and implement dynamic construction. Second, large language models, in specific application scenarios for dynamic SERVQUAL model construction, are still constrained by factors such as ambiguous user expressions leading to ambiguous identification of service dimensions, subjective user emotions increasing the difficulty of determining emotional polarity, and the high demands placed on model generalization capabilities due to contextual differences in different domains. These factors limit existing methods' ability to accurately capture real-world user experiences and the real-time construction and evaluation of dynamic service quality systems based on the SERVQUAL model. Therefore, constructing a service quality evaluation system that integrates the advanced reasoning capabilities of large language models, an emotion-driven mechanism, and a multi-factor dynamic weighting fusion strategy not only effectively complements existing traditional models but also provides a key technical foundation for achieving accurate service quality perception and real-time decision support in multi-source, heterogeneous data environments. Summary of the Invention

[0004] In order to solve the problems of abstract evaluation dimensions, strong subjectivity in weight allocation and lack of dynamic regulation capability in existing service quality evaluation system construction methods, the present invention proposes a dynamic construction method for a service quality evaluation system integrating a large language model. Based on the user evaluations of agricultural supplies and agricultural tools on the e-commerce platform, the unsupervised topic modeling method is used to extract keywords and representative evaluation content. Combined with the structured prompt construction mechanism, the initial indicator structure that can be input into the large language model is generated. At the same time, a multi-channel heterogeneous hypergraph neural network is constructed for semantic emotion-driven perception to enhance the fine modeling of service emotion expression. In the process of weight generation, a multi-factor screening strategy is integrated to hierarchically adjust the indicator structure to obtain a service quality evaluation system indicator scale, improve the reliability and discrimination of the service evaluation indicator results, effectively cope with the dynamic changes of multi-source heterogeneous service data, and support the automated evaluation of service quality for real scenarios.

[0005] The present invention is achieved through the following technical solutions.

[0006] A method for dynamically constructing a service quality evaluation system integrating a large language model, comprising the following steps: Step 1: Obtain a user review text dataset for product sales; Step 2: Extract topics, keywords, and representative documents from the user review text dataset, construct prompt words, and input them into the large language model to generate a list of candidate first-level service dimensions for the five service dimensions of the SERVQUAL model, as well as a list of candidate service quality indicators consisting of candidate second-level topics and candidate third-level indicators. Step 3: Use representative documents as the test set and the remaining user review text dataset as the training set to construct a multi-channel heterogeneous hypergraph collection that integrates sliding window co-occurrence and external knowledge graph; Step 4: The multi-channel heterogeneous hypergraph set is input into the multi-channel heterogeneous hypergraph neural network model to extract the graph structure information and perform joint prediction of the sentiment label and sentiment score of the user evaluation text; Step 5: Based on the predicted results of sentiment labels and sentiment scores, combined with user satisfaction scores, user review text content structure, and sentiment prediction information, a multi-class input factor space at the topic level is constructed. The final fusion weight of each topic, the fusion weight of the second-level topic, and the normalized fusion weight of the candidate first-level service dimension are calculated. Step 6: Comprehensively rank and dynamically screen the candidate third-level indicator items, and collect the core third-level indicator items to form a multi-level service quality evaluation system indicator scale.

[0007] Specifically, in step 2, the process of extracting topics, keywords, and representative documents is as follows: each user review text is converted into a fixed-dimensional semantic embedding vector through a sentence-level multilingual pre-trained model to form an overall text embedding vector matrix; then, the unified manifold approximation projection algorithm is used to reduce the dimensionality of the overall text embedding vector matrix, preserving the local density and global structure of the semantic space to obtain a reduced-dimensional feature matrix. The reduced-dimensional feature matrix is ​​then clustered using a hierarchical density clustering algorithm to output the topic set to which each user review text belongs; Keywords for each topic through the category-conditional inverse document frequency weight model; A representative document collection is selected based on the text distribution density and clustering confidence of each topic.

[0008] Specifically, in step 2, the keywords of the topic are spliced ​​with the corresponding representative documents, and combined with their semantic relevance to the five service dimensions of the SERVQUAL model to generate structured prompt input that conforms to the semantic topic. Prompt words are constructed for each topic, and the prompt words include a keyword set, a representative document set, and a user satisfaction score set.

[0009] Specifically, the specific process of step three is as follows: Each user review text in the training set and test set is divided into word sequences to form the basis of semantic units. All words constitute the vocabulary. Then, a continuous bag-of-words model based on word embedding is used to map each word into a q-dimensional embedding vector to obtain a word embedding matrix. For each user review text, the word sequence is divided into sliding windows to obtain multiple sliding window word sets; Define any two words based on the sliding window word set of all samples and calculate their point mutual information value, set the point mutual information threshold >0, select the candidate hyperedge set that meets the point mutual information threshold condition in the user evaluation text; An external knowledge graph (ConceptNet) is introduced as the second channel of the heterogeneous hypergraph. For each word, a semantic association set is defined by all connected words that have semantic edge relationships with it in ConceptNet retrieval. Then, a hyperedge set consisting of all the associated semantic adjacent words is constructed for each word. Then, the hyperedge sets of the two semantic channels of sliding window co-occurrence and external knowledge graph are unified and fused to form a hyperedge set of multi-channel heterogeneous hypergraph, and the triples of each user evaluation text for the multi-channel heterogeneous hypergraph are constructed, finally forming a multi-channel heterogeneous hypergraph set constructed by all user evaluation texts.

[0010] Specifically, in step 4, the processing of the multi-channel heterogeneous hypergraph neural network model includes: First, the graph structure of user review text is decoupled according to different semantic channels to form H channel graph structure information. The feature embedding of each user review text in each channel is constructed, and the node feature submatrix is ​​shared by all channels. For each feature embedding, multi-layer hypergraph convolution is used to embed the node features and aggregate the structural information of adjacent nodes at different scales layer by layer to obtain the local-global semantically aware embedding matrix. The node structure embedding finally obtained after passing L layers through multi-layer hypergraph convolution is dimensionally aligned for the node structure embeddings of all channels. Channel attention is introduced to generate fused node embedding, which is then input into the convolutional neural network. Local features are extracted through local convolution, and global pooling is performed to obtain a unified graph-level structure representation. Joint learning is then performed to achieve the prediction results of sentiment labels and sentiment scores.

[0011] Specifically, the multi-category input factor space described in step five includes word frequency factor, sentiment mean factor, document coverage factor, satisfaction driving factor, consistency factor, information density factor, text length factor and sentiment entropy factor, a total of eight types of indicator factors.

[0012] Specifically, in step 5, for each candidate third-level indicator item corresponding to each topic, all factor values ​​are normalized using the minimum-maximum method to obtain normalized factor values. Subsequently, a set of preset weight parameters are used to linearly weight each normalized factor value to obtain the initial fusion weight of each topic. The original fusion weight is then normalized and converted into the final fusion weight using the Softmax function. The final fusion weight of each topic is transferred to its corresponding candidate third-level indicator item to achieve downward binding of the weight. Then, a bottom-up weighted normalization strategy is used to complete the weight aggregation and obtain the normalized second-level topic fusion weight; at the same time, the normalized fusion weight of the candidate first-level service dimension is obtained.

[0013] Specifically, in step six, multiple topic screening strategies are selected to screen the final fusion weight of the topic. By traversing and comparing various topic screening strategies, the topic set with the highest fusion weight is selected as the final core three-level indicator item set. The final service quality evaluation system scale is constructed based on the selected three-level indicator items combined with the corresponding second-level topics and first-level service dimensions.

[0014] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the instructions are executed by the processor, the processor implements the method for dynamically constructing a service quality evaluation system.

[0015] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for dynamically constructing a service quality evaluation system is implemented.

[0016] The present invention adopts a service quality evaluation method that integrates topic modeling, large language model understanding, multi-channel heterogeneous hypergraph sentiment modeling and dynamic weight update mechanism. It automatically identifies the core service topics in user evaluation texts, introduces sliding window co-occurrence and external knowledge to construct a multi-source heterogeneous hypergraph structure, and simultaneously outputs sentiment label predictions and scoring results based on a multi-layer hypergraph neural network, thereby realizing multi-dimensional modeling of user perceived quality. Dynamic weight updates are performed based on multiple factors such as text length, frequency and sentiment consistency, and a weighted fusion strategy is used to dynamically sort and screen service quality indicators, ultimately constructing a service quality evaluation index system with semantic representativeness, sentiment recognition and hierarchical adaptability. This method not only achieves significant improvements in automation level and explanatory power, but also provides a scientific basis and practical support for the service platform to achieve intelligent quality monitoring and dynamic optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the overall structure of the present invention; Figure 2 It is a structural diagram of the multi-channel heterogeneous hypergraph modeling of the present invention; Figure 3 This is a data distribution chart of agricultural supplies and agricultural tools categories on the e-commerce platform; Figure 4 It is the weight ranking diagram of each theme indicator; Figure 5 It is a radar chart showing the distribution of the integrated weights of the five service dimensions in the final SERVQUAL service quality evaluation system scale. DETAILED DESCRIPTION

[0018] The present invention will be further described in detail below with reference to the accompanying drawings and examples.

[0019] Reference Figure 1 A dynamic construction method for a service quality evaluation system integrating a large language model is described in the following steps: Step 1: Use crawler technology to crawl original user evaluation texts on agricultural supplies and agricultural tools from e-commerce platforms. Collect them in batches by product. Each original user evaluation text contains the user's free text evaluation and the corresponding user satisfaction score (1-5 points). Clean and pre-process the original user evaluation texts to obtain the user evaluation text dataset. The specific distribution is as follows: Figure 3As shown, user reviews for agricultural implements include: seed drills, 7,132 reviews; sickles, 6,540 reviews; fertilizer spreaders, 8,276 reviews; and sprayers, 13,319 reviews. User reviews for agricultural supplies include: feed, 11,230 reviews; aquatic products, 9,756 reviews; fertilizers, 8,143 reviews; seeds, 10,278 reviews; and pesticides, 9,158 reviews.

[0020] Step 2: The user evaluation text dataset D={d1,d2,..,d n}, where d1, d2, .., d n Represent the 1st, 2nd, ..., nth user evaluation text respectively, and transform each user evaluation text d into i Convert to a fixed-dimensional semantic embedding vector , i is the user evaluation text index, Represents a set of real numbers, forming the overall text embedding vector matrix , where e1, e2, …, e n Represent the embedding vectors of the 1st, 2nd, ..., nth user evaluation text respectively, n represents the total number of user evaluation texts, and r represents the vector dimension of each user evaluation text after embedding. Subsequently, the unified manifold approximation projection algorithm (UMAP) is used to reduce the dimension of the overall text embedding vector matrix E, retaining the local density and global structure of the semantic space, and obtaining the feature matrix after dimensionality reduction. , where m represents the target dimension. On this basis, the hierarchical density clustering algorithm (HDBSCAN) is used to perform topic clustering on the feature matrix U after dimensionality reduction, and the topic set T = {t1, t2, ..., t K}, where K represents the number of topics finally obtained, t1, t2, ..., t K Represent the 1st, 2nd,…,Kth topics respectively.

[0021] To extract keywords for each topic, the category conditional inverse document frequency weight model (c-TF-IDF) is constructed according to formula (1): (1); where f k,j Indicates the frequency of occurrence of the jth word in the kth topic, df j Represents the number of topics containing word j. The higher the word weight obtained, the more representative the word is of the topic. Then, based on the text distribution density and clustering confidence of each topic, a representative document set D is selected. k ={d k,1 , d k,2 ,…, d k,l}, as the semantic core sample of the kth topic, where D krepresents the representative document set under the kth topic, l represents the number of representative documents selected in each topic, d k,l = represents the user evaluation text of the lth selected as the representative document in the kth topic. Finally, 95 topics were obtained, of which the top five topics, their keywords and weight scores are shown in Table 1.

[0022] Table 1

[0023] The keywords of the topic are spliced ​​with the corresponding representative documents, and combined with their semantic relevance to the five service dimensions of the SERVQUAL model (tangibility, reliability, assurance, responsiveness, and empathy), a structured prompt input that conforms to the semantic topic is generated. A prompt word is constructed for each topic. The specific construction process is shown in formula (2).

[0024] (2); Among them, P k represents the prompt word of the kth topic, W k represents the keyword set consisting of the kth topic, Denotes the representative document set D composed of the kth topic k A collection of user satisfaction ratings, Respectively represent the user satisfaction scores of the user evaluation texts of the 1st, 2nd, …, lth selected as representative documents in the kth topic. The prompt words are passed into the large language model to generate candidate first-level service dimensions divided by the five service dimensions of the SERVQUAL model. , where S∈{tangibility, reliability, assurance, responsiveness, empathy}, S k Represents the candidate first-level service dimensions of the k-th topic, as well as the candidate second-level topics and candidate third-level indicators A list of candidate service quality indicators composed of Denotes the kth topic t k The secondary theme, Represents the kth topic t k The 1st, 2nd, ..., Gth secondary topic names, Denotes the kth topic t k The third-level indicator items, Represents the kth topic t k 1, 2, ..., The obtained list of candidate indicators of the three-level hierarchical structure service quality is shown in Table 2, and the results obtained from the first five topics are specifically selected.

[0025] Table 2

[0026] Step 3: Use the representative documents as the test set and the remaining user evaluation text dataset as the training set. Each user evaluation text in the training set and the test set is divided into word sequences to form the semantic unit basis. The word sequence is represented as , where the word sequence O (i) is the set of all words in the i-th user evaluation text, represents the 1st, 2nd, ..., nth item of the i-th user evaluation text i words, n i Represents the number of words in the i-th user review text. All words constitute the vocabulary table V, and then the continuous bag of words model (Word2Vec) based on word embedding is used to map each word into a q-dimensional embedding vector to obtain the word embedding matrix , where |V| represents the total number of words in the vocabulary, x1,x2,…,x |V| Represent the word vector representation of the 1st, 2nd,…,|V|th word in the vocabulary respectively.

[0027] For each user evaluation text, define the window size as , for word sequence O (i) Perform sliding window division to obtain multiple sliding window word sets, as shown in formula (3) to formula (5): (3); (4); (5); in represents the sliding window word set of the i-th user evaluation text, They represent the 1st, 2nd, ..., Mth items in the i-th user evaluation text. (i) Sliding window word set, M (i) represents the total number of sliding windows generated in the i-th user evaluation text, represents the first sliding window set formed starting from the jth word in the i-th user evaluation text, Respectively represent the i-th user evaluation text Position word.

[0028] Based on the sliding window word set of all samples, any two words are defined as w b ,w p , its pointwise mutual information (PMI) value reflects the closeness of their local co-occurrence in the corpus. The calculation process is shown in formula (6): (6); (7); in, For word pairs (w b ,w p ), Represents word pair (w b ,w p ) appears in all sliding windows, and Respectively represent w b ,w p Total number of occurrences, T win Indicates the total number of sliding windows generated in the entire corpus, and n represents the total number of user evaluation texts. In order to suppress the introduction of low-frequency redundant edges, the point mutual information threshold is set >0, the set of candidate hyperedges that meet the point mutual information threshold condition in the i-th user evaluation text is defined as shown in formula (8): (8); in, Represents the point mutual information hyperedge set constructed by the i-th user evaluation text, that is, the set of all word windows that meet the high co-occurrence strength requirements as the hyperedge in the hypergraph, reflecting the co-occurrence structural relationship between words.

[0029] In order to further model high-order semantic relationships, an external knowledge graph (ConceptNet) is introduced as the second channel of the heterogeneous hypergraph. , the semantic association set defined by all connected words that have semantic edge relations with it in ConceptNet retrieval is shown in formula (9): (9); in Indicates the first item in the i-th user evaluation text words The set of semantically adjacent words associated in the external knowledge graph, Represents There is a tail word connected by a semantic edge, Indicates that there is a link in the external knowledge graph from point to The semantic edge of Indicates the semantic type of this edge, and ConceptNet is the external knowledge graph used. Construct a hyperedge set consisting of it and all associated semantic adjacent words, as shown in formula (10): (10); in represents the hyperedge set constructed by the semantic relationship of the i-th user evaluation text obtained through the external knowledge graph, Indicates that the current word and all its semantically adjacent words together form a hyperedge. Then, the sliding window co-occurrence mechanism is unified with the hyperedge sets of the two semantic channels of the external knowledge graph to form a multi-channel heterogeneous hypergraph, and the final hyperedge set of the multi-channel heterogeneous hypergraph is defined as shown in formula (11): (11); in represents the hyperedge set of the multi-channel heterogeneous hypergraph of the i-th user evaluation text, C represents the total number of channels of the multi-channel heterogeneous hypergraph, represents the set of hyperedges from the cth channel in the i-th user evaluation text. Then, the triple of the i-th user evaluation text for the multi-channel heterogeneous hypergraph can be constructed, as shown in formula (12): (12); (13); in, Represents the multi-channel heterogeneous hypergraph of the i-th user evaluation text, i=[1,2,…,n], represents the node feature submatrix obtained by embedding the words used in the i-th user evaluation text in the word embedding matrix X, |V (i) | represents the number of nodes in the i-th user evaluation text, u is the embedding dimension, S (i) Represents the graph structure of the i-th user evaluation text, represents the hyperedge connection matrix of the i-th user evaluation text, represents the user satisfaction score of the i-th user evaluation text, which ultimately constitutes a multi-channel heterogeneous hypergraph set constructed from all user evaluation texts .

[0030] Step 4: Establish a multi-channel heterogeneous hypergraph neural network model that integrates multi-layer hypergraph convolution and multi-scale feature extraction capabilities, such as Figure 2 As shown. First, the graph structure Decouple according to different semantic channels to form H channel graph structure information ,in Represents the graph structure information of the i-th graph structure under the h-th channel. The feature embedding of each user evaluation text is constructed by formula (14): (14); in Represents the feature embedding of the i-th user evaluation text under the h-th channel, the node feature submatrix Shared among all channels. , we use multi-layer hypergraph convolution as shown in formula (15) to embed node features and aggregate the structural information of adjacent nodes of different scales layer by layer to obtain the local-global semantically aware embedding matrix.

[0031] (15); in represents the node embedding matrix of the i-th user evaluation text in the l+1-th hypergraph convolutional layer of the h-th channel, represents the node embedding matrix of the i-th user evaluation text in the h-th channel and the l-th layer of the hypergraph convolutional layer, represents the node degree matrix, represents the square root inverse of the node degree matrix, represents the vertex-hyperedge incidence matrix of the h-th channel, express The transpose of represents the hyperedge weight matrix of the h-th channel, represents the hyperedge degree matrix, represents the inverse of the hyperedge degree matrix, represents the feature mapping weight from the lth layer of hypergraph convolution layer to the l+1th layer of hypergraph convolution layer of the hth channel, represents a non-linear activation function.

[0032] The i-th graph structure After passing L layers through multi-layer hypergraph convolution under the h-th channel, the final node structure embedding is , embedding the node structure of all channels After dimension alignment, channel attention is introduced to generate fused node embedding, as shown in formula (16)-formula (17).

[0033] (16); (17); in represents the fused node embedding of the i-th graph structure, represents the attention weight of the h-th channel, represents the attention score of each channel, for The transpose of is the i-th graph structure in the The final node structure embedding obtained under the channel is represents the learnable parameter matrix in channel attention, Indicates the average pooling of node features. Embed the fusion node into Z (i)Input the convolutional neural network, extract local features through local convolution, and perform global pooling to obtain a unified graph-level structure representation f (i) Then, according to formula (18) and formula (19), joint learning is performed to predict the sentiment label and sentiment score.

[0034] (18); (19); in represents the predicted sentiment label of the i-th user evaluation text, represents the weight matrix of the classification task, represents the classification bias vector, represents the predicted sentiment score of the i-th user review text, represents the authority matrix of the regression task, The multi-channel heterogeneous hypergraph neural network model constructed in this paper achieved an accuracy of 78.11% on a representative document test set, with weighted precision, recall, and F1 values ​​of 94.04%, 78.11%, and 84.24%, respectively.

[0035] Step 5: Based on the prediction results of sentiment labels and sentiment scores, the present invention combines user satisfaction scores, user evaluation text content structure and sentiment prediction information to construct a multi-category input factor space at the topic level, specifically including eight types of indicator factors as shown in Table 3.

[0036] Table 3

[0037] For each topic k ∈T corresponding candidate third-level indicator items , perform minimum-maximum normalization on all factor values ​​to obtain the normalized factor value, as shown in formula (20): (20); (twenty one); in Denotes the kth topic t k The vth normalization factor value of Denotes the kth topic t k The vth factor value of represents the numerical set of the vth factor on K topics, Represent the v-th factor value on the 1st, 2nd, ..., Kth topic respectively. Then a set of preset weight parameters are used , linearly weight each normalization factor value, the specific process is shown in formula (22): (twenty two); (twenty three); in Denotes the kth topic t k The initial fusion weight, m ​​represents the total number of factors involved in weighted fusion, Represents the weight coefficient of the vth factor. Then the original fusion weight Normalization is performed and converted into the final fusion weight using the Softmax function, as shown in formula (24).

[0038] (twenty four); in Denotes the kth topic t k The final fusion weight of , e represents the base of the natural logarithm.

[0039] The final fusion weight of each topic Passed to its corresponding candidate third-level indicator item , realize weight downward binding, and then adopt the bottom-up weighted normalization strategy to complete weight aggregation, and obtain the normalized secondary topic fusion weight through formula (25)-formula (26).

[0040] (25); (26); in Denotes the kth topic t k The fusion weight of the g-th candidate secondary topic, represents the set of candidate third-level indicator items for the g-th candidate second-level topic, Indicates the A set of candidate third-level indicator items for candidate second-level topics, Denotes the kth topic t k The second-level candidate topic candidate third-level indicator items, G represents the set of all candidate second-level topics, Denotes the kth topic t k Next The fusion weight of candidate secondary topics, Denotes the kth topic t k The normalized fusion weight of the g-th candidate secondary topic is obtained. At the same time, the normalized fusion weight of the candidate primary service dimension is obtained according to formula (27).

[0041] (27); in Denotes the kth topic t k In the candidate level service dimension S k The normalized fusion weight on represents its semantic importance contribution, S k represents the kth candidate first-level service dimension, G k Indicates that it belongs to S k The candidate secondary topic set, Indicates the Candidate first-level service dimensions, Indicates belonging A set of candidate secondary topics under a dimension.

[0042] Step 6: Candidate third-level indicators For comprehensive sorting and dynamic screening, the present invention designs three topic screening strategies to eliminate redundant topics with low fusion weight contribution from candidates, as shown in Table 4: Table 4

[0043] The three topic screening strategies shown in Table 4 screen the final fusion weights of the topics according to formulas (28) to (30).

[0044] (28); (29); (30); in Indicates that the first The filtered topic set consists of topics. Represents the topic set retained through dynamic cumulative screening, Represents the topic set with a fusion weight higher than the overall average. The present invention traverses and compares the above three topic screening strategies and selects the topic set with the highest fusion weight as the final core three-level indicator item set, as shown in formula (31).

[0045] (31); in represents the topic set with the highest weight fusion value, The core three-level indicator set finally retained is used as the core content of constructing the service quality evaluation system scale. The final fusion weight is as follows Figure 4As shown. And according to the selected three-level indicators, combined with the corresponding second-level themes and first-level service dimensions, the final service quality evaluation system scale is formed. The final system results are shown in Table 5, which shows the first 10 final service quality evaluation system indicators.

[0046] Table 5

[0047] Figure 5 This study demonstrates the normalized weight distribution of five primary service dimensions (tangibility, assurance, reliability, responsiveness, and empathy) in a service quality evaluation system constructed based on a user review dataset of agricultural supplies and implements on an e-commerce platform. This system integrates semantic parsing with a large language model, multi-channel heterogeneous hypergraph modeling, and a multi-factor weight update strategy. The results show that the assurance dimension has the highest weight of 0.26, indicating that user reviews focus on service aspects such as reliable product delivery, quality assurance, and after-sales commitment. The tangibility and responsiveness dimensions have weights of 0.22 and 0.17, respectively, reflecting a high level of user interest in perceptible service attributes such as product appearance, packaging experience, and customer service response speed. The reliability and empathy dimensions have relatively low weights of 0.19 and 0.15, respectively, indicating that some users prioritize service stability and humanistic care.

[0048] The above results fully verify the effectiveness of the SERVQUAL service quality evaluation system proposed in the present invention, which integrates the large language model understanding ability, graph structure emotion perception modeling ability and multi-factor dynamic control mechanism. It effectively breaks through the limitations of traditional service quality evaluation methods based on questionnaires and static dimension designs, and shows good transferability and practicality in e-commerce scenarios. By introducing a structured evaluation indicator scale with strong interpretability, wide semantic coverage and dynamically adjustable indicators, it provides reliable technical support for intelligent service quality evaluation, operation and maintenance optimization and strategic decision-making.

[0049] Another embodiment of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores computer-readable instructions, and wherein when the instructions are executed by the processor, the processor implements the method for dynamically constructing a service quality evaluation system.

[0050] Another embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method for dynamically constructing a service quality evaluation system.

[0051] The above description merely represents preferred embodiments of the present invention and is not intended to limit the present invention in any other manner. Any person skilled in the art may utilize the above disclosure to modify or modify the present invention into equivalent embodiments. However, any simple modifications, equivalent variations, and modifications to the above embodiments that do not depart from the technical content of the present invention and are based on the technical essence of the present invention remain within the scope of protection of the present invention.

Claims

1. A method for dynamically constructing a service quality evaluation system integrating a large language model, characterized by the following steps: as follows: Step 1: Obtain a user review text dataset for product sales; Step 2: Extract topics, keywords, and representative documents from the user review text dataset, construct prompt words, and input them into the large language model to generate a list of candidate first-level service dimensions for the five service dimensions of the SERVQUAL model, as well as a list of candidate service quality indicators consisting of candidate second-level topics and candidate third-level indicators. Step 3: Use representative documents as the test set and the remaining user review text dataset as the training set to construct a multi-channel heterogeneous hypergraph collection that integrates sliding window co-occurrence and external knowledge graph; Step 4: The multi-channel heterogeneous hypergraph set is input into the multi-channel heterogeneous hypergraph neural network model to extract the graph structure information and perform joint prediction of the sentiment label and sentiment score of the user evaluation text; Step 5: Based on the predicted results of sentiment labels and sentiment scores, combined with user satisfaction scores, user review text content structure, and sentiment prediction information, a multi-class input factor space at the topic level is constructed. The final fusion weight of each topic, the fusion weight of the second-level topic, and the normalized fusion weight of the candidate first-level service dimension are calculated. Step 6: Comprehensively rank and dynamically screen the candidate third-level indicator items, and collect the core third-level indicator items to form a multi-level service quality evaluation system indicator scale.

2. The method for dynamically constructing a service quality evaluation system according to claim 1, wherein: In step 2, the process of extracting topics, keywords, and representative documents is as follows: each user review text is converted into a fixed-dimensional semantic embedding vector using a sentence-level multilingual pre-trained model, forming an overall text embedding vector matrix. Subsequently, the unified manifold approximation projection algorithm is used to reduce the dimensionality of the overall text embedding vector matrix, preserving the local density and global structure of the semantic space to obtain a reduced-dimensional feature matrix. The reduced-dimensional feature matrix is ​​then clustered using a hierarchical density clustering algorithm to output the topic set to which each user review text belongs. Keywords for each topic through the category-conditional inverse document frequency weight model; A representative document collection is selected based on the text distribution density and clustering confidence of each topic.

3. The method for dynamically constructing a service quality evaluation system according to claim 2, wherein: In step 2, the keywords of the topic are spliced ​​with the corresponding representative documents, and combined with their semantic relevance to the five service dimensions of the SERVQUAL model to generate structured prompt input that conforms to the semantic topic. Prompt words are constructed for each topic, and the prompt words include a keyword set, a representative document set, and a user satisfaction score set.

4. The method for dynamically constructing a service quality evaluation system according to claim 1, wherein: The specific process of step three is as follows: Each user review text in the training set and test set is divided into word sequences to form the basis of semantic units. All words constitute the vocabulary. Then, a continuous bag-of-words model based on word embedding is used to map each word into a q-dimensional embedding vector to obtain a word embedding matrix. For each user review text, the word sequence is divided into sliding windows to obtain multiple sliding window word sets; Define any two words based on the sliding window word set of all samples and calculate their point mutual information value, set the point mutual information threshold >0, select the candidate hyperedge set that meets the point mutual information threshold condition in the user evaluation text; The external knowledge graph ConceptNet is introduced as the second channel of the heterogeneous hypergraph. For each word, a semantic association set is defined by all connected words that have semantic edge relationships with it in ConceptNet retrieval. Then, a hyperedge set consisting of all the associated semantic adjacent words is constructed for each word. Then, the hyperedge sets of the two semantic channels of sliding window co-occurrence and external knowledge graph are unified and fused to form a hyperedge set of multi-channel heterogeneous hypergraph, and the triples of each user evaluation text for the multi-channel heterogeneous hypergraph are constructed, finally forming a multi-channel heterogeneous hypergraph set constructed by all user evaluation texts.

5. The method for dynamically constructing a service quality evaluation system according to claim 1, wherein: In step 4, the processing of the multi-channel heterogeneous hypergraph neural network model includes: First, the graph structure of user review text is decoupled according to different semantic channels to form H channel graph structure information. The feature embedding of each user review text in each channel is constructed, and the node feature submatrix is ​​shared by all channels. For each feature embedding, multi-layer hypergraph convolution is used to embed the node features and aggregate the structural information of adjacent nodes at different scales layer by layer to obtain the local-global semantically aware embedding matrix. The node structure embedding finally obtained after passing L layers through multi-layer hypergraph convolution is dimensionally aligned for the node structure embeddings of all channels. Channel attention is introduced to generate fused node embedding, which is then input into the convolutional neural network. Local features are extracted through local convolution, and global pooling is performed to obtain a unified graph-level structure representation. Joint learning is then performed to achieve the prediction results of sentiment labels and sentiment scores.

6. The method for dynamically constructing a service quality evaluation system according to claim 1, wherein: The multi-category input factor space described in step five includes word frequency factor, sentiment mean factor, document coverage factor, satisfaction driving factor, consistency factor, information density factor, text length factor and sentiment entropy factor, a total of eight types of indicator factors.

7. The method for dynamically constructing a service quality evaluation system according to claim 1, wherein: In step 5, for each candidate third-level indicator item corresponding to each topic, all factor values ​​are normalized using the minimum-maximum method to obtain the normalized factor value. Then, a set of preset weight parameters are used to linearly weight each normalized factor value to obtain the initial fusion weight of each topic. The original fusion weight is then normalized and converted into the final fusion weight using the Softmax function. The final fusion weight of each topic is transferred to its corresponding candidate third-level indicator item to achieve downward binding of the weight. Then, a bottom-up weighted normalization strategy is used to complete the weight aggregation and obtain the normalized second-level topic fusion weight; at the same time, the normalized fusion weight of the candidate first-level service dimension is obtained.

8. The method for dynamically constructing a service quality evaluation system according to claim 1, wherein: In step six, multiple topic screening strategies are selected to screen the final fusion weight of the topic. By traversing and comparing various topic screening strategies, the topic set with the highest fusion weight is selected as the final core three-level indicator item set. The final service quality evaluation system scale is constructed based on the selected three-level indicator items combined with the corresponding second-level topics and first-level service dimensions.

9. An electronic device comprising a memory and a processor, wherein the memory stores computer-readable instructions, wherein: When the instruction is executed by the processor, the processor implements the method for dynamically constructing a service quality evaluation system according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for dynamically constructing a service quality evaluation system according to any one of claims 1 to 8 is implemented.

Images