Method and device for generating a multi-dimensional service evaluation word cloud chart based on customer perception

Abstract

Embodiments of the present disclosure disclose a method and device for generating a multi-dimensional service evaluation word cloud chart based on customer perception. A specific embodiment of the method comprises: obtaining customer review texts from a service online review platform to obtain a set of customer review texts; performing text compensation on each customer review text in the set of customer review texts to generate a set of compensated review texts; performing text topic word extraction on each compensated review text in the set of compensated review texts to generate a set of multi-dimensional topic word groups of review texts; determining a preset customer topic perception model corresponding to each multi-dimensional topic word group of review texts in the set of multi-dimensional topic word groups of review texts as a current topic perception model; performing semantic recognition processing on each multi-dimensional topic word in the set of multi-dimensional topic word groups of review texts to generate a sequence of recognized topic word information; and constructing a multi-dimensional word cloud chart according to the sequence of recognized topic word information. The embodiment can reduce waste of computing resources.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, and specifically to a method and apparatus for generating word cloud charts for multi-dimensional service evaluation based on customer perception. Background Technology

[0002] Improving merchant service quality to better adapt to current customer preferences and identifying service quality shortcomings affecting customer satisfaction can promote the circulation of offline goods. Currently, the common method for generating word cloud charts is to collect all words from user reviews and then create a word cloud chart using all those words.

[0003] However, in practice, it has been found that when using the above method to generate word cloud charts, the following technical problems often occur:

[0004] User comments often contain simple expressions and missing sentences, resulting in incomplete vocabulary extraction and difficulty in distinguishing the semantics of words. Consequently, the generated word cloud chart data is incomplete and needs to be regenerated, leading to a waste of computing resources.

[0005] The information disclosed in this background section is only intended to enhance the understanding of the background of the inventive concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this disclosure propose a method and apparatus for generating multi-dimensional service evaluation word cloud charts based on customer perception, in order to solve one or more of the technical problems mentioned in the background section above.

[0008] In a first aspect, some embodiments of this disclosure provide a method for generating a multi-dimensional service evaluation word cloud chart based on customer perception. The method includes: obtaining customer review texts from an online service review platform to obtain a customer review text set; performing text compensation on each customer review text in the aforementioned customer review text set to generate a compensated review text set; extracting text keywords from each compensated review text in the aforementioned compensated review text set to generate a multi-dimensional keyword group set for review texts; determining a preset customer topic perception model corresponding to each multi-dimensional keyword group in the aforementioned multi-dimensional keyword group set for review texts as the current topic perception model; performing semantic recognition processing on each multi-dimensional keyword in the aforementioned multi-dimensional keyword group set for review texts based on the current topic perception model and the aforementioned customer review text set to generate a sequence of recognized keyword information; and constructing a multi-dimensional word cloud chart based on the sequence of recognized keyword information.

[0009] Secondly, some embodiments of this disclosure provide a multi-dimensional service evaluation word cloud chart generation device based on customer perception. The device includes: an acquisition unit configured to acquire customer review text from an online service review platform to obtain a customer review text set; a text compensation unit configured to perform text compensation on each customer review text in the aforementioned customer review text set to generate a compensated review text set; a topic word extraction unit configured to extract text topic words from each compensated review text in the aforementioned compensated review text set to generate a multi-dimensional topic word group set for review texts; a model determination unit configured to determine a preset customer topic perception model corresponding to each multi-dimensional topic word group in the aforementioned multi-dimensional topic word group set for review texts as the current topic perception model; a semantic recognition processing unit configured to perform semantic recognition processing on each multi-dimensional topic word in the aforementioned multi-dimensional topic word group set for review texts based on the current topic perception model and the aforementioned customer review text set to generate a sequence of recognized topic word information; and a word cloud chart construction unit configured to construct a multi-dimensional word cloud chart based on the sequence of recognized topic word information.

[0010] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0011] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0012] The above embodiments of this disclosure have the following beneficial effects: the method for generating multi-dimensional service evaluation word cloud charts based on customer perception according to some embodiments of this disclosure can reduce the waste of computing resources. Specifically, the reason for the waste of computing resources is that user reviews often contain simple expressions and missing sentences, resulting in incomplete extraction of all words and difficulty in distinguishing the semantics of words. Consequently, the generated word cloud chart data is incomplete and needs to be regenerated. Based on this, the method for generating multi-dimensional service evaluation word cloud charts based on customer perception according to some embodiments of this disclosure first obtains customer review text from an online service review platform to obtain a customer review text set. Then, considering that the original review text usually contains a lot of non-standard content, such as spelling errors, special symbols, emoticons, etc., which may interfere with text analysis and reduce model performance, and that noisy data (such as HTML tags, advertisements, redundant characters, etc.) will affect word segmentation, word frequency statistics, etc., and that a large number of meaningless characters or words will increase the time cost of data processing, it is necessary to clean the original reviews. Therefore, text compensation is performed on each customer review text in the above customer review text set to generate a compensated review text set. Thus, text compensation improves data quality, reduces noise, and enhances data accuracy. Next, text keyword extraction is performed on each compensated comment text in the compensated comment text set to generate a multi-dimensional keyword set. Here, keyword extraction helps identify keywords and other vocabulary in the comments, facilitating semantic differentiation and text feature extraction. Furthermore, a pre-defined customer topic perception model corresponding to each multi-dimensional keyword set in the multi-dimensional keyword set is determined as the current topic perception model. Then, based on the current topic perception model and the customer comment text set, semantic recognition processing is performed on each multi-dimensional keyword in the multi-dimensional keyword set to generate a sequence of identified keywords. Here, the pre-defined customer topic perception model helps determine the sentiment information of keywords in the comment text based on the multi-dimensional keyword set. This allows for perception of keywords in customer comment texts from the customer's perspective, improving the accuracy of keyword extraction. This facilitates the construction of a multi-dimensional word cloud chart based on the identified keyword information sequence. This can improve the accuracy of the generated multidimensional word cloud charts. Furthermore, it can reduce the computational resources wasted on regeneration. Attached Figure Description

[0013] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0014] Figure 1 This is a flowchart of some embodiments of the customer perception-based multidimensional service evaluation word cloud chart generation method according to the present disclosure;

[0015] Figure 2 This is a result display diagram of model perplexity and model consistency data according to some embodiments of the customer perception-based multidimensional service evaluation word cloud chart generation method disclosed herein;

[0016] Figure 3 This is a schematic diagram of the model structure of a review quality identification model according to some embodiments of the customer perception-based multidimensional service evaluation word cloud chart generation method disclosed herein;

[0017] Figure 4 This is a schematic diagram of a review performance analysis chart based on some embodiments of the customer perception-based multidimensional service evaluation word cloud chart generation method disclosed herein;

[0018] Figure 5 These are schematic diagrams of positive and negative sentiment word clouds based on some embodiments of the customer perception-based multidimensional service evaluation word cloud generation method disclosed herein;

[0019] Figure 6 This is a framework diagram of a customer satisfaction regression model based on some embodiments of the customer perception-based multidimensional service evaluation word cloud chart generation method disclosed herein;

[0020] Figure 7 This is a schematic diagram of the structure of some embodiments of the customer perception-based multidimensional service evaluation word cloud chart generation device according to the present disclosure;

[0021] Figure 8 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0022] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0023] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0024] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0025] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0026] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0027] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] Figure 1 A flow 100 of some embodiments of the customer-perception-based multidimensional service evaluation word cloud chart generation method according to this disclosure is shown. This customer-perception-based multidimensional service evaluation word cloud chart generation method includes the following steps:

[0029] Step 101: Obtain customer review texts from the online service review platform to obtain a set of customer review texts.

[0030] In some embodiments, the entity executing the customer-perception-based multidimensional service evaluation word cloud chart generation method can obtain customer review texts from an online service review platform via wired or wireless means, thus obtaining a set of customer review texts. The online service review platform can be a public platform for reviewing businesses.

[0031] It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other currently known or future wireless connection methods.

[0032] Step 102: Perform text compensation on each customer comment text in the customer comment text set to generate a compensated comment text set.

[0033] In some embodiments, the aforementioned executing entity may perform text compensation on each customer comment text in the aforementioned customer comment text set to generate a compensated comment text set.

[0034] In some optional implementations of certain embodiments, the execution entity performs text compensation on each customer comment text in the aforementioned customer comment text set to generate a compensated comment text set, which may include the following steps:

[0035] The first step is to perform character recognition on each customer review text in the aforementioned customer review text set to remove meaningless characters, resulting in a clean review text set. Specifically, character identifiers in the customer review text can be indexed for removal, yielding the clean review text.

[0036] The second step involves using a pre-established stop word dictionary to perform stop word analysis on each of the removed comment texts in the aforementioned removed comment text set, in order to generate a compensated comment text set. The stop word analysis is used to determine the stop words and their part-of-speech meanings in each removed comment text.

[0037] In practice, stop word dictionaries contain a series of words and phrases related to specific user needs. During data cleaning, referring to the user dictionary allows for more accurate identification and processing of key information in the text. In word segmentation, the user dictionary can provide the necessary lexical information, thereby optimizing segmentation results. For example, for certain compound words or technical terms, the user dictionary can break them down into smaller units for a more accurate understanding of the text's meaning. The user dictionary can also provide information such as the part of speech and meaning of words, further contributing to improved accuracy and efficiency in word segmentation.

[0038] Step 103: Extract text keywords from each compensated comment text in the compensated comment text set to generate a multidimensional keyword set for the comment text.

[0039] In some embodiments, the aforementioned executing entity may extract text keywords from each compensated comment text in the compensated comment text set to generate a multidimensional keyword set for the comment text.

[0040] In some optional implementations of certain embodiments, the execution entity extracts text keywords from each compensated comment text in the compensated comment text set to generate a multidimensional keyword set for the comment text, which may include the following steps:

[0041] The first step is to extract keywords from each compensated comment text in the aforementioned compensated comment text set to generate the current comment text's keyword phrases, thus obtaining the current comment text's keyword phrase set. Specifically, a term frequency-inverse term frequency algorithm can be used to extract keywords from each compensated comment text in the aforementioned compensated comment text set to generate the current comment text's keyword phrases, thus obtaining the current comment text's keyword phrase set.

[0042] In practice, considering that the bag-of-words model only records the frequency of word occurrences and does not consider word weights or word positional relationships in the text, term frequency (TF) and inverse document frequency (IDF) algorithms are used for word segmentation.

[0043] The second step is to perform deduplication on the current comment text topic term set to generate a deduplicated text topic term set.

[0044] The third step involves dividing the deduplicated text keywords in the aforementioned deduplicated text keyword set into themes to generate a multidimensional keyword set for the comment text. The theme division can involve assigning each deduplicated text keyword to any number of different themes (e.g., 2-24 themes), with no limit on the number of deduplicated text keywords under each theme. Each multidimensional keyword set for the comment text can correspond to one division method. Here, the multidimensional keyword sets for the comment text under different themes can be repeatedly divided.

[0045] As an example, one possible approach is to divide the topic into 14 subtopics. The result of this division can be shown in the table below:

[0046] Theme 1 Direct access, nice environment, many restaurants and shopping options, convenient transportation via subway. Theme Two The center provides good products and experiences to meet your needs for shopping. Theme 3 Food, discounts, clothes, brands, promotions, great deals, good, lots of discounts Theme 4 Eating, Kapibala, Joy City, Check-in, Cute, Like, Many, Queuing, Flash Mob, Event Theme 5 Experience, restrooms, parking lot, queues, special, clothing, pop-up, events, each, cute Theme Six The food was excellent, I really liked it; it was clean, the taste was great, the service was excellent, and the ambiance was nice. Theme Seven Parking, shops, subway, goods, snacks, check-in spots, convenience, dining, activities, many Theme 8 Environment, air conditioning, many, check-in, shops, frequently, pop-up, many, like, activities Theme Nine Check-in, suitable, things, good place, nice, eating, like, big, weekend, many Theme 10 Queuing, clear, good, found, environment, brand, elevator, special, many, things Theme Eleven Frequently, like, brands, plazas, Tianshan, many, cities, Youke, Parkson, events Theme Twelve Supermarket, I like it, it's good, there are promotions, department stores, Taikoo, Jiuguang, many brands Theme Thirteen Attitude, dining, security, service desk, trash, parking fee, parking lot, parking, points Theme Fourteen Also, shops, nearby, points, Joy City, parking fees, clothes, activities, many, restaurants

[0047] In practice, when using technical solutions to address the problems mentioned in the background technology, the following technical problem often arises: even if all words in the text can be found, simply calculating word importance based on word frequency is insufficient to reflect the emotional value and customer satisfaction of the customer review text. This leads to keyword bias, resulting in significant deviations in the final generated multidimensional word cloud chart. Consequently, it also requires regeneration, consuming more computational resources. To address this second technical problem, the inventors decided to adopt the following solution.

[0048] Step 104: Determine the preset customer topic perception model corresponding to each multi-dimensional topic phrase in the multi-dimensional topic phrase set of comment text, and use it as the current topic perception model.

[0049] In some embodiments, the execution entity may determine a preset customer topic perception model corresponding to each multi-dimensional topic phrase in the multi-dimensional topic phrase set of the comment text as the current topic perception model.

[0050] In some optional implementations of certain embodiments, the execution entity determines a preset customer topic perception model corresponding to each multi-dimensional topic phrase in the multi-dimensional topic phrase set of the comment text as the current topic perception model, which may include the following steps:

[0051] The first step is to obtain the initial topic-aware model from the database.

[0052] As an example, the initial topic-aware model can be the LDA (Latent Dirichlet Allocation) unsupervised topic modeling algorithm.

[0053] The second step involves iterating the initial topic-aware model using the various multi-dimensional topic terms from the aforementioned set of multi-dimensional topic phrases in the comment text, resulting in a set of topic quantity-aware models. Each topic quantity-aware model corresponds to one multi-dimensional topic phrase from the aforementioned set of multi-dimensional topic phrases in the comment text. Each multi-dimensional topic phrase corresponds to a different number of topics. Each multi-dimensional topic phrase can be divided into multiple sub-groups based on the number of topics. Here, each multi-dimensional topic phrase from the aforementioned set of multi-dimensional topic phrases in the comment text can be input into the initial topic-aware model for iteration to train and obtain the topic quantity-aware model.

[0054] The third step is to determine the model perplexity and model consistency data for each topic quantity perception model in the aforementioned topic quantity perception model set. This can be achieved by using a pre-set test set, which can be input into each topic quantity perception model to test its model perplexity and model consistency data.

[0055] As an example, model perplexity and model consistency data can be as follows: Figure 2 As shown. Figure 2 The left-hand plot shows the model perplexity. The horizontal axis represents the number of topics, and the vertical axis represents the perplexity. The right-hand plot shows the model consistency data. The horizontal axis represents the number of topics, and the vertical axis represents the consistency value. Here, each point in the plot corresponds to the model perplexity and model consistency value of a topic-aware model.

[0056] The fourth step is to determine the topic-aware model corresponding to the model perplexity and model consistency data that meets the preset topic segmentation conditions as the current topic-aware model. Specifically, the multi-dimensional topic phrases of the comment text corresponding to the above-mentioned current topic-aware model are the current comment text multi-dimensional topic phrases. The preset topic segmentation conditions can be the highest comprehensive evaluation of model perplexity and model consistency data. Specifically, from... Figure 2 The data also shows that the consistency scores of the 14 topics reached a local peak, indicating an improvement in semantic consistency within the topics. The model may be capturing a richer topic structure at this point. Although the perplexity curve shows an upward trend, the 14 topics are still within an acceptable range. Therefore, the multi-dimensional topic phrases of the comment text, divided into 14 topics, can be identified as the current comment text topic phrases. The corresponding subject quantity perception model can then be identified as the current topic perception model.

[0057] Step 105: Based on the current topic perception model and the customer review text set, perform semantic recognition processing on each multi-dimensional topic word in the multi-dimensional topic word set of the review text to generate a sequence of recognized topic word information.

[0058] In some embodiments, the aforementioned execution entity may perform semantic recognition processing on each multidimensional topic word in the multidimensional topic word set of the aforementioned comment text based on the aforementioned current topic perception model and the aforementioned customer comment text set, so as to generate a sequence of recognized topic word information.

[0059] In some optional implementations of certain embodiments, the execution entity performs semantic recognition processing on each multidimensional topic word in the multidimensional topic word set of the comment text based on the current topic awareness model and the customer comment text set, in order to generate a sequence of recognized topic word information, which may include the following steps:

[0060] The first step involves using the aforementioned topic-aware model to perform word frequency identification on the multi-dimensional topic words in the multi-dimensional topic word set of the comment text, thereby generating a topic word frequency sequence. Specifically, the current topic-aware model can be used again to extract topic words from each compensated comment text in the compensated comment text set, resulting in a secondary extracted topic word set. Then, topic words that overlap with those in the secondary extracted topic word set and the aforementioned multi-dimensional topic word set of the comment text are identified as potential topic words, thus obtaining a potential topic word set. Finally, the word frequencies of each potential topic word in the potential topic word set can be summarized using Python, resulting in a topic word frequency sequence.

[0061] The second step involves sorting the multi-dimensional keywords of each comment text in the aforementioned set of multi-dimensional keyword groups based on the keyword frequency sequence, in order to generate a high-frequency vocabulary statistics table. This sorting can be done by word frequency.

[0062] The third step is to determine the multi-dimensional topic words and corresponding topic word frequencies of each comment text in the above high-frequency word statistics table as the identified topic word information, and obtain the identified topic word information sequence.

[0063] Step 106: Construct a multi-dimensional word cloud chart based on the identified keyword information sequence.

[0064] In some embodiments, the aforementioned executing entity can construct a multidimensional word cloud chart based on the identified keyword information sequence. Specifically, the identified keyword information sequence can be input into a pre-set word cloud generator to generate the multidimensional word cloud chart.

[0065] As examples, word cloud generators may include: WordClouds.com online word cloud generator, WordItOut word cloud generator, Easy Word Cloud generator, Micro Word Cloud generator, etc.

[0066] The optional implementation methods and related content under steps 104-106 above, as an inventive point of this disclosure, solve the above-mentioned technical problem two: "Even if all words in the text can be found, it is difficult to match the emotional value and customer satisfaction of the customer review text by only counting word importance based on word frequency. This leads to topic word bias. Consequently, the final generated multi-dimensional word cloud chart also has a large bias. Therefore, it also needs to be regenerated, which consumes more computing resources." The factors that lead to more computing resources are often as follows: Even if all words in the text can be found, it is difficult to match the emotional value and customer satisfaction of the customer review text by only counting word importance based on word frequency. This leads to topic word bias. Consequently, the final generated multi-dimensional word cloud chart also has a large bias. Therefore, it also needs to be regenerated. If the above factors are solved, the consumption of computing resources can be reduced. To achieve this effect, firstly, in order to better extract topic words based on customer sentiment, an initial topic perception model is introduced. Here, in order to better distinguish topic words in the text, by dividing topic words, topic words can be randomly assigned into multiple topic combination methods. Then, the optimal topic combination is determined through the initial topic-aware model. This allows for the selection of accurate potential keywords and topic segmentation methods. Finally, this can be used to generate more accurate multi-dimensional word cloud charts, thereby reducing computational resource consumption.

[0067] In practice, when using technical solutions to address the problems mentioned in the background, the following technical problem three often arises: Due to the difficulty in fine-grained segmentation of customer evaluations of merchant services, it becomes difficult to construct accurate positive and negative sentiment word clouds, and the accuracy of the generated models is also difficult to assess. Consequently, errors occur in the stored multi-dimensional word cloud charts and customer satisfaction data tables, leading to wasted storage space. To address this technical problem three, the inventors decided to adopt the following solution.

[0068] Optionally, the aforementioned implementing entity may also perform the following steps:

[0069] The first step involves using a pre-trained comment quality recognition model to perform text recognition on each compensated comment text in the aforementioned compensated comment text set, generating negative comment text label groups and positive comment text label groups. Specifically, the compensated comment texts can be input into the comment quality recognition model to obtain negative and positive comment text labels. Here, the negative and positive comment text labels can be classification labels for topic words.

[0070] As an example, a comment quality identification model can be constructed in the following way:

[0071] Specifically, considering the impact of severe dataset imbalance on model recognition, the BCEWithLogitsLoss loss function, primarily used for multi-label binary classification tasks, was employed to address this issue. Label weights were introduced through the pos_weight parameter, thus resolving the label imbalance problem. The specific calculation method for label weights is as follows: the weight of each label is calculated by determining the number of positive samples for each label, and then applied to the loss function.

[0072] The model is based on BERT-Base-Chinese as its basic architecture and has a custom multi-label classification head, which enables the model output layer to adapt to multi-label binary classification tasks. When the model is initialized, num_labels = 28, where each label is predicted independently using the sigmoid activation function.

[0073] Regarding word segmentation, BertTokenizer was used to ensure that the input text was transformed into an input format suitable for the BERT (Bi Direction Encoder Representations from Transformers) pre-trained model based on the Transformer encoder. By setting padding=True and truncation=True, it was ensured that the input sequence had a uniform length (maximum length 512) and would not exceed the maximum limit.

[0074] In addition, the model incorporates early stopping callbacks to prevent overfitting during training. Training automatically terminates if performance on the validation set no longer improves. In practice, the weighted F1 score is used as the primary evaluation metric, suitable for multi-label problems because it considers the performance of each label and can adjust for imbalanced data. The secondary metric is Hamming Accuracy, which measures accuracy by calculating whether each label is correctly predicted.

[0075] Finally, regarding model configuration and scalability: by mapping id2label and label2id, the label ID and label name are associated, which facilitates the model's label processing and output. Furthermore, the model's problem_type is set to multi_label_classification, making it suitable for multi-label classification problems. Ultimately, the model will output the probability value of each label.

[0076] Here, the AdamW optimizer is used for model optimization, combined with a weight decay strategy to effectively prevent overfitting and accelerate convergence. Furthermore, backpropagation is used to update model parameters and optimize classification performance. Therefore, the model is not only suitable for multi-label classification tasks, but can also be transferred to other text classification tasks, such as single-label classification or multi-task learning, by adjusting the classification head and loss function.

[0077] As an example, the model structure of the comment quality identification model can be found by referring to... Figure 3 As shown.

[0078] Optionally, the comment quality identification model can be generated through the following steps:

[0079] First, after constructing 14 fine-grained dimensions of service quality, 7706 data points were extracted and manually labeled. Relevant dimensions appearing in each comment were marked, and scores were assigned based on their positive or negative sentiment. Sentiment was categorized into two types: positive sentiment was labeled 1, negative sentiment was labeled 2, and any comment without a mentioned variable was labeled 0, resulting in three possible values: {0, 1, 2}. The labeling process strictly adhered to established variable definitions and a keyword database to ensure the accuracy of sentiment classification.

[0080] Next, feature variable decomposition is performed: After labeling, to improve the model's accuracy in learning from samples, all variables are decomposed into binary classification tasks. For example, facility {0,1,2} is decomposed into facility positive {0,1} and facility negative {0,1}. This process allows the model to focus on the discrimination of a single sentiment direction during prediction, improving classification accuracy. At this point, the 14 low-level feature variables are split into positive and negative categories, resulting in 28 low-level feature variables.

[0081] Subsequently, each input embedding is a combination of three embeddings. During preprocessing of the input text, BERT uses positional embeddings to represent the relative position of each word in the sentence, fragment embeddings to distinguish different sentences from the sentence as task input, and labeled embeddings to help the model understand the structure of the input. BERT's pre-training tasks are divided into two main tasks: Masked Language Model (MLM) and Next Sentence Prediction (NSP). In the MLM task, some words are randomly replaced with "[MASK]", and the model needs to predict the replaced words, allowing BERT to capture the context and grammatical relationships of the words. In the NSP task, two sentences are given, and the model needs to determine whether the second sentence logically follows the first sentence. This task helps BERT understand the logical and semantic relationships between sentences.

[0082] As an example, the labeled data in this paper is divided into training, validation, and test sets in an 8:1:1 ratio. The differences between the data results returned by the algorithm and the manually labeled results are statistically analyzed using the values ​​TP, TN, FP, and FN. TP (Truepositive): The number of samples where the model predicts a positive result and the actual label is also positive; TN (Truene negative): The number of samples where the model predicts a negative result and the actual label is also negative; FP (False positive): The number of samples where the model predicts a positive result and the actual label is negative; FN (False negative): The number of samples where the model predicts a negative result and the actual label is positive. Based on TP, TN, FP, and FN, accuracy, precision, recall, and F1 score are calculated as evaluation metrics. Accuracy refers to the proportion of samples correctly predicted by the model out of the total number of samples.

[0083] The second step involves performing the following steps for each comment variable dimension corresponding to each current comment text topic terminology subgroup:

[0084] Step 1: Identify the various compensated comment texts in the compensated comment text set that correspond to the topic word subgroups of the current comment text, and designate them as the target comment text set.

[0085] Step two: Determine the number of each target comment text in the above target comment text group, which will be used as the total number of sentiment-oriented comments.

[0086] Step 3: Determine the number of target comment texts corresponding to the positive comment text tag group in the target comment text group as the number of positive sentiment comments.

[0087] The third step is to create a comment performance analysis chart for each comment variable dimension, based on the total number of sentiment comments and the number of positive sentiment comments corresponding to those dimensions. In this chart, the horizontal axis represents the importance of each comment variable dimension, and the vertical axis represents the satisfaction level of that dimension.

[0088] As an example, refer to Figure 4 . Figure 4The horizontal axis represents performance, measured by frequency of occurrence, i.e., the percentage of the total sentiment sentiment mentioning this dimension variable in the overall population, denoted as I (importance). The vertical axis represents importance, measured by positive ratings, i.e., the percentage of positive sentiment sentiment mentioning this dimension variable in the total sentiment sentiment of the entire dimension variable, denoted as P (positive). The formulas are as follows: I = a / c. P = g / a. Where c is the total number of all comments in the sample, a is the total number of comments with all sentiment sentiments for a certain dimension variable (total number of sentiment sentiment comments), and g is the total number of comments with positive sentiment sentiment mentioning a certain dimension variable (number of positive sentiment sentiment comments). Importing the data from the table into SPSS, both the horizontal and vertical axes are divided into opportunity zones, maintenance zones, strength zones, and repair zones, using the average value of each rating as a reference line.

[0089] The fourth step involves creating positive and negative sentiment word clouds for each comment variable dimension based on the aforementioned comment performance analysis chart, and storing the comment performance analysis chart, each positive sentiment word cloud, and each negative sentiment word cloud. Specifically, for each comment variable dimension, a corresponding positive and negative sentiment word cloud can be generated.

[0090] As an example, taking the comment variable dimension of "facilities" as an example, the generated positive sentiment word cloud and negative sentiment word cloud can be as follows: Figure 5 As shown.

[0091] The fifth step involves determining the performance coordinates of each compensated comment text in the aforementioned compensated comment text set within the comment performance analysis graph. These performance coordinates are then used to label each compensated comment text, resulting in a labeled comment text set. Here, the performance coordinates can be the horizontal and vertical coordinate values ​​in the comment performance analysis graph. Labeling involves marking the performance coordinates at the corresponding word positions within the text.

[0092] The sixth step is to input the tagged comment text set into the preset customer satisfaction regression model to generate customer satisfaction data tables for each comment variable dimension.

[0093] As an example, the framework diagram of a customer satisfaction regression model can be as follows: Figure 6As shown, various data points can be input into the customer satisfaction regression model to generate a customer satisfaction data table. Additionally, a person correlation test can be performed on all comment variable dimensions. The test results show that promotional activities have no significant correlation with tangibility, interpersonal interaction, reliability, hedonism, or policy. Among the other variables with significant correlations, the values ​​are low, indicating no high correlation and generally not causing high multicollinearity. Therefore, regression analysis can continue. To reduce the impact of multicollinearity among independent variables, the independent variables were standardized, and the moderating variables were centered. The interaction term was constructed using the product of the standardized independent variables and the decentralized moderating variables. Then, a multiple linear regression analysis was performed using a hierarchical regression method. The first layer contains control variables, the second layer contains all standardized independent variables, the third layer contains the moderating variables, and the fourth layer contains the interaction term. This results in four corresponding models.

[0094] As an example, the equation for the customer satisfaction regression model could be: Consumer Satisfaction = 0.069 Dining + 0.045 Promotional Activities + 0.123 Tangibles + 0.083 Hedonics + 0.057 Reliability + 0.055 Policy + 0.061 Interpersonal Interaction + 0.053 Shopping Mall Type + 0.051 (Hedonics * Shopping Mall Type) - 0.062 (Dining * Shopping Mall Type) + 0.425. Here, the significance values ​​for interpersonal interaction and tangibles are 0.107 and 0.124 respectively, which do not reach the traditional significance level and show a marginal significance trend; therefore, they are not included in the regression equation for this purpose.

[0095] In practice, statistical results show that there is no collinearity among the variables in the four models. In Model 1, summer and autumn have a significant impact on the scores, with the regression coefficient for autumn being lower than that for summer. The impact of winter on the scores did not reach the commonly used significance level. The R-squared value of the model is 0.006, meaning that summer, autumn, and winter can explain 0.6% of the score variation. An F-test was performed on the model, which passed the F-test (F = 3.030, p < 0.05), indicating that at least one of summer, autumn, and winter has an impact on the scores. The regression coefficient for summer is 0.121 and is significant (t = 2.804, p = 0.005 < 0.01), and the regression coefficient for autumn is 0.101 and is also significant (t = 2.431, p = 0.015 < 0.05), indicating that summer and autumn have a significant positive impact on the scores. The regression coefficient for winter was 0.076, which was not significant (t = 1.937, p = 0.053 > 0.05), indicating that winter does not affect the score. In Model 2, after adding independent variables for tangibility, policy, interpersonal interaction, reliability, hedonism, dining, and promotional activities, all control variables and independent variables became significant, indicating that the addition of independent variables improved the explanatory power of the model for the dependent variable, and the selection of independent variables was reliable. The R-squared value increased from 0.006 to 0.320, and the F-value changed significantly (p < 0.05). The independent variables contributed 31.4% of the explanatory power for the score, and all had a significant positive impact on the score. The degree of influence of each independent variable on the dependent variable, from highest to lowest, was: tangibility, hedonism, interpersonal interaction, policy, dining, promotional activities, and reliability. In Model 3, the shopping mall type was added to Model 2. The change in the F-value was significant (p<0.05), indicating that the addition of the shopping mall type had explanatory value for the model. The R-squared value increased from 0.320 to 0.323, contributing 0.3% to the explanatory power of the score. The selection of the moderating variable was meaningful and had a positive impact on the score. In Model 4, the interaction term between the independent variables and the moderating variable was added. The change in the F-value was significant (p<0.05), indicating that the model had explanatory value. The R-squared value increased from 0.323 to 0.331, contributing 0.8% to the explanatory power of the score. After adding the moderating effect, the influence of the independent variables on the dependent variable also changed. The influence from highest to lowest was: tangibility, hedonism, dining, interpersonal interaction, policy, reliability, and promotional activities. Among them, only the interaction items between dining and shopping mall types showed significance and had a significant negative impact on the scores. The interaction items between people interaction (p=0.107) and hedonism (p=0.085) and shopping mall type showed marginal significance.

[0096] The seventh step is to store each customer satisfaction data table and send each customer satisfaction data table to the target terminal for display.

[0097] The aforementioned optional steps and related content, as an inventive point of this disclosure, solve the third technical problem mentioned above: "Due to the difficulty in finely segmenting customer evaluations of merchant services, it is difficult to construct accurate positive and negative sentiment word clouds, and the accuracy of the generated model is also difficult to determine. Consequently, errors occur in the stored word cloud charts and customer satisfaction data tables, leading to excessive storage space consumption." Factors contributing to increased storage space consumption often include: the difficulty in finely segmenting customer evaluations of merchant services, resulting in the difficulty in constructing accurate positive and negative sentiment word clouds, and the difficulty in determining the accuracy of the generated model. This leads to errors in the stored word cloud charts and customer satisfaction data tables. Solving these factors can reduce storage space consumption. To achieve this, firstly, text recognition can be used to distinguish the sentiment tendencies of keywords in the text, i.e., negative and positive comment text tags. Then, comment performance analysis charts for each comment variable dimension can be determined based on these tags. This characterizes the potential customer intent reflected in customer comments. This allows for more accurate positive and negative sentiment word clouds. Furthermore, performance coordinates are used for labeling, providing deeper textual association features for the introduced customer satisfaction regression model. This ensures the generated customer satisfaction data table aligns with user reviews, thereby improving the accuracy of the generated customer satisfaction data table. Consequently, repeated generation and storage are eliminated, reducing storage resource consumption.

[0098] The above embodiments of this disclosure have the following beneficial effects: the method for generating multi-dimensional service evaluation word cloud charts based on customer perception according to some embodiments of this disclosure can reduce the waste of computing resources. Specifically, the reason for the waste of computing resources is that user reviews often contain simple expressions and missing sentences, resulting in incomplete extraction of all words and difficulty in distinguishing the semantics of words. Consequently, the generated word cloud chart data is incomplete and needs to be regenerated. Based on this, the method for generating multi-dimensional service evaluation word cloud charts based on customer perception according to some embodiments of this disclosure first obtains customer review text from an online service review platform to obtain a customer review text set. Then, considering that the original review text usually contains a lot of non-standard content, such as spelling errors, special symbols, emoticons, etc., which may interfere with text analysis and reduce model performance, and that noisy data (such as HTML tags, advertisements, redundant characters, etc.) will affect word segmentation, word frequency statistics, etc., and that a large number of meaningless characters or words will increase the time cost of data processing, it is necessary to clean the original reviews. Therefore, text compensation is performed on each customer review text in the above customer review text set to generate a compensated review text set. Thus, text compensation improves data quality, reduces noise, and enhances data accuracy. Next, text keyword extraction is performed on each compensated comment text in the compensated comment text set to generate a multi-dimensional keyword set. Here, keyword extraction helps identify keywords and other vocabulary in the comments, facilitating semantic differentiation and text feature extraction. Furthermore, a pre-defined customer topic perception model corresponding to each multi-dimensional keyword set in the multi-dimensional keyword set is determined as the current topic perception model. Then, based on the current topic perception model and the customer comment text set, semantic recognition processing is performed on each multi-dimensional keyword in the multi-dimensional keyword set to generate a sequence of identified keywords. Here, the pre-defined customer topic perception model helps determine the sentiment information of keywords in the comment text based on the multi-dimensional keyword set. This allows for perception of keywords in customer comment texts from the customer's perspective, improving the accuracy of keyword extraction. This facilitates the construction of a multi-dimensional word cloud chart based on the identified keyword information sequence. This can improve the accuracy of the generated multidimensional word cloud charts. Furthermore, it can reduce the computational resources wasted on regeneration.

[0099] Further reference Figure 7 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a multi-dimensional service evaluation word cloud chart generation device based on customer perception. These device embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0100] like Figure 7 As shown, a multi-dimensional service evaluation word cloud chart generation device 700 based on customer perception in some embodiments includes: an acquisition unit 701, a text compensation unit 702, a topic word extraction unit 703, a model determination unit 704, a semantic recognition processing unit 705, and a word cloud chart construction unit 706. The system includes the following components: an acquisition unit 701, configured to acquire customer review texts from an online service review platform to obtain a customer review text set; a text compensation unit 702, configured to perform text compensation on each customer review text in the aforementioned customer review text set to generate a compensated review text set; a topic word extraction unit 703, configured to extract text topic words from each compensated review text in the aforementioned compensated review text set to generate a multi-dimensional topic word set for review texts; a model determination unit 704, configured to determine a preset customer topic perception model corresponding to each multi-dimensional topic word set in the aforementioned multi-dimensional topic word set for review texts, as the current topic perception model; a semantic recognition processing unit 705, configured to perform semantic recognition processing on each multi-dimensional topic word in the aforementioned multi-dimensional topic word set for review texts based on the current topic perception model and the aforementioned customer review text set to generate a sequence of recognized topic word information; and a word cloud construction unit 706, configured to construct a multi-dimensional word cloud chart based on the sequence of recognized topic word information.

[0101] It is understandable that the units described in the device 700 are related to the reference. Figure 1 The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method also apply to the device 700 and the units contained therein, and will not be repeated here.

[0102] The following is for reference. Figure 8 It shows a schematic diagram of the structure of an electronic device (e.g., a computing device) 800 suitable for implementing some embodiments of the present disclosure. Figure 8 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0103] like Figure 8 As shown, the electronic device 800 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory 802 or a program loaded from a storage device 808 into a random access memory 803. The random access memory 803 also stores various programs and data required for the operation of the electronic device 800. The processing unit 801, the read-only memory 802, and the random access memory 803 are interconnected via a bus 804. An input / output interface 805 is also connected to the bus 804.

[0104] Typically, the following devices can be connected to I / O interface 805: input devices 806 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 807 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 808 including, for example, magnetic tapes, hard disks, etc.; and communication devices 809. Communication device 809 allows electronic device 800 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 8 An electronic device 800 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 8 Each box shown can represent a device or multiple devices as needed.

[0105] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 809, or installed from a storage device 808, or installed from a read-only memory 802. When the computer program is executed by the processing device 801, it performs the functions defined above in the methods of some embodiments of this disclosure.

[0106] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0107] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or future-developed networks.

[0108] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: obtain customer review text from an online service review platform to obtain a customer review text set; perform text compensation on each customer review text in the aforementioned customer review text set to generate a compensated review text set; extract text keywords from each compensated review text in the aforementioned compensated review text set to generate a multi-dimensional keyword set for review texts; determine a preset customer topic perception model corresponding to each multi-dimensional keyword set in the aforementioned multi-dimensional keyword set for review texts as the current topic perception model; based on the current topic perception model and the aforementioned customer review text set, perform semantic recognition processing on each multi-dimensional keyword in the aforementioned multi-dimensional keyword set for review texts to generate a sequence of recognized keyword information; and construct a multi-dimensional word cloud chart based on the sequence of recognized keyword information.

[0109] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0110] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0111] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including an acquisition unit, a text compensation unit, a topic word extraction unit, a model determination unit, a semantic recognition processing unit, and a word cloud construction unit. The names of these units do not necessarily limit the specific unit; for example, the acquisition unit may also be described as a "unit for acquiring customer review text."

[0112] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0113] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for generating a multi-dimensional service evaluation word cloud chart based on customer perception, comprising: Obtain customer review texts from online service review platforms to obtain a set of customer review texts; Text compensation is performed on each customer comment text in the customer comment text set to generate a compensated comment text set; Text keyword extraction is performed on each compensated comment text in the compensated comment text set to generate a multidimensional keyword set of comment texts; The process of determining a preset customer topic perception model corresponding to each multi-dimensional topic phrase in the set of multi-dimensional topic phrases in the comment text, as the current topic perception model, includes: obtaining an initial topic perception model from a database; iterating the initial topic perception model using each multi-dimensional topic phrase in the set of multi-dimensional topic phrases in the comment text to obtain a set of topic quantity perception models, wherein each topic quantity perception model corresponds to one multi-dimensional topic phrase in the set of multi-dimensional topic phrases in the comment text, each multi-dimensional topic phrase corresponds to a different number of topics, and each multi-dimensional topic phrase is divided into multiple subgroups of multi-dimensional topic phrases in the comment text according to the number of topics; determining the model perplexity and model consistency data corresponding to each topic quantity perception model in the set of topic quantity perception models; and determining the topic quantity perception model corresponding to the model perplexity and model consistency data that meets the preset topic division conditions as the current topic perception model, wherein the multi-dimensional topic phrase in the comment text corresponding to the current topic perception model is the current comment text topic phrase; Based on the current topic perception model and the customer review text set, semantic recognition processing is performed on each multidimensional topic word in the multidimensional topic word set of the review text to generate a sequence of recognized topic word information; Based on the identified keyword information sequence, a multidimensional word cloud chart is constructed.

2. The method according to claim 1, wherein, The method further includes: The multidimensional word cloud chart is stored and then sent to a display terminal for display.

3. The method according to claim 1, wherein, The step of performing text compensation on each customer review text in the customer review text set to generate a compensated review text set includes: Character recognition is performed on each customer comment text in the customer comment text set to remove meaningless characters, resulting in a comment text set after character removal; Using a pre-established stop word dictionary, stop word analysis is performed on each of the removed comment texts in the removed comment text set to generate a compensated comment text set. The stop word analysis is used to determine the stop words and their part-of-speech meanings in each removed comment text.

4. The method according to claim 3, wherein, The step of extracting text keywords from each compensated comment text in the compensated comment text set to generate a multi-dimensional keyword set for the comment text includes: For each compensated comment text in the compensated comment text set, keyword extraction is performed to generate the current comment text keyword group, resulting in the current comment text keyword group set; The current comment text topic word set is deduplicated to generate a deduplicated text topic word set; The keywords in the deduplicated text keyword set are divided into themes to generate a multidimensional keyword set for the comment text.

5. The method according to claim 1, wherein, The step involves semantic recognition processing of each multi-dimensional topic word in the multi-dimensional topic word set of the comment text, based on the current topic perception model and the customer comment text set, to generate a sequence of recognized topic word information, including: Using the current topic perception model, word frequency identification is performed on each multidimensional topic word in the multidimensional topic word group set of the comment text to generate a topic word frequency sequence; Based on the keyword frequency sequence, the multi-dimensional keywords of each comment text in the set of main multi-dimensional keyword groups are sorted to generate a high-frequency vocabulary statistics table; The multidimensional topic words and their corresponding word frequencies of each comment text in the high-frequency word statistics table are determined as the identified topic word information, thus obtaining the identified topic word information sequence.

6. The method according to claim 5, wherein, The step of constructing a multi-dimensional word cloud chart based on the identified keyword information sequence includes: Using the high-frequency vocabulary statistics table, word clouds are drawn for the multi-dimensional keywords of the comment text included in each identified keyword information sequence, resulting in a multi-dimensional word cloud chart.

7. A device for generating a multi-dimensional service evaluation word cloud chart based on customer perception, comprising: The acquisition unit is configured to retrieve customer review text from the online review platform to obtain a set of customer review texts. The text compensation unit is configured to perform text compensation on each customer comment text in the customer comment text set to generate a compensated comment text set; The keyword extraction unit is configured to extract keywords from each compensated comment text in the compensated comment text set to generate a multidimensional keyword set for the comment text. The model determination unit is configured to determine a preset customer topic perception model corresponding to each comment text multi-dimensional topic phrase in the comment text multi-dimensional topic phrase set as the current topic perception model. This includes: obtaining an initial topic perception model from a database; iterating the initial topic perception model using each comment text multi-dimensional topic phrase in the comment text multi-dimensional topic phrase set to obtain a topic quantity perception model set, wherein each topic quantity perception model corresponds to one comment text multi-dimensional topic phrase in the comment text multi-dimensional topic phrase set, each comment text multi-dimensional topic phrase corresponds to a different topic quantity, and each comment text multi-dimensional topic phrase is divided into multiple comment text multi-dimensional topic phrase subgroups based on the topic quantity; determining the model perplexity and model consistency data corresponding to each topic quantity perception model in the topic quantity perception model set; and determining the topic quantity perception model corresponding to the model perplexity and model consistency data that satisfies the preset topic division conditions as the current topic perception model, wherein the comment text multi-dimensional topic phrase corresponding to the current topic perception model is the current comment text topic phrase. The semantic recognition processing unit is configured to perform semantic recognition processing on each multidimensional topic word in the multidimensional topic word set of the comment text based on the current topic perception model and the customer comment text set, so as to generate a sequence of recognized topic word information. The word cloud construction unit is configured to construct a multidimensional word cloud chart based on the identified topic word information sequence.

8. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.

9. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

Images