A method, apparatus, device, and readable storage medium for extracting service quality evaluation data.

Abstract

This invention provides a service quality evaluation extraction method, apparatus, device, and readable storage medium, relating to the field of data processing technology. This method improves the generalization performance of the algorithm by using dual-granularity sentiment extraction, double deduplication, and preventing data misalignment, thereby establishing a more accurate service quality evaluation model for the banking industry. It also reduces the reliance of traditional service quality evaluation methods on expert subjective opinions and the relatively one-sided extraction results obtained by single feature value evaluation models.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method, apparatus, device, and readable storage medium for extracting service quality evaluation data. Background Technology

[0002] The sentiment descriptions in service quality texts are highly arbitrary, the data is mixed, and it is constrained by the subjective feelings of customers. Much of the text data is also non-standardized. Traditional convolutional neural network methods cannot effectively extract sentiment information from texts, nor can they handle long texts well, thus failing to effectively extract service quality evaluations. Summary of the Invention

[0003] The purpose of this invention is to provide a service quality evaluation extraction method, apparatus, device, and readable storage medium to improve the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows:

[0004] In a first aspect, this application provides a service quality evaluation extraction method, comprising: acquiring first information, the first information including at least one service quality evaluation data of a customer on a service; performing coarse-grained extraction calculation on the first information to obtain second information, the second information including a dimensionality-reduced sentiment dataset, the sentiment dataset including a set of service evaluation results extracted from all evaluations and a set of sentiment expressions; performing fine-grained extraction and inductive calculation on the second information to obtain service quality evaluation information, each third element in the service quality evaluation set including a service evaluation and the sentiment expression corresponding to the service evaluation; and performing a double induction on the service quality evaluation information to obtain the inductive service quality evaluation information.

[0005] Furthermore, the step of performing coarse-grained extraction and calculation of the first information to obtain the second information includes: preprocessing the first information to obtain a vectorized sample set; and performing coarse-grained extraction of sentiment evaluation on the sample set to obtain the second information.

[0006] Further, the step of performing coarse-grained extraction of sentiment evaluation from the sample set to obtain the second information includes: classifying the sample set by part-of-speech tags to obtain the part-of-speech categorized sample set; extracting the sentiment based on subject-verb relations from the part-of-speech categorized sample set to obtain a sentiment dataset, wherein the sentiment dataset includes a set of service evaluation results and a set of sentiment expressions obtained from all evaluation extractions, and each first element in the service evaluation result set uniquely corresponds to a second element in the sentiment expression set; and transforming and reducing the dimensionality of the sentiment dataset to obtain the second information.

[0007] Further, the step of performing fine-grained extraction and inductive calculation on the second information to obtain service quality evaluation information includes: establishing an LSTM mathematical model; using the second information as input information to the LSTM mathematical model, solving the LSTM mathematical model to obtain third information, wherein the third information is the set of sentiment expressions after fine-grained extraction of the sentiment results from the second information, the set of sentiment expressions including the set of service evaluation results after fine-grained extraction and the set of sentiment expressions after fine-grained extraction; performing phase-correction matching on the set of sentiment expressions after fine-grained extraction to obtain a service quality evaluation set, wherein each third element in the service quality evaluation set includes a service evaluation and the sentiment expression corresponding to the service evaluation.

[0008] Further, the step of performing phase-corrected matching on the finely extracted sentiment expression set to obtain a service quality evaluation set includes: extracting the number of elements from the finely extracted sentiment expression set to obtain the total number of a first element and the total number of a second element in the finely extracted service evaluation result set; if the total number of the first element in the finely extracted service evaluation result set is equal to the total number of the second element in the finely extracted sentiment expression set, then the evaluation result set is matched and extracted to obtain the phase-corrected sentiment expression set. If the total number of the first element in the fine-grained extracted evaluation result set is greater than the total number of the second element in the fine-grained extracted emotion expression set, then error correction matching calculation is performed based on the fine-grained extracted evaluation result set and the emotion expression set to obtain the error-corrected emotion expression set; if the total number of elements in the fine-grained extracted evaluation result set is less than the total number of elements in the fine-grained extracted emotion expression set, then error correction matching calculation is performed based on the fine-grained extracted emotion expression set and the evaluation result set to obtain the error-corrected emotion expression set.

[0009] Further, the step of performing error correction matching calculation based on the finely extracted evaluation result set and the sentiment expression set to obtain the error-corrected sentiment expression set includes: for each first element in the finely extracted evaluation result set, using the minimum text matching distance method to find the sentiment expression corresponding to each element in the sentiment expression set; denoting each first element and the sentiment expression corresponding to each first element in the finely extracted evaluation result set as the third element; and denoting all the third elements as the error-corrected sentiment expression set.

[0010] Secondly, this application also provides a service quality evaluation extraction device, comprising: a first acquisition unit, a coarse-grained extraction unit, a fine-grained extraction unit, and a double induction unit; the first acquisition unit is used to acquire first information, the first information including at least one service quality evaluation data of a customer on a service; the coarse-grained extraction unit is used to perform coarse-grained extraction calculation on the first information to obtain second information, the second information including a dimensionality-reduced sentiment dataset, the sentiment dataset including a set of service evaluation results extracted from all evaluations and a set of sentiment expressions; the fine-grained extraction unit is used to perform fine-grained extraction and induction calculation on the second information to obtain service quality evaluation information, each third element in the service quality evaluation set including a service evaluation and the sentiment expression corresponding to the service evaluation; the double induction unit is used to perform double induction on the service quality evaluation information to obtain the inducted service quality evaluation information.

[0011] Furthermore, the coarse-grained extraction unit includes: a preprocessing unit for preprocessing the first information to obtain a vectorized sample set; and a first granularity unit for performing coarse-grained emotion evaluation extraction on the sample set to obtain the second information.

[0012] Further, the first granularity unit includes: a classification unit, used to classify the sample set by part of speech to obtain the sample set after part of speech classification; an extraction unit, used to extract the sentiment of subject-predicate relationship from the sample set after part of speech classification to obtain a sentiment dataset, the sentiment dataset including a service evaluation result set and a sentiment expression set obtained from all evaluations, each first element in the service evaluation result set uniquely corresponds to a second element in the sentiment expression set; and a transformation unit, used to transform and reduce the dimensionality of the sentiment dataset to obtain second information.

[0013] Further, the fine-grained extraction unit includes: a model building unit for building an LSTM mathematical model; a second granularity unit for using the second information as input information to the LSTM mathematical model and solving the LSTM mathematical model to obtain third information, wherein the third information is the sentiment expression set after fine-grained extraction of the sentiment result from the second information, the sentiment expression set including the service evaluation result set after fine-grained extraction and the sentiment expression set after fine-grained extraction; and a correction matching unit for performing phase correction matching on the sentiment expression set after fine-grained extraction to obtain a service quality evaluation set, wherein each third element in the service quality evaluation set includes a service evaluation and the sentiment expression corresponding to the service evaluation.

[0014] Further, the correction matching unit includes: a quantity extraction unit, used to extract the number of elements from the fine-grained extracted sentiment expression set to obtain the total number of first elements in the fine-grained extracted service evaluation result set and the total number of second elements in the fine-grained extracted sentiment expression set; a first judgment unit, used to perform matching extraction on the evaluation result set if the total number of first elements in the fine-grained extracted service evaluation result set is equal to the total number of second elements in the fine-grained extracted sentiment expression set, to obtain the sentiment expression set after misalignment correction; and a second judgment unit, used to perform matching extraction on the fine-grained extracted sentiment expression set if the total number of first elements in the fine-grained extracted service evaluation result set is equal to the total number of second elements in the fine-grained extracted sentiment expression set. If the total number of the first element in the evaluation result set after fine-grained extraction is greater than the total number of the second element in the emotion expression set after fine-grained extraction, then error correction matching calculation is performed based on the evaluation result set after fine-grained extraction and the emotion expression set to obtain the emotion expression set after error correction; the third judgment unit is used to perform error correction matching calculation based on the emotion expression set after fine-grained extraction and the evaluation result set if the total number of elements in the evaluation result set after fine-grained extraction is less than the total number of elements in the emotion expression set after fine-grained extraction to obtain the emotion expression set after error correction.

[0015] Further, the second judgment unit includes: a traversal unit, used to find the corresponding sentiment expression for each first element in the set of sentiment expressions after fine-grained extraction using the minimum text matching distance method; a pairing unit, used to record each first element and the corresponding sentiment expression in the set of sentiment expressions after fine-grained extraction as the third element; and a marking unit, used to record all the third elements as the set of sentiment expressions after phase error correction.

[0016] Thirdly, this application also provides a service quality evaluation extraction device, comprising:

[0017] Memory, used to store computer programs;

[0018] A processor is used to implement the steps of the service quality evaluation extraction method when executing the computer program.

[0019] Fourthly, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described service quality evaluation extraction method.

[0020] The beneficial effects of this invention are as follows:

[0021] 1. This invention enhances the understanding of genuine customer emotions through two-stage granularity extraction.

[0022] 2. This invention improves the accuracy of subsequent neural network algorithms by extracting the sentiment of the subject-verb relationship in the evaluation and performing syntactic matching on the data, while reducing the problem of manually labeled datasets.

[0023] 3. This method improves the generalization performance of the algorithm by using dual-granularity sentiment extraction, double deduplication, and preventing data misalignment, and establishes a more accurate banking service quality evaluation model. It reduces the reliance of traditional service quality evaluation methods on expert subjective opinions and the relatively one-sided results obtained by single feature value evaluation models.

[0024] 4. By using clustering evaluation, the sentiment elements extracted by the neural network are further processed, and the results of similar evaluation categories are integrated to reduce the classification of sentiment elements and further improve the accuracy of the model's sentiment evaluation.

[0025] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a schematic diagram of the service quality evaluation extraction method described in this embodiment of the invention;

[0028] Figure 2 This is a schematic diagram of the service quality evaluation extraction device described in this embodiment of the invention;

[0029] Figure 3 This is a schematic diagram of the service quality evaluation extraction device described in this embodiment of the invention;

[0030] Figure 4 This is a schematic diagram of the service quality evaluation extraction device described in an embodiment of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0032] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0033] Example 1:

[0034] This embodiment provides a method for extracting service quality evaluation data.

[0035] See Figure 1 The figure shows that the method includes steps S100, S200, S300, S400 and S500.

[0036] S100. Obtain first information, which includes at least one service quality evaluation data point from the customer regarding the business.

[0037] Understandably, this step is to obtain multiple service quality evaluation data, such as entering all evaluation information corresponding to a certain service.

[0038] S200. The first information is coarsely extracted and calculated to obtain the second information. The second information includes the dimensionality-reduced sentiment dataset, which includes the set of service evaluation results extracted from all evaluations and the set of sentiment expressions.

[0039] Understandably, this step is for preliminary data processing, initial standardization of text data, and annotation of evaluation data to reduce the workload of manual data annotation. Specifically, this step includes steps S210 and S220.

[0040] S210. Preprocess the first information to obtain a vectorized sample set.

[0041] Specifically, this step includes steps S211, S212 and S213.

[0042] S211. Filter out invalid data in the first information and update the first information.

[0043] It is understandable that in actual service quality evaluation, some customers will not enter useful information, but will enter some pure symbols or meaningless words. This step is to filter out useless data and reduce its interference with the subsequent deep neural network model. The specific implementation of this step is common knowledge in the field and can be implemented by simply following some rules, so it will not be described in detail in this application.

[0044] S212. Perform Chinese text segmentation on each service quality evaluation in the first information and update the first information.

[0045] It should be noted that the Chinese text segmentation method used in this step is calculated using the CRF algorithm.

[0046] S213. The first information is vectorized to obtain a vectorized sample set.

[0047] It should be noted that the vectorization process mentioned in this step is common knowledge in the field, and those skilled in the art can implement it using the One-hot algorithm.

[0048] S220. Perform coarse-grained extraction of sentiment evaluation on the sample set to obtain the second information.

[0049] Specifically, this step addresses the issues of arbitrary sentiment descriptions and mixed data in service quality evaluations by first extracting coarse-grained data from the service quality evaluations to reduce their dimensionality. It also merges duplicate evaluations to mitigate subsequent data processing problems.

[0050] Specifically, this step includes steps S221, S222, and S223.

[0051] S221. Perform part-of-speech classification on the sample set to obtain the part-of-speech classified sample set;

[0052] Specifically, the open-source pyltp used in this step performs part-of-speech tagging for service quality evaluation. By dividing a long text into different parts of speech, the original input text "The business staff are very enthusiastic, but the queue time is too long." becomes "Business staff / very / enthusiastic, but / queue / time / too / long / " after segmentation. This improves the accuracy of subject-verb part-of-speech tagging in the later stage and ensures that the word processing in this method is based on words with independent meaning.

[0053] S222. Subject-verb relationship sentiment extraction is performed on the sample set after part-of-speech classification to obtain the sentiment dataset. The sentiment dataset includes the service evaluation result set and the sentiment expression set obtained from all evaluations. Each first element in the service evaluation result set uniquely corresponds to a second element in the sentiment expression set.

[0054] Specifically, this step is combined with step S221, using a single-sentence dependency syntax method. Taking the example of S221 as an illustration, after extracting the subject-verb relationship, we can obtain "business personnel, enthusiasm" and "time, long". After the coarse-grained extraction in this step, the correspondence between the subject and the verb is matched with the data.

[0055] S223. Transform and reduce the dimensionality of the sentiment dataset to obtain the second information.

[0056] Specifically, in this step, the sentiment dataset obtained through step S222 is transformed and its dimensionality reduced using the word2vec word vector model.

[0057] S300. The second information is extracted and summarized in a fine-grained manner to obtain service quality evaluation information. Each third element in the service quality evaluation set includes a service evaluation and the corresponding emotional expression of the service evaluation.

[0058] Understandably, this step involves further refining the already coarsely processed service quality evaluation to categorize what aspects are "good" and what specific manifestations are "bad".

[0059] Specifically, S300 also includes steps S310, S320 and S330.

[0060] S310. Establish the LSTM mathematical model.

[0061] S320. The second information is used as the input information of the LSTM mathematical model. The LSTM mathematical model is solved to obtain the third information. The third information is the sentiment expression set after fine-grained extraction of the sentiment results from the second information. The sentiment expression set includes the service evaluation result set after fine-grained extraction and the sentiment expression set after fine-grained extraction.

[0062] Understandably, this step utilizes more accurate user review data obtained through unsupervised learning. This data is then combined with the long sentence segmentation data from S200 to improve the extraction accuracy of service quality evaluations.

[0063] S330. Perform phase error correction matching on the fine-grained extracted sentiment expression set to obtain the service quality evaluation set. Each third element in the service quality evaluation set includes a service evaluation and the sentiment expression corresponding to the service evaluation.

[0064] It should be noted that this step is an optimization to address the issue of multiple matching results that may occur during the object classification and extraction process of the LSTM algorithm.

[0065] Specifically, this step also includes steps S331, S332, S333 and S334.

[0066] S331. Extract the number of elements from the fine-grained extracted sentiment expression set to obtain the total number of the first element in the fine-grained extracted service evaluation result set and the total number of the second element in the fine-grained extracted sentiment expression set.

[0067] S332. If the total number of the first element in the service evaluation result set after fine-grained extraction is equal to the total number of the second element in the sentiment expression set after fine-grained extraction, then the evaluation result set is matched and extracted to obtain the sentiment expression set after phase correction.

[0068] It should be noted that the evaluation result set matching and extraction mentioned in this step refers to constructing a third element from each service evaluation and its corresponding sentiment expression, and then constructing a sentiment expression set using all third elements. This embodiment illustrates this with the following example: the service evaluation result set includes {U1, U2, ... U...} q The set of emotion expressions includes {T1, T2, ..., T}. p Since q = p in this step, we only need to use {U1, T1} as the third element in this step, and finally form the emotion expression set S = {(U1, T1), (U2, T2), ... (U... q ,T p )}.

[0069] S333. If the total number of the first element in the fine-grained extracted evaluation result set is greater than the total number of the second element in the fine-grained extracted sentiment expression set, then error correction matching calculation is performed based on the fine-grained extracted evaluation result set and sentiment expression set to obtain the error-corrected sentiment expression set.

[0070] It should be noted that the purpose of this step is to determine whether the second element of the sentiment expression set after fine-grained extraction is missing after LSTM fine-grained extraction. Since it is unknown which second element is missing from the sentiment expression set after fine-grained extraction, the matching relationship between the sentiment expression set after fine-grained extraction and the evaluation result set after fine-grained extraction is disordered.

[0071] Therefore, in order to solve the above problems, this step also includes steps S3331, S3331 and S3331.

[0072] S3331. For each first element in the set of evaluation results after fine-grained extraction, use the minimum text matching distance method to find the corresponding sentiment expression in the sentiment expression set.

[0073] It should be noted that in this step, the second element corresponding to each first element is found using the minimum text matching distance method.

[0074] S3332. Denote each first element and the sentiment expression corresponding to each first element in the set of evaluation results after fine-grained extraction as the third element;

[0075] This can be understood as follows: this step involves combining a first element with the first and second elements to form a third element, which is (U1,T1) in step S332, ultimately forming an emotion expression set S = {(U1,T1),(U2,T2),…(U...}}. q ,T q )}.

[0076] S3333, denote all third elements as the set of emotion expressions after phase correction.

[0077] S334. If the total number of elements in the evaluation result set after fine-grained extraction is less than the total number of elements in the sentiment expression set after fine-grained extraction, then error correction matching calculation is performed based on the sentiment expression set after fine-grained extraction and the evaluation result set to obtain the sentiment expression set after error correction.

[0078] It should be noted that this step is similar to step S333, only the objects being processed are different.

[0079] By processing step S330, the mismatch results that may occur in the object classification and extraction process of the LSTM algorithm are reduced, thus improving the accuracy of this method.

[0080] S400. Perform a double summarization of the service quality evaluation information to obtain the summarized service quality evaluation information.

[0081] It should be noted that the LSTM algorithm suffers from accuracy degradation due to the problem of repeated sentiment elements during sentiment extraction. Therefore, this step further includes step S410 to summarize the sentiment elements.

[0082] S410, Obtain the preset K value;

[0083] It should be noted that the value of K is an empirical value, which is roughly judged based on the quantity in the first information. In this embodiment, the sum is preferably 7.

[0084] S420. Based on the service quality evaluation information, the first cluster center point is calculated using the density method;

[0085] For ease of understanding, in S400, the service quality evaluation information is represented by a set S, where set S includes {S1, S2, ... S}. n}. Where each third element has S i The formula for calculating the density value is as follows:

[0086] p i =∑x(d ij -d c )

[0087] Where, d ij For S i and S j Euclidean distance, d c Let x be the average distance between the sample sets of S, and let x be (d ij -d c When ) is less than zero, it is zero, and x is (d ij -d c When it is greater than zero, it is one.

[0088] d c The calculation formula is as follows:

[0089]

[0090] S430. Based on the service quality evaluation information, calculate K-1 cluster centers using a preset formula, with each cluster center corresponding to a class.

[0091] Specifically, the preset formula used in this step is:

[0092] dist i =max{min(d i1 d i2 , ..., d in ,)}(i=1,2…n)

[0093] S440, Double Induction Extraction: Calculate the weighted distance from each third element in the service quality evaluation information to each cluster center point in turn, and assign each third element to the class corresponding to the cluster center point with the smallest weighted distance; calculate the mean of the third elements in each class, and use the mean to update the cluster center point corresponding to the class;

[0094] It should be noted that the weighted distance calculation formula used in this step is as follows:

[0095]

[0096] In the formula, and Denotes the sample S in space (l) dimension. Iand S j , where m is the number of spatial dimensions.

[0097] S450. Repeat the double inductive extraction until the mean of the third element in each class remains constant.

[0098] Compared to the traditional dependency parsing + LSTM model, this method addresses the LSTM model's poor handling of multiple evaluation results from a single text by employing text segmentation and part-of-speech tagging. Furthermore, to address the issue of misplaced sentiment elements during sentiment extraction in the LSTM model, it utilizes error-correcting matching and subsequent double induction to reduce the probability of sentiment element repetition. In the double induction, a combination of density-based and minimum-maximum distance methods is used to traverse and find K centroids, resolving the issue of biased initial centroid selection. A weighted distance allocation method is also employed to mitigate the problem of getting trapped in local optima. This method achieves higher accuracy in extracting public Chinese service quality evaluations, accurately locating users' true emotions, and obtaining more precise user profiles.

[0099] S500, Send a first display command, the first display command including a command to display the summarized service quality evaluation information through a word cloud diagram.

[0100] See Figure 2 The figure shows the word cloud after processing in this embodiment.

[0101] Example 2:

[0102] As shown in the figure, this embodiment provides a service quality evaluation extraction device. See [link / reference]. Figure 3 The device includes:

[0103] The first acquisition unit 1 is used to acquire first information, which includes at least one service quality evaluation data of the customer on the business;

[0104] The coarse-grained extraction unit 2 is used to perform coarse-grained extraction calculations on the first information to obtain the second information. The second information includes the dimensionality-reduced sentiment dataset, which includes the set of service evaluation results extracted from all evaluations and the set of sentiment expressions.

[0105] The fine-grained extraction unit 3 is used to perform fine-grained extraction and inductive calculation on the second information to obtain service quality evaluation information. Each third element in the service quality evaluation set includes a service evaluation and the sentiment expression corresponding to the service evaluation.

[0106] The double summarization unit 4 is used to perform double summarization on the service quality evaluation information to obtain the summarized service quality evaluation information.

[0107] In some specific embodiments, the coarse-grained extraction unit 2 includes:

[0108] Preprocessing unit 21 is used to preprocess the first information to obtain a vectorized sample set;

[0109] The first granularity unit 22 is used to perform coarse-grained extraction of sentiment evaluation from the sample set to obtain the second information.

[0110] In some specific embodiments, the first granularity unit 22 includes:

[0111] Classification unit 221 is used to perform part-of-speech classification on the sample set to obtain a part-of-speech classified sample set;

[0112] Extraction unit 222 is used to extract the sentiment of subject-predicate relation from the sample set after part-of-speech classification to obtain the sentiment dataset. The sentiment dataset includes the service evaluation result set and the sentiment expression set obtained from all evaluations. Each first element in the service evaluation result set uniquely corresponds to a second element in the sentiment expression set.

[0113] Transformation unit 223 is used to transform and reduce the dimensionality of the sentiment dataset to obtain the second information.

[0114] In some specific embodiments, the fine-grained extraction unit 3 includes:

[0115] Model building unit 31 is used to build the LSTM mathematical model;

[0116] The second granularity unit 32 is used to take the second information as the input information of the LSTM mathematical model, solve the LSTM mathematical model to obtain the third information, the third information is the sentiment expression set after fine-grained extraction of the sentiment result of the second information, the sentiment expression set includes the service evaluation result set after fine-grained extraction and the sentiment expression set after fine-grained extraction.

[0117] The correction matching unit 33 is used to perform phase error correction matching on the fine-grained extracted sentiment expression set to obtain the service quality evaluation set. Each third element in the service quality evaluation set includes a service evaluation and the sentiment expression corresponding to the service evaluation.

[0118] In some specific embodiments, the correction matching unit 33 includes:

[0119] The quantity extraction unit 331 is used to extract the number of elements from the fine-grained extracted sentiment expression set to obtain the total number of the first element in the fine-grained extracted service evaluation result set and the total number of the second element in the fine-grained extracted sentiment expression set.

[0120] The first judgment unit 332 is used to match and extract the evaluation result set if the total number of the first element in the service evaluation result set after fine-grained extraction is equal to the total number of the second element in the sentiment expression set after fine-grained extraction, so as to obtain the sentiment expression set after phase correction.

[0121] The second judgment unit 333 is used to perform error correction matching calculation based on the fine-grained evaluation result set and the emotion expression set if the total number of the first element in the fine-grained evaluation result set is greater than the total number of the second element in the fine-grained emotion expression set, so as to obtain the error-corrected emotion expression set.

[0122] The third judgment unit 334 is used to perform error correction matching calculation based on the fine-grained extracted emotion expression set and the evaluation result set if the total number of elements in the fine-grained extracted evaluation result set is less than the total number of elements in the fine-grained extracted emotion expression set, so as to obtain the error-corrected emotion expression set.

[0123] In some specific embodiments, the second determination unit 333 includes:

[0124] Traverse unit 3341, which is used to find the sentiment expression corresponding to each element in the sentiment expression set for each first element in the fine-grained extracted evaluation result set using the minimum text matching distance method.

[0125] The pairing unit 3342 is used to denote each first element and the sentiment expression corresponding to each first element in the fine-grained extracted evaluation result set as the third element.

[0126] Labeling unit 3343 is used to denote all third elements as the set of sentiment expressions after phase error correction.

[0127] In some specific embodiments, the double induction unit 4 includes:

[0128] The second acquisition unit 41 is used to acquire a preset K value;

[0129] The first calculation unit 42 is used to calculate the first cluster center point using the density method based on the service quality evaluation information;

[0130] The second calculation unit 43 is used to calculate K-1 cluster center points based on service quality evaluation information using a preset formula, and each cluster center point corresponds to a class.

[0131] Repeating unit 44 is used for double inductive extraction: sequentially calculate the weighted distance from each third element in the service quality evaluation information to each cluster center point, and assign each third element to the class corresponding to the cluster center point with the smallest weighted distance; calculate the mean of the third elements in each class, and use the mean to update the cluster center point corresponding to the class;

[0132] Repeat judgment unit 45 is used to repeatedly perform double inductive extraction until the mean of the third element in each class remains unchanged.

[0133] It should be noted that the specific manner in which each module performs its operation in the apparatus described in the above embodiments has been described in detail in the embodiments of the method, and will not be elaborated here.

[0134] Example 3:

[0135] Corresponding to the above method embodiments, this embodiment also provides a service quality evaluation extraction device. The service quality evaluation extraction device described below and the service quality evaluation extraction method described above can be referred to in correspondence.

[0136] Figure 4 This is a block diagram illustrating a service quality evaluation extraction device 800 according to an exemplary embodiment. For example... Figure 4 As shown, the service quality assessment extraction device 800 may include: a processor 801 and a memory 802. The service quality assessment extraction device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0137] The processor 801 controls the overall operation of the service quality evaluation extraction device 800 to complete all or part of the steps in the service quality evaluation extraction method described above. The memory 802 stores various types of data to support the operation of the service quality evaluation extraction device 800. This data may include, for example, instructions for any application or method operating on the service quality evaluation extraction device 800, and application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the service quality evaluation extraction device 800 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0138] In an exemplary embodiment, the service quality assessment extraction device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the service quality assessment extraction method described above.

[0139] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the service quality assessment extraction method described above. For example, the computer-readable storage medium may be the memory 802 including the program instructions described above, which may be executed by the processor 801 of the service quality assessment extraction device 800 to complete the service quality assessment extraction method described above.

[0140] Example 4:

[0141] Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the service quality evaluation extraction method described above.

[0142] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the service quality evaluation extraction method described in the above method embodiments.

[0143] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0144] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0145] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for extracting service quality evaluation criteria, characterized in that, include: Obtain first information, which includes at least one service quality evaluation data point from a customer regarding the business; The first information is coarsely extracted and calculated to obtain the second information, which includes a dimensionality-reduced sentiment dataset. The sentiment dataset includes a set of service evaluation results extracted from all evaluations and a set of sentiment expressions. The second information is extracted and summarized in a fine-grained manner to obtain service quality evaluation information. Each third element in the service quality evaluation set includes a service evaluation and the sentiment expression corresponding to the service evaluation. The service quality evaluation information is subjected to a double induction to obtain the induced service quality evaluation information. The process of extracting and summarizing the second information in a fine-grained manner to obtain service quality evaluation information includes: Establish an LSTM mathematical model; The second information is used as the input information of the LSTM mathematical model, and the LSTM mathematical model is solved to obtain the third information. The third information is the emotional expression set after fine-grained extraction of the emotional results from the second information. The emotional expression set includes the service evaluation result set after fine-grained extraction and the emotional expression set after fine-grained extraction. The sentiment expression set after fine-grained extraction is subjected to phase error correction matching to obtain a service quality evaluation set. Each third element in the service quality evaluation set includes a service evaluation and the sentiment expression corresponding to the service evaluation. Also includes: Obtain a preset experience value K; the experience value K is determined based on the quantity present in the first information; Based on service quality evaluation information, the first cluster center point was calculated using the density method. The service quality evaluation information is a set S, which includes {S1, S2, ..., S}. n }, S of each third element i The formula for calculating the density value is as follows: p i =∑x(d ij -d c ) Where, d ij For S i and S j Euclidean distance, d c Let S be the average distance between the sample sets, and x be (d ij -d c When ) is less than zero, it is zero, and x is (d ij -d c When the value is greater than zero, it is one; d c The calculation formula is as follows: Based on the service quality evaluation information, K-1 cluster centers are calculated using a preset formula, and each cluster center corresponds to a class. The default formula is: dist i =max{min(d i1 ,d i2 ,…,d in ,)}(i=1,2…n) Double inductive extraction: Calculate the weighted distance from each third element in the service quality evaluation information to each cluster center point in turn, and assign each third element to the class corresponding to the cluster center point with the smallest weighted distance; calculate the mean of the third elements in each class, and use the mean to update the cluster center point corresponding to the class.

2. The service quality evaluation extraction method according to claim 1, characterized in that... The step of performing coarse-grained extraction and calculation on the first information to obtain the second information includes: The first information is preprocessed to obtain a vectorized sample set; The second information is obtained by performing coarse-grained sentiment evaluation on the sample set.

3. The service quality evaluation extraction method according to claim 2, characterized in that... The step of performing coarse-grained sentiment evaluation extraction on the sample set to obtain the second information includes: The sample set is classified by part-of-speech tagging to obtain the sample set after part-of-speech tagging; The subject-predicate relationship of the sample set after part-of-speech classification is extracted to obtain a sentiment dataset. The sentiment dataset includes a set of service evaluation results and a set of sentiment expressions extracted from all evaluations. Each first element in the set of service evaluation results uniquely corresponds to a second element in the set of sentiment expressions. The sentiment dataset is transformed and its dimensionality reduced to obtain the second information.

4. The service quality evaluation extraction method according to claim 1, characterized in that... The process of performing phase error correction matching on the finely extracted sentiment expression set to obtain a service quality evaluation set includes: The number of elements in the fine-grained extracted sentiment expression set is extracted to obtain the total number of the first element in the fine-grained extracted service evaluation result set and the total number of the second element in the fine-grained extracted sentiment expression set. If the total number of the first element in the service evaluation result set after fine-grained extraction is equal to the total number of the second element in the sentiment expression set after fine-grained extraction, then the evaluation result set is matched and extracted to obtain the sentiment expression set after phase correction. If the total number of the first element in the fine-grained extracted evaluation result set is greater than the total number of the second element in the fine-grained extracted emotion expression set, then error correction matching calculation is performed based on the fine-grained extracted evaluation result set and the emotion expression set to obtain the error-corrected emotion expression set. If the total number of elements in the evaluation result set after fine-grained extraction is less than the total number of elements in the sentiment expression set after fine-grained extraction, then error correction matching calculation is performed based on the sentiment expression set after fine-grained extraction and the evaluation result set to obtain the sentiment expression set after error correction.

5. The service quality evaluation extraction method according to claim 4, characterized in that... The step of performing error correction matching calculations based on the finely extracted evaluation result set and the sentiment expression set to obtain the error-corrected sentiment expression set includes: For each first element in the evaluation result set after fine-grained extraction, the minimum text matching distance method is used to find the corresponding sentiment expression of each element in the sentiment expression set. Each first element and the corresponding sentiment expression in the set of evaluation results after fine-grained extraction are denoted as the third element; All the aforementioned third elements are denoted as the set of emotion expressions after phase correction.

6. A service quality evaluation extraction device, characterized in that, include: The first acquisition unit is used to acquire first information, which includes at least one service quality evaluation data of the customer regarding the business. A coarse-grained extraction unit is used to perform coarse-grained extraction calculations on the first information to obtain the second information. The second information includes a dimension-reduced sentiment dataset, which includes a set of service evaluation results extracted from all evaluations and a set of sentiment expressions. The fine-grained extraction unit is used to perform fine-grained extraction and inductive calculation on the second information to obtain service quality evaluation information. Each third element in the service quality evaluation set includes a service evaluation and the sentiment expression corresponding to the service evaluation. The double induction unit is used to perform double induction on the service quality evaluation information to obtain the induced service quality evaluation information. The fine-grained extraction unit includes: Model building unit, used to build LSTM mathematical models; The second granularity unit is used to take the second information as input information of the LSTM mathematical model, solve the LSTM mathematical model to obtain the third information, the third information being the sentiment expression set after fine-grained extraction of the sentiment results from the second information, the sentiment expression set including the service evaluation result set after fine-grained extraction and the sentiment expression set after fine-grained extraction; A correction matching unit is used to perform phase error correction matching on the fine-grained extracted sentiment expression set to obtain a service quality evaluation set. Each third element in the service quality evaluation set includes a service evaluation and the sentiment expression corresponding to the service evaluation. Also includes: Obtain a preset experience value K; the experience value K is determined based on the quantity present in the first information; Based on service quality evaluation information, the first cluster center point was calculated using the density method. The service quality evaluation information is a set S, which includes {S1, S2, ..., S}. n }, S of each third element i The formula for calculating the density value is as follows: p i =∑x(d ij -d c ) Where, d ij For S i and S j Euclidean distance, d c Let S be the average distance between the sample sets, and x be (d ij -d c When ) is less than zero, it is zero, and x is (d ij -d c When the value is greater than zero, it is one; d c The calculation formula is as follows: Based on the service quality evaluation information, K-1 cluster centers are calculated using a preset formula, and each cluster center corresponds to a class. The default formula is: dist i =max{min(d i1 ,d i2 ,…,d in ,)}(i=1,2…n) Double inductive extraction: Calculate the weighted distance from each third element in the service quality evaluation information to each cluster center point in turn, and assign each third element to the class corresponding to the cluster center point with the smallest weighted distance; calculate the mean of the third elements in each class, and use the mean to update the cluster center point corresponding to the class.

7. The service quality evaluation extraction device according to claim 6, characterized in that, The coarse-grained extraction unit includes: A preprocessing unit is used to preprocess the first information to obtain a vectorized sample set; The first granularity unit is used to perform coarse-grained extraction of sentiment evaluation on the sample set to obtain the second information.

8. The service quality evaluation extraction device according to claim 7, characterized in that, The first granularity unit includes: A classification unit is used to perform part-of-speech classification on the sample set to obtain the part-of-speech classified sample set; An extraction unit is used to extract the sentiment of the subject-predicate relationship from the sample set after part-of-speech classification to obtain a sentiment dataset. The sentiment dataset includes a set of service evaluation results extracted from all evaluations and a set of sentiment expressions. Each first element in the set of service evaluation results uniquely corresponds to a second element in the set of sentiment expressions. The transformation unit is used to transform and reduce the dimensionality of the sentiment dataset to obtain the second information.

9. The service quality evaluation extraction device according to claim 6, characterized in that, The correction matching unit includes: The quantity extraction unit is used to extract the number of elements from the fine-grained extracted sentiment expression set to obtain the total number of the first element in the fine-grained extracted service evaluation result set and the total number of the second element in the fine-grained extracted sentiment expression set. The first judgment unit is used to match and extract the evaluation result set to obtain the sentiment expression set after phase correction if the total number of the first elements in the service evaluation result set after fine-grained extraction is equal to the total number of the second elements in the sentiment expression set after fine-grained extraction. The second judgment unit is used to perform error correction matching calculation based on the fine-grained extracted evaluation result set and the emotion expression set if the total number of the first elements in the fine-grained extracted evaluation result set is greater than the total number of the second elements in the fine-grained extracted emotion expression set, so as to obtain the error-corrected emotion expression set. The third judgment unit is used to perform error correction matching calculation based on the fine-grained extracted emotion expression set and the evaluation result set if the total number of elements in the fine-grained extracted evaluation result set is less than the total number of elements in the fine-grained extracted emotion expression set, so as to obtain the error-corrected emotion expression set.

10. The service quality evaluation extraction device according to claim 9, characterized in that, The second determination unit includes: The traversal unit is used to find the sentiment expression corresponding to each element in the sentiment expression set for each first element in the fine-grained extraction of the evaluation result set using the minimum text matching distance method. A pairing unit is used to denote each first element and the sentiment expression corresponding to each first element in the finely extracted evaluation result set as the third element. A labeling unit is used to denote all the third elements as the set of emotion expressions after phase error correction.

11. A service quality evaluation extraction device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the service quality evaluation extraction method as described in any one of claims 1 to 5 when executing the computer program.

12. A readable storage medium, characterized in that: The readable storage medium stores a computer program that, when executed by a processor, implements the steps of the service quality evaluation extraction method as described in any one of claims 1 to 5.

Images