Webpage quality evaluation method and device, readable storage medium and program product

By obtaining the results of multiple sets of web page dialing and using self-attention model to integrate the correlation of multiple features, the one-sidedness and inaccuracy of web page quality assessment in the prior art is solved, and a more accurate and comprehensive web page quality assessment is achieved.

CN118964771APending Publication Date: 2024-11-15CHINA MOBILE GROUP SHANDONG +1

Patent Information

Application Number
CN202411024898.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing web page quality evaluation methods cannot accurately evaluate web page quality and cannot truly reflect users' subjective perceptions, which are mainly due to the evaluation of single indicator parameters, which lead to one-sidedness and inaccuracy.

Method used

By obtaining the results of multiple sets of web page dialing, using a preset evaluation algorithm to determine the evaluation vector, and splicing multiple indicator parameters with the evaluation vector into a feature matrix, input a self-attention model to determine the web page quality evaluation score, and integrate it with the self-attention mechanism. The correlation of multiple features improves the comprehensiveness and accuracy of the assessment.

Benefits of technology

It effectively improves the comprehensiveness and accuracy of web page quality evaluation, avoids the one-sidedness and inaccuracy of single indicator parameter evaluation, and truly reflects the user's subjective perception.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964771A_ABST
    Figure CN118964771A_ABST
Patent Text Reader

Abstract

The invention discloses a webpage quality evaluation method and device, a readable storage medium and a program product, and the method comprises the steps: obtaining a plurality of groups of dial test results of target webpage quality evaluation, one group of dial test results comprising a plurality of different index parameters; determining an evaluation vector of the target webpage quality evaluation corresponding to each group of dial test results based on a preset evaluation algorithm; splicing a plurality of index parameters included in each group of dial test results with the evaluation vectors corresponding to each group of dial test results to obtain a spliced feature matrix; inputting the splicing feature matrix into a self-attention model, so as to determine a quality evaluation score of the target webpage on the basis of relevancy among features in the splicing feature matrix and the evaluation vector; and evaluating the quality of the target webpage based on the quality evaluation score of the target webpage.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a webpage quality assessment method and device, a readable storage medium, and a program product. Background Art

[0002] At present, the main method of evaluating web page quality is to use an automatic dialing system to simulate web page access actions by actively dialing web pages, and record and analyze the dialing results to achieve web page quality evaluation. The results of automatic web page dialing contain a series of web page quality index parameters. Each index parameter is used to quantitatively evaluate the quality of network communication at a certain stage of web page access, and can be used to evaluate the subjective feelings of users when accessing web pages. Specifically, a threshold is set for each index parameter. When any index parameter exceeds the set threshold, the web page is judged to be a web page of poor quality. The index parameters of poor quality web pages are optimized and improved to improve web page quality and user access perception.

[0003] However, the web page quality assessment method based on automatic dialing technology can only make a simple judgment through the degradation of a single indicator parameter. The size and threshold of the impact of a single indicator parameter on the web page quality cannot be simply determined, resulting in the inability to accurately assess the quality of the web page and the inability to truly reflect the user's subjective perception. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a web page quality assessment method and device, a readable storage medium, and a program product to solve the problem that the existing web page quality assessment method cannot accurately assess the quality of the web page.

[0005] In order to solve the above technical problems, this specification is implemented as follows:

[0006] In a first aspect, a webpage quality assessment method is provided, comprising:

[0007] Obtain multiple groups of dial test results for evaluating the quality of a target web page, where a group of dial test results includes multiple different indicator parameters;

[0008] Determine, based on a preset evaluation algorithm, an evaluation vector for the target webpage quality evaluation corresponding to each group of dialing test results;

[0009] The plurality of indicator parameters included in each group of dialing test results and the evaluation vectors corresponding to each group of dialing test results are respectively concatenated to obtain a concatenated feature matrix;

[0010] Inputting the concatenated feature matrix into a self-attention model to determine a quality evaluation score of the target web page based on the correlation between each feature in the concatenated feature matrix and the evaluation vector;

[0011] The quality of the target webpage is evaluated based on the quality evaluation score of the target webpage.

[0012] Optionally, multiple groups of dial test results for evaluating the quality of the target web page are obtained, including:

[0013] Simulating the access action to the target webpage through a preset automatic dialing test system to obtain multiple groups of dialing test results for the quality evaluation of the target webpage;

[0014] Multiple different indicator parameters for quality assessment are selected from each group of dial test results. The indicator parameters include a combination of multiple items including DNS resolution delay, connection establishment delay, first byte response delay, web page display delay, first screen delay, average web page download rate, web page opening success rate, element introduction rate and website size.

[0015] Optionally, the preset evaluation algorithm includes a least squares method, and determining the evaluation vector of the target webpage quality evaluation corresponding to each group of dialing test results based on the preset evaluation algorithm includes:

[0016] Selecting some groups from the multiple groups of dialing test results;

[0017] Obtaining manual scores of the target webpage quality assessment corresponding to the partial group;

[0018] Performing least squares based on the indicator parameters of the partial groups and the corresponding manual scores to obtain a corresponding least squares solution;

[0019] Based on the least squares solution, an evaluation vector score of each group of dialing test results corresponding to the target webpage quality evaluation is determined.

[0020] Optionally, the self-attention model includes a self-attention layer, a normalized exponential function layer, a feedforward neural network layer and a summation layer connected in sequence,

[0021] Inputting the concatenated feature matrix into a self-attention model to determine a quality evaluation score of the target webpage based on the correlation between the features in the concatenated feature matrix and the evaluation vector includes:

[0022] Inputting the concatenated feature matrix into the self-attention layer to obtain a self-attention encoding matrix;

[0023] Inputting the self-attention encoding matrix into the normalized exponential function layer to calculate the probability of each dimensional feature, so as to obtain a probability matrix representing the correlation between each feature in the concatenated feature matrix;

[0024] Inputting the probability matrix into the feedforward neural network layer to obtain a feedforward vector encoding matrix;

[0025] The column vector corresponding to the evaluation vector in the feedforward vector encoding matrix is ​​extracted and input into the summation layer together with the evaluation vector to obtain the quality evaluation score output by the self-attention model.

[0026] Optionally, inputting the concatenated feature matrix into the self-attention layer to obtain a self-attention encoding matrix includes:

[0027] Based on the least squares solution, construct a first weight matrix, a second weight matrix, and a third weight matrix of the self-attention layer respectively;

[0028] By cross-multiplying the concatenated feature matrix with the first weight matrix, the second weight matrix, and the third weight matrix, respectively, an input encoding matrix, an input weight matrix, and an output weight matrix of the self-attention layer are obtained;

[0029] Based on the input encoding matrix and the input weight matrix, the encoding matrix of the self-attention layer is obtained.

[0030] Optionally, inputting the probability matrix into the feedforward neural network layer to obtain a feedforward vector encoding matrix includes:

[0031] The probability matrix is ​​input into the feedforward neural network layer and cross-multiplied with the output weight matrix to obtain a feedforward vector encoding matrix.

[0032] Optionally, before inputting the concatenated feature matrix into the self-attention model, the method further includes:

[0033] The self-attention model is trained using the concatenated feature matrices corresponding to the dialing test results of different web pages as samples and the quality assessment scores corresponding to the dialing test results as labels;

[0034] Based on the loss between the quality assessment score output by the summation layer of the self-attention model and the corresponding label, the parameters of the self-attention model are adjusted until the loss converges to train the self-attention model.

[0035] Optionally, the self-attention model further includes a fully connected layer and a hyperparameter loss function,

[0036] The step of adjusting the parameters of the self-attention model based on the loss between the quality assessment score output by the sum layer of the self-attention model and the corresponding label until the loss converges to train the self-attention model comprises:

[0037] Inputting the quality assessment score output by the summation layer into the fully connected layer to normalize the quality assessment score;

[0038] Inputting the normalized quality assessment score into the hyperparameter loss function to output a final quality assessment score;

[0039] The parameters of the self-attention model are adjusted until the hyperparameter loss function reaches a preset loss threshold to train the self-attention model.

[0040] In a second aspect, a web page quality assessment device is provided, comprising a processor and a memory, wherein the memory stores a program or instruction that can be executed on the processor, and when the program or instruction is executed by the processor, the steps of the method described in the first aspect are implemented.

[0041] In a third aspect, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented.

[0042] According to a fourth aspect, a computer program product is provided, the computer program product comprising a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a computer to execute the steps of the method according to the first aspect.

[0043] In an embodiment of the present application, multiple groups of dialing test results for evaluating the quality of a target web page are obtained, wherein one group of dialing test results includes multiple different indicator parameters; an evaluation vector for the quality evaluation of the target web page corresponding to each group of dialing test results is determined based on a preset evaluation algorithm; the multiple indicator parameters included in each group of dialing test results and the evaluation vector corresponding to each group of dialing test results are respectively spliced ​​to obtain a spliced ​​feature matrix; the spliced ​​feature matrix is ​​input into a self-attention model to determine the quality evaluation score of the target web page based on the correlation between each feature in the spliced ​​feature matrix and the evaluation vector; the quality of the target web page is evaluated based on the quality evaluation score of the target web page, thereby utilizing the self-attention mechanism to determine the correlation of each feature in a multi-feature fusion matrix constructed by multiple indicator parameters, and by adding the mutual influence relationship between each feature, the comprehensiveness and accuracy of the web page quality evaluation is effectively improved, thereby avoiding the one-sidedness and inaccuracy of evaluating the web page quality with a single indicator parameter in the traditional web page quality evaluation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0045] Figure 1 This is one of the flow charts of the webpage quality assessment method according to the embodiment of the present application.

[0046] Figure 2This is the second flow chart of the webpage quality assessment method according to the embodiment of the present application.

[0047] Figure 3 This is the third flow chart of the webpage quality assessment method according to the embodiment of the present application.

[0048] Figure 4 It is a structural block diagram of a webpage quality assessment device according to an embodiment of the present application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application. The numbering of the drawings in this application is only used to distinguish the various steps in the scheme, and is not used to limit the execution order of the various steps. The specific execution order is subject to the description in the specification.

[0050] In order to solve the problems existing in the prior art, the present application provides a webpage quality assessment method, such as Figure 1 As shown, the following steps 102 to 110 are included.

[0051] Step 102: obtaining multiple groups of dialing test results for evaluating the quality of the target webpage, wherein one group of dialing test results includes multiple different indicator parameters.

[0052] The dialing test result can be obtained by using an automatic dialing test system to actively dial the target web page by simulating the web page access action, thereby obtaining a dialing test result containing a series of web page quality index parameters. Each index parameter is used to quantitatively evaluate the quality of network communication at a certain stage of the target web page access, such as Domain Name System (DNS) resolution delay, connection establishment delay, first byte response delay, web page display delay, first screen delay, average web page download rate, web page opening success rate, element introduction rate, website size, etc.

[0053] By deploying web page test tasks on the automatic dialing system, the pre-defined web page test cases are published to the automatic dialing system, and the start and end time of the dialing test, the number of dialing rounds, the type of monitoring indicators, etc. are set. According to the set start and end time, web page test cases and other parameters, the user is simulated to automatically execute the dialing task, and the dialing results and indicator parameters are recorded. A set of dialing results includes multiple different indicator parameters, and a set of dialing results corresponds to a round of dialing. If multiple rounds of automatic dialing are performed, multiple sets of dialing results will be obtained accordingly. The dialing results of the automatic dialing system can be stored in the management platform of the automatic dialing system. By obtaining the dialing results corresponding to the target web page from the management platform, it can be used for quality analysis and evaluation of the target web page.

[0054] Step 104: determining an evaluation vector for evaluating the quality of the target webpage corresponding to each group of dialing test results based on a preset evaluation algorithm.

[0055] Specifically, the preset evaluation algorithm includes a least squares method, and determining the evaluation vector of each group of dialing test results corresponding to the target web page quality evaluation based on the preset evaluation algorithm includes: selecting some groups from the multiple groups of dialing test results; obtaining manual scores of the target web page quality evaluation corresponding to the some groups; performing least squares based on the indicator parameters of the some groups and the corresponding manual scores to obtain the corresponding least squares solution; and determining the evaluation vector score of each group of dialing test results corresponding to the target web page quality evaluation based on the least squares solution.

[0056] From multiple groups of dial test results of the target web page, select some groups of dial test results, for example, if there are 100 groups of dial test results for the target web page, select 10 of them. At the same time, combine the indicator parameters in the selected dial test results to manually score the quality of the target web page. For example, the web page quality score is set to 10. To avoid the subjectivity of manual scoring, multiple people score the dial test results of the same group, remove the highest score and the lowest score, and take the average as the manual score for the target web page quality assessment.

[0057] Then, the feature matrix F corresponding to the exponential parameters of each group of dial test results is used t1×d The feature matrix S corresponding to the manual scores of each group of dial test results t1×1 , perform the least squares solution:

[0058] F×W=S

[0059]

[0060] The least squares solution W is:

[0061] W=(F T F) -1 F T S

[0062] Where d represents the number or dimension of indicator parameters included in each group of dial test results, t1 represents the number of selected groups, and f t1d Indicates the characteristic value of the dth indicator parameter in the t1th group of dial test results; s t11 The feature value representing the manual scoring of the dial test results of group t1.

[0063] Then, for the characteristic matrix F corresponding to the index parameters of the multiple groups of dialing test results of the target webpage obtained in step 102, t×d and the least squares solution W d×1 According to the formula F×W=S, we can get the preliminary estimated quality scores of each group of dialing test results of the target web page, that is, the characteristic matrix S of the evaluation vector of the target web page quality evaluation t×1 , where t represents the number of groups of dial test results of the target web page obtained.

[0064] Step 106 , concatenating the multiple indicator parameters included in each group of dialing test results and the evaluation vectors corresponding to each group of dialing test results to obtain a concatenated feature matrix.

[0065] The characteristic matrix F corresponding to the index parameters of multiple groups of dial-up test results of the target web page t×d and the characteristic matrix S of the evaluation vector t×1 Perform matrix concatenation to obtain the concatenated feature matrix E t×(d+1) .

[0066] Combination Figure 2 , normalize the corresponding indicator parameters of the target web page, including the first byte response delay, first screen delay, web page opening success rate, website size, etc., and obtain the feature matrix F corresponding to the indicator parameters t×d Based on the preliminary estimated quality scores of each group of dial-up test results of the target web page, the characteristic matrix S of the evaluation vector of the target web page quality evaluation is obtained. t×1 , the feature matrix F t×d And the feature matrix S t×1 Concatenate to get the concatenated feature matrix E t×(d+1) .

[0067] Step 108: Input the concatenated feature matrix into a self-attention model to determine the quality evaluation score of the target web page based on the correlation between the features in the concatenated feature matrix and the evaluation vector.

[0068] Specifically, the self-attention model includes a self-attention layer, a normalized exponential function layer, a feedforward neural network layer and a summation layer connected in sequence. The concatenated feature matrix is ​​input into the self-attention model to determine the quality evaluation score of the target web page based on the correlation between each feature in the concatenated feature matrix and the evaluation vector, including: inputting the concatenated feature matrix into the self-attention layer to obtain a self-attention encoding matrix; inputting the self-attention encoding matrix into the normalized exponential function layer to calculate the probability of each dimensional feature to obtain a probability matrix representing the correlation between each feature in the concatenated feature matrix; inputting the probability matrix into the feedforward neural network layer to obtain a feedforward vector encoding matrix; extracting the column vector corresponding to the evaluation vector in the feedforward vector encoding matrix and inputting it into the summation layer with the evaluation vector to obtain the quality evaluation score output by the self-attention model.

[0069] Combination Figure 2 The self-attention model includes a Transformer module, which includes a self-attention layer 41, a normalized exponential function layer 42, a feedforward neural network layer 43 and a summation layer 44 connected in sequence.

[0070] The specific architecture and working principle of the Transformer module are as follows:

[0071] The concatenation matrix E t×(d+1) The input Transformer module is used for feature encoding, and then passes through the self-attention layer 41, the normalized exponential function layer 42, the feedforward neural network layer 43 and the summation layer 44 to obtain the output feature matrix O t×(d+1) From the output feature matrix O t×(d+1) Extract the column vector corresponding to the evaluation vector, that is, the d+1th column vector, and compare it with the characteristic matrix S of the evaluation vector t×1 The input is summed up in the summation layer 44 to obtain the quality assessment score output by the Transformer module.

[0072] Specifically, the step of inputting the concatenated feature matrix into the self-attention layer to obtain a self-attention encoding matrix includes: constructing a first weight matrix Wq, a second weight matrix Wk, and a third weight matrix Wv of the self-attention layer based on the least squares solution; obtaining an input encoding matrix Q, an input weight matrix K, and an output weight matrix V of the self-attention layer by cross-multiplying the concatenated feature matrix with the first weight matrix Wq, the second weight matrix Wk, and the third weight matrix Wv; and obtaining an encoding matrix A of the self-attention layer based on the input encoding matrix Q and the input weight matrix K.

[0073] The step of inputting the probability matrix into the feedforward neural network layer to obtain a feedforward vector encoding matrix includes: inputting the probability matrix into the feedforward neural network layer and performing cross multiplication with the output weight matrix V to obtain a feedforward vector encoding matrix.

[0074] The encoding process inside each Transformer module is as follows:

[0075] (1) Construct the weight matrix Wq (d+1)×(d+1) , and the concatenated feature matrix E of the input t×(d+1) Perform cross multiplication to obtain the input encoding matrix Q t×(d+1) .

[0076] (2) Construct the weight matrix Wk (d+1)×(d+1) , and the concatenated feature matrix E of the input t×(d+1) Perform cross multiplication to obtain the input weight matrix K t×(d+1) .

[0077] (3) Construct the weight matrix Wv (d+1)×(d+1) , and the concatenated feature matrix E of the input t×(d+1) Perform cross multiplication to obtain the output weight matrix V t×(d+1) .

[0078] (4) Calculate the self-attention encoding matrix A by inputting the encoding matrix Q and the input weight matrix K t×t ,in,

[0079] (5) The self-attention encoding matrix A is input into the normalized exponential function layer 42, such as the Softmax layer, to calculate the probability of each dimension feature and obtain the probability matrix P t×t , to simulate the contribution of each eigenvector to the current eigenvalue. The higher the contribution, the higher the correlation between the eigenvalues ​​and the greater the dependency.

[0080] (6) The probability matrix P output by the normalized exponential function layer 42 is t×t Input to the feedforward neural network layer 43, cross-multiplied to the output weight matrix V in (3) t×(d+1) In the above equation, we get the feedforward vector encoding matrix O t×(d+1) , thus obtaining the output feature matrix O of the Transformer module t×(d+1) :

[0081] O = Softmax(A) × V

[0082] In one embodiment, the self-attention model includes multiple series-connected Transformer modules. t×(d+1), input to the next series of Transformer modules, enhance the encoding depth, until the output feature matrix O of the last Transformer module is obtained t×(d+1) .

[0083] The output feature matrix O t×(d+1) The d+1th column vector Od+1 t×1 and the feature matrix S of the web page quality evaluation vector t×1 , are added element by element at the summation layer 44, whereby the self-attention model outputs a quality assessment score of the target web page.

[0084] Step 110: Evaluate the quality of the target webpage based on the quality evaluation score of the target webpage.

[0085] The web page quality evaluation score obtained according to the above steps can more accurately evaluate the quality of the web page and truly reflect the user's subjective perception.

[0086] Based on the solution provided by the above embodiment, optionally, before inputting the spliced ​​feature matrix into the self-attention model, it also includes: training the self-attention model with the spliced ​​feature matrices corresponding to the dialing results of different web pages as samples and the quality assessment scores corresponding to the dialing results as labels; based on the loss between the quality assessment scores output by the summation layer of the self-attention model and the corresponding labels, adjusting the parameters of the self-attention model until the loss converges to train the self-attention model.

[0087] During training, the dialing test results of different web pages are obtained, and the evaluation vectors of the corresponding quality evaluation are obtained based on the least squares method, and then spliced ​​to obtain the spliced ​​feature matrix. The spliced ​​feature matrices of each web page are used as samples, and the labels are the quality evaluation scores of each web page, which are obtained by manual evaluation. The self-attention model is trained based on the samples and labels, and the loss between the quality evaluation score output by the summation layer of the self-attention model and the corresponding label is calculated. The parameters of the self-attention model are adjusted based on the loss until the training ends when the loss converges. The obtained self-attention model can be used to predict the evaluation score of web page quality.

[0088] The least squares method is used as the preliminary estimated quality score of the web page quality of each sample. By splicing the preliminary estimated quality score into the splicing feature matrix as an input of the model, the convergence speed of the self-attention model training can be accelerated.

[0089] The residual value between the webpage quality evaluation score and the evaluation vector output by Transformer training is the loss Od+1. t×1 , and adding the evaluation vector of the web page quality evaluation to the residual value can effectively improve the accuracy of web page quality evaluation.

[0090] Combination Figure 2 The summation layer 44 outputs the quality evaluation score of the web page, which is usually scored on a 10-point scale, and the web page quality score is [0,10]. According to the web page quality score statistics, the quality difference of web pages with scores in [0,1] is not obvious, and they have all been seriously degraded; the quality difference of web pages with scores in [9,10] is also not obvious, and they can all be browsed smoothly.

[0091] According to the web page quality scoring characteristics, in one embodiment, the self-attention model also includes a fully connected layer and a hyperparameter loss function, and the loss between the quality assessment score output by the summation layer of the self-attention model and the corresponding label is used to adjust the parameters of the self-attention model until the loss converges to train the self-attention model, including: inputting the quality assessment score output by the summation layer into the fully connected layer to normalize the quality assessment score; inputting the normalized quality assessment score into the hyperparameter loss function to output a final quality assessment score; and adjusting the parameters of the self-attention model until the hyperparameter loss function reaches a preset loss threshold to train the self-attention model.

[0092] In the above embodiment, the output feature matrix O t×(d+1) The d+1th column vector Od+1 t×1 and the evaluation vector S of web page quality assessment t×1 Add them element by element and input them into the fully connected layer to get the normalized quality assessment score O t t ×1 .

[0093] Design a hyperparameter function similar to the Sigmoid activation function as the loss function of the self-attention model, and define the hyperparameter loss function as:

[0094]

[0095] According to the model experimental data, in the embodiment of the present application, for example, the hyperparameter λ is set to 10, the hyperparameter μ is set to 33, and the hyperparameter σ is set to 0.7, then the hyperparameter loss function is:

[0096]

[0097] Then, O t ′ ×1 Input the hyperparameter loss function L(x) to obtain the final quality assessment score of the web page. t×1 :

[0098] Score = L(O ′ )

[0099] During the training process of the self-attention model, the training ends when the hyperparameter loss function value output after multiple cycles of iterations reaches the minimum, and the obtained self-attention model can be used to predict the evaluation score of web page quality.

[0100] According to the characteristics of web page quality rating, a hyperparameter loss function is added to the self-attention model to make the rating results of web page quality assessment closer to the user's actual visit perception.

[0101] Next, combine Figure 3 , an example of a web page quality assessment method according to an embodiment of the present application is described.

[0102] like Figure 3 As shown, the following steps are included:

[0103] Step 202, obtaining the dialing result index parameters of the target webpage, performing normalization processing to form a feature matrix, and obtaining a feature matrix of the index parameters;

[0104] Step 204, calculating the least squares estimated score of the target webpage quality, obtaining an estimated score feature vector, and concatenating it with the feature matrix of the indicator parameters;

[0105] Step 206, input the concatenated feature matrix into N Transformer modules for feature encoding to obtain an output feature matrix;

[0106] Step 208, sum the d+1th column vector of the output feature matrix and the estimated score feature vector, input into the fully connected layer and the hyperparameter loss function, and obtain the evaluation score of the target web page quality.

[0107] In an embodiment of the present application, multiple groups of dialing test results for evaluating the quality of a target web page are obtained, wherein one group of dialing test results includes multiple different indicator parameters; an evaluation vector for the quality evaluation of the target web page corresponding to each group of dialing test results is determined based on a preset evaluation algorithm; the multiple indicator parameters included in each group of dialing test results and the evaluation vector corresponding to each group of dialing test results are respectively spliced ​​to obtain a spliced ​​feature matrix; the spliced ​​feature matrix is ​​input into a self-attention model to determine the quality evaluation score of the target web page based on the correlation between each feature in the spliced ​​feature matrix and the evaluation vector; the quality of the target web page is evaluated based on the quality evaluation score of the target web page, thereby utilizing the self-attention mechanism to determine the correlation of each feature in a multi-feature fusion matrix constructed by multiple indicator parameters, and by adding the mutual influence relationship between each feature, the comprehensiveness and accuracy of the web page quality evaluation is effectively improved, thereby avoiding the one-sidedness and inaccuracy of evaluating the web page quality with a single indicator parameter in the traditional web page quality evaluation method.

[0108] Alternatively, if Figure 4As shown, the embodiment of the present application also provides a web page quality assessment device 2000, including a processor 2400 and a memory 2200, and the memory 2200 stores a program or instruction that can be executed on the processor 2400. When the program or instruction is executed by the processor 2400, the various steps of the above-mentioned web page quality assessment method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0109] The embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, each process of any of the above-mentioned webpage quality assessment method embodiments is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. Among them, the readable storage medium includes a computer-readable storage medium, such as a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0110] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0111] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0112] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A webpage quality assessment method, characterized in that: include: Obtain multiple groups of dial test results for evaluating the quality of a target web page, where a group of dial test results includes multiple different indicator parameters; Determine, based on a preset evaluation algorithm, an evaluation vector for the target webpage quality evaluation corresponding to each group of dialing test results; The plurality of indicator parameters included in each group of dialing test results and the evaluation vectors corresponding to each group of dialing test results are respectively concatenated to obtain a concatenated feature matrix; Inputting the concatenated feature matrix into a self-attention model to determine a quality evaluation score of the target web page based on the correlation between each feature in the concatenated feature matrix and the evaluation vector; The quality of the target webpage is evaluated based on the quality evaluation score of the target webpage.

2. The method according to claim 1, characterized in that: Get multiple sets of test results for target web page quality assessment, including: Simulating the access action to the target webpage through a preset automatic dialing test system to obtain multiple groups of dialing test results for the quality evaluation of the target webpage; Multiple different indicator parameters for quality assessment are selected from each group of dial test results, including a combination of multiple items including domain name system DNS resolution delay, connection establishment delay, first byte response delay, web page display delay, first screen delay, average web page download rate, web page opening success rate, element introduction rate and website size.

3. The method according to claim 1, characterized in that The preset evaluation algorithm includes the least square method, and the step of determining the evaluation vector of the target webpage quality evaluation corresponding to each group of dialing test results based on the preset evaluation algorithm includes: Selecting some groups from the multiple groups of dialing test results; Obtaining manual scores of the target webpage quality assessment corresponding to the partial group; Performing least squares based on the indicator parameters of the partial groups and the corresponding manual scores to obtain a corresponding least squares solution; Based on the least squares solution, an evaluation vector score of each group of dialing test results corresponding to the target webpage quality evaluation is determined.

4. The method according to claim 3, characterized in that: The self-attention model includes a self-attention layer, a normalized exponential function layer, a feedforward neural network layer and a summation layer connected in sequence. Inputting the concatenated feature matrix into a self-attention model to determine a quality evaluation score of the target webpage based on the correlation between the features in the concatenated feature matrix and the evaluation vector includes: Inputting the concatenated feature matrix into the self-attention layer to obtain a self-attention encoding matrix; Inputting the self-attention encoding matrix into the normalized exponential function layer to calculate the probability of each dimensional feature, so as to obtain a probability matrix representing the correlation between each feature in the concatenated feature matrix; Inputting the probability matrix into the feedforward neural network layer to obtain a feedforward vector encoding matrix; The column vector corresponding to the evaluation vector in the feedforward vector encoding matrix is ​​extracted and input into the summation layer together with the evaluation vector to obtain the quality evaluation score output by the self-attention model.

5. The method according to claim 4, characterized in that The step of inputting the concatenated feature matrix into the self-attention layer to obtain a self-attention encoding matrix includes: Based on the least squares solution, construct a first weight matrix, a second weight matrix, and a third weight matrix of the self-attention layer respectively; By cross-multiplying the concatenated feature matrix with the first weight matrix, the second weight matrix, and the third weight matrix, respectively, an input encoding matrix, an input weight matrix, and an output weight matrix of the self-attention layer are obtained; Based on the input encoding matrix and the input weight matrix, the encoding matrix of the self-attention layer is obtained.

6. The method according to claim 5, characterized in that The step of inputting the probability matrix into the feedforward neural network layer to obtain a feedforward vector encoding matrix comprises: The probability matrix is ​​input into the feedforward neural network layer and cross-multiplied with the output weight matrix to obtain a feedforward vector encoding matrix.

7. The method according to any one of claims 4 to 6, characterized in that: Before inputting the concatenated feature matrix into the self-attention model, the method further includes: The self-attention model is trained using the concatenated feature matrices corresponding to the dialing test results of different web pages as samples and the quality assessment scores corresponding to the dialing test results as labels; Based on the loss between the quality assessment score output by the summation layer of the self-attention model and the corresponding label, the parameters of the self-attention model are adjusted until the loss converges to train the self-attention model.

8. The method according to claim 7, characterized in that The self-attention model also includes a fully connected layer and a hyperparameter loss function. The step of adjusting the parameters of the self-attention model based on the loss between the quality assessment score output by the sum layer of the self-attention model and the corresponding label until the loss converges to train the self-attention model comprises: Inputting the quality assessment score output by the summation layer into the fully connected layer to normalize the quality assessment score; Inputting the normalized quality assessment score into the hyperparameter loss function to output a final quality assessment score; The parameters of the self-attention model are adjusted until the hyperparameter loss function reaches a preset loss threshold to train the self-attention model.

9. A webpage quality assessment device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer program product, characterized in that The computer program product comprises a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a computer to execute the steps of the method according to any one of claims 1 to 8.

Citation Information

Patent Citations

  • Image-text content quality identification method and device based on artificial intelligence

    CN113407663A

  • Webpage quality evaluation method, neural network training method, device and equipment

    CN115130041A

  • Webpage quality evaluation method and device, electronic equipment and storage medium

    CN117009705A

  • Website scoring system

    US20200151187A1

Cited By

  • Website performance evaluation method based on multi-modal data fusion and AI deep learning

    CN122243266A

  • Website performance evaluation method based on multi-modal data fusion and AI deep learning

    CN122243266B