CN112613775A - Resource quality evaluation method and device, electronic device and storage medium - Google Patents
Resource quality evaluation method and device, electronic device and storage medium Download PDFInfo
- Publication number
- CN112613775A CN112613775A CN202011582561.9A CN202011582561A CN112613775A CN 112613775 A CN112613775 A CN 112613775A CN 202011582561 A CN202011582561 A CN 202011582561A CN 112613775 A CN112613775 A CN 112613775A
- Authority
- CN
- China
- Prior art keywords
- resource
- quality
- quality evaluation
- evaluated
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013441 quality evaluation Methods 0.000 title claims abstract description 396
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000003860 storage Methods 0.000 title claims abstract description 13
- 230000004927 fusion Effects 0.000 claims description 25
- 238000000605 extraction Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 14
- 238000009826 distribution Methods 0.000 claims description 14
- 230000015556 catabolic process Effects 0.000 claims description 10
- 238000006731 degradation reaction Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 7
- 238000012163 sequencing technique Methods 0.000 claims description 5
- 238000001303 quality assessment method Methods 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 description 25
- 238000010586 diagram Methods 0.000 description 11
- 238000004458 analytical method Methods 0.000 description 3
- 238000012015 optical character recognition Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005056 compaction Methods 0.000 description 1
- 230000009193 crawling Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
Abstract
A resource quality evaluation method and device, electronic equipment and storage medium are provided, the method comprises: performing quality evaluation on resource content matched with each quality evaluation model in the resource to be evaluated through N quality evaluation models corresponding to the resource to be evaluated to obtain N quality scores, wherein N is a positive integer greater than or equal to 2; and fusing the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated. By implementing the embodiment of the application, the image-text resources can be comprehensively evaluated through the multiple quality evaluation models, and the accuracy and reliability of resource quality evaluation are improved.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for evaluating resource quality, an electronic device, and a storage medium.
Background
At present, resources such as pictures and characters in the internet are too expensive as the sea, but not all the resources have good quality and are suitable for being searched, displayed or stored. However, in practice, it is found that when the image-text resources are evaluated, the image-text resources are often evaluated by using artificially established rules, so that the quality evaluation results have the problems of strong subjectivity and irregularity, and the evaluation results matched with the actual quality of the image-text resources cannot be accurately obtained, thereby reducing the accuracy and reliability of resource quality evaluation.
Disclosure of Invention
The embodiment of the application discloses a resource quality evaluation method and device, electronic equipment and a storage medium, which can comprehensively evaluate image-text resources through a plurality of quality evaluation models and improve the accuracy and reliability of resource quality evaluation.
A first aspect of an embodiment of the present application discloses a resource quality evaluation method, including:
performing quality evaluation on resource content matched with each quality evaluation model in the resources to be evaluated through N quality evaluation models corresponding to the resources to be evaluated to obtain N quality scores, wherein N is a positive integer greater than or equal to 2;
and fusing the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated.
A second aspect of the embodiments of the present application discloses a resource quality evaluation apparatus, including:
the quality evaluation unit is used for carrying out quality evaluation on resource content matched with each quality evaluation model in the resources to be evaluated through N quality evaluation models corresponding to the resources to be evaluated to obtain N quality scores, wherein N is a positive integer greater than or equal to 2;
and the fusion unit is used for fusing the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated.
A third aspect of the embodiments of the present application discloses an electronic device, including:
a memory storing executable program code;
a processor coupled with the memory;
the processor calls the executable program code stored in the memory to execute all or part of the steps of any one of the resource quality evaluation methods disclosed in the first aspect of the embodiments of the present application.
A fourth aspect of the embodiments of the present application discloses a computer-readable storage medium, which stores a computer program, where the computer program enables a computer to execute all or part of the steps in any one of the resource quality evaluation methods disclosed in the first aspect of the embodiments of the present application.
Compared with the related art, the embodiment of the application has the following beneficial effects:
in the embodiment of the application, N corresponding quality evaluation models (N is a positive integer greater than or equal to 2) can be set for the Internet resources to be evaluated, and through the N quality evaluation models corresponding to the resources to be evaluated, the quality of the resource content matched with each quality evaluation model in the resources to be evaluated can be evaluated respectively to obtain N quality scores; and fusing the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated. Therefore, by implementing the embodiment of the application, the multi-dimensional quality evaluation can be performed on the resource to be evaluated by adopting the plurality of correspondingly matched quality evaluation models, so that the resource content of the resource to be evaluated on each dimension can be subjected to quality evaluation by using the targeted models, and the corresponding quality score is output, so that the accuracy of the resource quality evaluation is improved. The obtained quality scores are fused, so that a comprehensive result after multi-dimensional quality evaluation can be obtained, the bias of evaluation results caused by evaluating the resource quality from only a single dimension (for example, evaluating only the text quality, evaluating only the picture quality and the like) can be avoided, and the reliability of resource quality evaluation is ensured. Therefore, comprehensive evaluation of resources to be evaluated can be achieved through the quality evaluation models, and accuracy and reliability of resource quality evaluation are improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic diagram of a resource quality evaluation framework disclosed in an embodiment of the present application;
fig. 2 is a schematic flowchart of a resource quality evaluation method disclosed in an embodiment of the present application;
fig. 3 is a schematic diagram of an image-text resource disclosed in an embodiment of the present application;
fig. 4 is a schematic flowchart of another resource quality evaluation method disclosed in the embodiment of the present application;
fig. 5 is a schematic flowchart of another resource quality evaluation method disclosed in the embodiment of the present application;
FIG. 6 is a schematic diagram of another resource quality assessment framework disclosed in embodiments of the present application;
fig. 7 is a schematic block diagram of a resource quality evaluation apparatus disclosed in an embodiment of the present application;
FIG. 8 is a schematic block diagram of another resource quality assessment apparatus disclosed in an embodiment of the present application;
fig. 9 is a schematic block diagram of an electronic device according to an embodiment of the present disclosure.
Detailed Description
The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
It should be noted that the terms "comprises" and "comprising," and any variations thereof, in the embodiments of the present application, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The embodiment of the application discloses a resource quality evaluation method and device, electronic equipment and a storage medium, which can comprehensively evaluate image-text resources through a plurality of quality evaluation models and improve the accuracy and reliability of resource quality evaluation.
The following detailed description is made with reference to the accompanying drawings.
Referring to fig. 1, fig. 1 is a schematic diagram of a resource quality evaluation framework disclosed in an embodiment of the present application, and the framework includes a model evaluation module 10 and a fusion module 20, where the model evaluation module 10 may include a plurality of different quality evaluation models 11, each quality evaluation model 11 may be independently used for performing feature extraction and quality evaluation on resource content in a resource to be evaluated, and the fusion module 20 may be used for fusing quality evaluation results of the plurality of quality evaluation models 11 to finally obtain a comprehensive quality evaluation result (such as a quality evaluation score, a quality evaluation grade, and the like) corresponding to the resource to be evaluated. Illustratively, when the resource quality evaluation framework is used for evaluating the quality of the teletext resource, the quality evaluation model 11 may include a model for evaluating the quality of a title of the teletext resource, a model for evaluating the quality of a text content of the teletext resource, a model for evaluating the quality of a picture contained in the teletext resource, and the like. On this basis, the resource content of the resource to be evaluated may be used as the input of the multiple quality evaluation models 11 included in the model evaluation module 10, the output of each quality evaluation model 11 may be used as the input of the fusion module 20, and the fusion module 20 outputs the quality evaluation result corresponding to the resource to be evaluated.
In the related art, when resources in the internet need to be evaluated, a single rule established manually is often adopted for evaluation, or resources to be evaluated are input into a certain quality evaluation model, but the above methods can only evaluate the quality of the resources to be evaluated from a single dimension (for example, only text quality, only picture quality, only overall image quality, and the like). Because the single-dimension evaluation realized by the single model is usually higher in evaluation accuracy only for specific types of resources, when the resources to be evaluated (such as plain text resources, plain picture resources, image-text resources and the like) with complicated types, sizes and sources are faced, it is difficult to accurately evaluate the quality of each resource to be evaluated, and thus the quality evaluation accuracy is easily reduced.
In order to solve the above problem, an embodiment of the present application discloses the above resource quality evaluation framework. In one embodiment, the resource content of the resource to be evaluated may be first evaluated by the model evaluation module 10 of the resource quality evaluation framework. Specifically, the model evaluation module 10 may include N (N is a positive integer greater than or equal to 2) quality evaluation models 11, and each quality evaluation model 11 may be a quality evaluation model corresponding to the resource to be evaluated. Through the N quality evaluation models, the quality of the resource content matched with each quality evaluation model in the resource to be evaluated can be evaluated, and N quality scores can be obtained. Then, the N quality scores may be fused by the fusion module 20 to obtain a quality evaluation result corresponding to the resource to be evaluated. The resource content matched with each quality evaluation model in the resource to be evaluated can be different, and the characteristics extracted from each quality evaluation model can also be different, so that the quality evaluation scores of the resource to be evaluated in multiple dimensions can be obtained in a targeted manner through the multiple quality evaluation models, and then the quality evaluation scores of the multiple dimensions are fused, so that the comprehensive quality evaluation result corresponding to the resource to be evaluated can be obtained, the comprehensive evaluation of the resource to be evaluated is realized, and the accuracy and reliability of the resource quality evaluation are improved.
Referring to fig. 2, fig. 2 is a schematic flow chart of a resource quality evaluation method according to an embodiment of the present application. As shown in fig. 2, the resource quality evaluation method may include the steps of:
202. and performing quality evaluation on the resource content matched with each quality evaluation model in the resource to be evaluated through N quality evaluation models corresponding to the resource to be evaluated to obtain N quality scores, wherein N is a positive integer greater than or equal to 2.
In some embodiments, the resource to be evaluated may include multiple types, such as a plain text resource type, a plain picture resource type, an image resource type, and the like, and each type of resource to be evaluated may include multiple resource contents. The resource type of the resource to be evaluated may be determined by the resource content included in the resource to be evaluated, as shown in fig. 3, if the resource 300 to be evaluated includes the resource contents such as a title 310, a text content 320, a picture content 330, and the like, the resource type of the resource 300 to be evaluated may be an image-text resource type. The title 310 and the text content 320 are both texts, and the picture content 330 may include at least one picture (e.g., a cover picture, a text illustration, etc.). In some embodiments, if the resource 300 to be evaluated only includes the title 310 and the body text content 320, the resource type of the resource 300 to be evaluated may be a plain text resource type; in other embodiments, if the resource 300 to be evaluated only includes a title 310 (e.g., a picture name, a picture summary, a picture tag, etc.) and picture content 330, the resource type of the resource 300 to be evaluated may be a pure picture resource type. It is to be understood that in the above embodiments, the header 310 is not necessary, and the resource 300 to be evaluated in some embodiments may not include the header 310.
The resource quality evaluation method disclosed in the embodiment of the present application may be executed by an electronic device or system such as a server and a terminal device, and is not particularly limited in the embodiment of the present application. By way of example, the following description will be given mainly in terms of servers. In this embodiment of the application, the server may perform quality evaluation on resource content in the resource to be evaluated through N (N is a positive integer greater than or equal to 2) quality evaluation models corresponding to the resource to be evaluated, where the N quality evaluation models may perform quality evaluation on the resource content in the resource to be evaluated from different dimensions. Different resources to be evaluated, the corresponding quality evaluation model types may be the same or different; the number N of corresponding quality evaluation models may be the same or different. Optionally, the N quality evaluation models corresponding to the resource to be evaluated may be different according to differences in resource types, topics, source ways (e.g., from search engine crawling, designated website acquisition, etc.) of the resource to be evaluated. And if the evaluation dimensions corresponding to each quality evaluation model are different, the resource content corresponding to the evaluation dimensions can be obtained from the resource to be evaluated and input into the corresponding quality evaluation model for analysis.
After the quality evaluation is performed on the resource content in the resource to be evaluated through the N quality evaluation models corresponding to the resource to be evaluated, the server may obtain N quality scores corresponding to the resource to be evaluated, that is, each quality evaluation model may output a corresponding quality score for the evaluated resource content. Based on the N quality scores, the quality of the resource to be evaluated in each dimension can be evaluated in a targeted manner for the resource content of the resource to be evaluated, so that the accuracy of resource quality evaluation is improved, meanwhile, the comprehensive quality evaluation of the resource to be evaluated in the subsequent steps is facilitated, and the subsequent applications of resource classification, sorting, recommendation and distribution and the like are realized.
204. And fusing the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated.
In this embodiment of the application, after the resource content in the resource to be evaluated is evaluated through the N quality evaluation models and the N quality scores are obtained respectively, the server may further fuse the N quality scores to obtain a comprehensive quality evaluation result corresponding to the resource to be evaluated. The quality evaluation result may include a quality evaluation score, a quality evaluation grade, and the like.
In one embodiment, the above-mentioned fusion may be performed by way of summation. For example, the server may directly add the N quality scores, and use the result of the addition as a quality evaluation score corresponding to the resource to be evaluated; or each quality score may be given a corresponding weight, and the N quality scores may be weighted and summed, and the result of weighted and summed may be used as the quality evaluation score corresponding to the resource to be evaluated. Optionally, after obtaining the quality evaluation score, the server may further determine a quality evaluation level corresponding to the quality evaluation score, and use the quality evaluation score and the quality evaluation level together as a quality evaluation result of the resource to be evaluated.
In another embodiment, the fusion may be performed by averaging. For example, after directly adding and summing the N quality scores to obtain a summation result, the server may divide the summation result by N to obtain a mean value of the N quality scores, which is used as a quality evaluation score corresponding to the resource to be evaluated; or after the N quality scores are subjected to weighted summation to obtain a weighted summation result, dividing the weighted summation result by N to obtain a weighted average value of the N quality scores, which is used as the quality evaluation score corresponding to the resource to be evaluated. Optionally, after obtaining the quality evaluation score, the server may also determine a quality evaluation level corresponding to the quality evaluation score, and use the quality evaluation score and the quality evaluation level together as a quality evaluation result of the resource to be evaluated.
Therefore, by implementing the resource quality evaluation method described in the above embodiment, a plurality of quality evaluation models which are correspondingly matched can be used for performing multi-dimensional quality evaluation on the resource to be evaluated, so that the resource content of the resource to be evaluated in each dimension can be subjected to quality evaluation by using a targeted model, and a corresponding quality score is output, thereby improving the accuracy of resource quality evaluation. The obtained quality scores are fused, so that a comprehensive result after multi-dimensional quality evaluation can be obtained, the bias of evaluation results caused by evaluating the resource quality from only a single dimension (for example, evaluating only the text quality, evaluating only the picture quality and the like) can be avoided, and the reliability of resource quality evaluation is ensured. Therefore, comprehensive evaluation of resources to be evaluated can be achieved through the quality evaluation models, and accuracy and reliability of resource quality evaluation are improved.
Referring to fig. 4, fig. 4 is a schematic flow chart of another resource quality evaluation method disclosed in the embodiment of the present application. As shown in fig. 4, the resource quality evaluation method may include the steps of:
402. and determining a resource type corresponding to the resource to be evaluated, wherein the resource type comprises any one of a plain text resource type, a plain picture resource type and an image-text resource type.
In the embodiment of the application, the resource to be evaluated may include internet resources with different resource types, such as a plain text resource, a plain picture resource, an image-text resource, and the like on the internet. It can be understood that the resource quality evaluation method disclosed in the embodiment of the present application is also applicable to quality evaluation of locally stored resources. After the resource to be evaluated is obtained, the server can determine the resource type of the resource to be evaluated by analyzing the resource content in the resource to be evaluated. For example, it may be determined whether resource content in the resource to be evaluated includes picture content, and if not, it may be determined that the resource to be evaluated is a plain text resource type; if the picture content is contained, whether the contained picture content is a character picture (namely, the picture only contains the text content) can be further identified, and if all the picture content is the character picture, the resource to be evaluated is still determined to be a pure text resource type; if not, determining that the resource to be evaluated is the type of the image-text resource; if the text content does not contain any text content, the resource to be evaluated can be determined to be a pure picture resource type.
404. And determining N quality evaluation models corresponding to the resource to be evaluated according to the resource type corresponding to the resource to be evaluated, wherein N is a positive integer greater than or equal to 2.
Specifically, resources to be evaluated of different resource types may correspond to different quality evaluation model combinations, where the quality evaluation model combination may include N quality evaluation models corresponding to the resources to be evaluated, where N is a positive integer greater than or equal to 2 and may vary with the resource type of the resources to be evaluated.
Illustratively, the number and types of the N quality assessment models may vary for resources to be assessed for different resource types. For example, if the resource to be evaluated is a pure text resource type, the quality evaluation model corresponding to the resource to be evaluated may include a title quality model for evaluating the quality of a text title, a body quality model for evaluating the quality of a body text content, and may further include a body consistency model for evaluating the degree of association between a text title and a body text content (in this case, N is 3).
For another example, if the resource to be evaluated is a pure picture resource type, the quality evaluation model corresponding to the resource to be evaluated may include a title quality model for evaluating the quality of a picture name, a picture abstract, a picture label, and the like, a picture quality model for evaluating picture attribute parameters (such as resolution, aspect ratio, and the like), a picture aesthetic model for evaluating a picture aesthetic experience, and a thematic consistency model for evaluating the degree of association between the picture name, the picture abstract, the picture label, and the like and the picture content (in this case, N is 4).
For another example, if the resource to be evaluated is a type of a graphic resource, the quality evaluation model corresponding to the resource to be evaluated may include the above-mentioned title quality model, text consistency model, picture quality model, and picture aesthetics model, and may further include a graphic consistency model for evaluating a degree of association between the title, the text content of the text, and the picture content (in this case, N is 6).
By adopting the N quality evaluation models, the server can perform targeted quality evaluation on various resource contents contained in the resources to be evaluated from multiple dimensions, so that the multi-dimensional quality evaluation of the resources to be evaluated is realized, and the accuracy of the resource quality evaluation is favorably improved. It should be noted that, what is shown above is only the quality evaluation models corresponding to some types of resources to be evaluated, and no limitation is made to the number and types of the quality evaluation models corresponding to a specific resource to be evaluated, and in some embodiments, other numbers and types of quality evaluation model combinations may also be used to implement multidimensional quality evaluation on the resource to be evaluated.
After the server determines the N quality evaluation models corresponding to the resource to be evaluated according to the resource type corresponding to the resource to be evaluated, the server may extract the resource content matching each quality evaluation model from the resource to be evaluated, so that the resource content matching each quality evaluation model is used as the input of the corresponding quality evaluation model in the next step. Taking a resource to be evaluated as an image-text resource as an example, in one embodiment, a server can extract a title, a text content and a picture content from the resource to be evaluated in a mode of cutting the resource to be evaluated, and then the server can determine the title as the resource content matched with a title quality model; determining the text content of the text as resource content matched with the text quality model; determining the title and the text content as the resource content matched with the title text consistency model; determining the picture content as the resource content matched with the picture quality model and the picture aesthetic model; and jointly determining the title, the text content of the text and the picture content as the resource content matched with the image-text consistency model.
Optionally, an auxiliary Recognition model, such as an OCR (Optical Character Recognition) model, a face Recognition model, etc., may be further introduced on the basis of the quality evaluation models, so as to improve the accuracy of quality evaluation on the specific resource to be evaluated. Illustratively, when the server detects that a character picture exists in picture content contained in the resource to be evaluated, text content in the character picture can be identified through the OCR model and is input into the text quality model, the text consistency model or the image-text consistency model as a part of text content. In another exemplary embodiment, when the server detects that a face picture exists in picture content included in the resource to be evaluated, the face picture may be identified through the face identification model, and the identification result is used as a label of the face picture and is input to the picture quality model, the picture aesthetic model or the image-text consistency model together, so that model parameters (such as an adjustment feature extraction parameter, a quality evaluation parameter, and the like) may be adjusted according to the label, and the pertinence and accuracy of resource quality evaluation are improved.
406. And performing quality evaluation on the resource content matched with each quality evaluation model in the resource to be evaluated through the N quality evaluation models corresponding to the resource to be evaluated to obtain N quality scores.
Wherein, the output quality score has different meanings due to different resource contents evaluated by each quality evaluation model. For example, for the above title quality model, the higher the quality score output by the above title quality model, the better the quality of the title (e.g., high compaction degree, strong attractiveness, etc.); for the text quality model, the higher the output quality score is, the better the quality of the text content concept writing of the resource to be evaluated is (such as clear text structure, few wrongly written characters and the like); for the topic consistency model, the output quality score can be in positive correlation with the summarization degree of the topic on the text content; for the picture quality model, the higher the output quality score is, the higher the picture attribute parameters of the resource to be evaluated can be represented (such as higher resolution, less noise and the like); for the picture aesthetic model, the higher the output quality score is, the higher the aesthetic feeling experience and the more correct picture emotion of the picture content of the resource to be evaluated can be represented; for the image-text consistency model, the output quality score can form a positive correlation with the title and the correlation degree of the text content and the image content of the text.
Specifically, when the server performs quality evaluation on the resource content in the resource to be evaluated through the N quality evaluation models, each quality evaluation model may perform steps including feature extraction on the resource content, and quality evaluation using the extracted features. For example, the server may extract, from the resource to be evaluated, a first resource content that matches a first quality evaluation model (belonging to any one of the N quality evaluation models), perform, as an input to the first quality evaluation model, feature extraction on the first resource content by using the first quality evaluation model, and perform quality evaluation calculation according to a feature extraction result, thereby obtaining a first quality score corresponding to the first resource content. Meanwhile, the server may extract, from the resource to be evaluated, a second resource content that matches a second quality evaluation model (belonging to any one of the N quality evaluation models that is different from the first quality evaluation model) as an input of the second quality evaluation model, perform feature extraction on the second resource content through the second quality evaluation model, and perform quality evaluation calculation according to a feature extraction result to obtain a second quality score corresponding to the second resource content.
It can be understood that, for each quality evaluation model, because the resource content of the evaluation is different, the features to be extracted are different, and the quality evaluation models can be performed independently from each other in the quality evaluation. For example, for the resource content of text nature such as title, text content of body, etc., the matched quality evaluation model may include the ALBERT series model (for extracting text features) and the corresponding quality evaluation model; for resource content of picture properties such as picture content, the matched quality evaluation model can include an EfficientNet series model, a MobileNet series model and the like (for extracting picture features) and a corresponding quality evaluation model.
In order to realize the above feature extraction and quality evaluation calculation, each quality evaluation model needs to be trained in advance. Illustratively, a certain amount of image-text resources can be used as samples to respectively complete the training of each quality evaluation model. For example, the title quality model may be trained according to the titles in the sample resources; the text quality model can be trained according to text contents in sample resources; the topic consistency model can be trained simultaneously according to the titles and the text contents in the sample resources; the picture quality model can be trained according to picture attribute parameters (such as resolution, aspect ratio and the like) in the sample resources; the picture aesthetic model can be trained according to picture contents in sample resources; the image-text consistency model can be trained according to the title, the text content and the picture content in the sample resource. The trained quality evaluation model can be directly used in the server to realize quality evaluation on the matched resource content, so that a corresponding quality score can be obtained.
As an alternative implementation manner, if there is no resource content matching the quality evaluation model in the resource to be evaluated, the server may use a default alternative resource content as an input of the quality evaluation model, and the default alternative resource content may correspond to a determined quality score output (e.g., 0 score, a certain fixed value, etc.), so that the influence thereof may be accurately eliminated or reduced (e.g., deleted, weighted down, etc.) in the next step. For example, if there is no title in the resource to be evaluated, the title cannot be extracted from the resource to be evaluated as an input of the title quality model, and at this time, the server may obtain a default alternative resource content (such as an empty title, a sample title, etc.) and input the default alternative resource content into the title quality model, so as to obtain a quality score output by the title quality model. On the basis, the server can obtain the default quality score corresponding to the default substitute resource content, and adjust the quality score output by the title quality model according to the default quality score, so that the score bias caused by the fact that the resource to be evaluated does not have a title is eliminated or reduced before the subsequent fusion step. For example, before performing the subsequent fusion step, the server may subtract the default quality score from the output quality score to obtain a quality score with a value of 0; the weight of the output quality score may also be reduced to obtain a reduced weight quality score. By implementing the method, the quality evaluation model can be ensured not to fail when facing to special resources to be evaluated on the basis of not changing a resource quality evaluation framework, and the quality score corresponding to the resources to be evaluated can be prevented from being accidentally increased or decreased, so that the accuracy and the reliability of resource quality evaluation are favorably improved.
408. And according to the weight corresponding to each quality score, carrying out weighted summation calculation on the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated.
Specifically, different weights may be set for the quality scores output by the quality evaluation models, and according to the weight corresponding to each quality score, the server may perform weighted summation on the N quality scores, and use the quality evaluation score obtained through the weighted summation as the quality evaluation result corresponding to the resource to be evaluated. The weights can be determined according to empirical values, or according to training results of quality evaluation training performed by taking a large number of resources to be evaluated as samples.
Illustratively, according to a resource type corresponding to a resource to be evaluated, the server may determine N quality evaluation models corresponding to the resource to be evaluated, and may further obtain weights corresponding to N quality scores output by the N quality evaluation models one to one, respectively, and then perform fusion calculation of the weighted sum. For example, for a plain text resource type, the above weights may include a weight r1 corresponding to a title quality score, a weight r2 corresponding to a text quality score, and a weight r3 corresponding to a title consistency score, and then the quality evaluation score S corresponding to the resource to be evaluated may be calculated by the following formula 1:
equation 1:
S＝A·r1+B·r2+C·r3
wherein, A represents the title quality score and is the quality score output by the title quality model; b represents the text quality score which is the quality score output by the text quality model; and C represents the consistency score of the subject, which is the quality score output by the consistency model of the subject.
For another example, for the type of the image-text resource, besides the above weights r1, r2, and r3, a weight r4 corresponding to the score of the image attribute parameter, a weight r5 corresponding to the score of the image aesthetic, and a weight r6 corresponding to the score of the image-text consistency can be included, and then the quality evaluation score S corresponding to the resource to be evaluated can be calculated by using the following formula 2:
equation 2:
S＝A·r1+B·r2+C·r3+D·r4+E·r5+F·r6
wherein D represents the picture attribute parameter score and is the quality score output by the picture quality model; e represents the aesthetic score of the picture, and is the quality score output by the aesthetic model of the picture; and F represents the image-text consistency score which is the quality score output by the image-text consistency model.
Further, according to the quality evaluation score, the server may further determine a quality evaluation level corresponding to the quality evaluation score, and may use the quality evaluation level and the quality evaluation score together as a quality evaluation result corresponding to the resource to be evaluated. When the quality evaluation grade corresponding to the quality evaluation score is determined according to the quality evaluation score, the grade can be divided according to a fixed score interval, and the grade can also be divided according to the proportion of the resource to be evaluated in different scores. For example, if the score range of the quality evaluation score is [0,5] and the resource to be evaluated needs to be divided into 5 grades, the corresponding resource to be evaluated can be divided into appropriate quality evaluation grades according to whether the quality evaluation score belongs to the score interval of [0,1 ], [1,2 ], [2,3 ], [3,4) or [4,5] respectively; the resources to be evaluated can also be divided according to the proportion of 20% in each grade, and the higher the average quality evaluation score of the resources to be evaluated divided into the same grade is, the higher the corresponding quality evaluation grade can also be. By implementing the method, the corresponding quality evaluation score and quality evaluation grade can be accurately determined for the resource to be evaluated, so that simple and clear quality evaluation result output can be provided, and the method is favorable for directly applying to subsequent applications such as resource classification, sorting, recommendation and distribution and the like.
Therefore, by implementing the resource quality evaluation method described in the above embodiment, the multiple quality evaluation models that are correspondingly matched can be used for performing the multidimensional quality evaluation according to the resource type of the resource to be evaluated, and a concise and clear comprehensive quality evaluation result is obtained and output in a weighted fusion manner, so that the quality evaluation condition of the resource to be evaluated in multiple dimensions can be synthesized, the comprehensive evaluation of the resource to be evaluated is realized, and the accuracy and reliability of the resource quality evaluation are further improved.
Referring to fig. 5, fig. 5 is a schematic flow chart of another resource quality evaluation method disclosed in the embodiment of the present application. As shown in fig. 5, the resource quality evaluation method may include the steps of:
502. and determining a resource type corresponding to the resource to be evaluated, wherein the resource type at least comprises any one of a plain text resource type, a plain picture resource type and an image-text resource type.
504. And determining N quality evaluation models corresponding to the resource to be evaluated according to the resource type corresponding to the resource to be evaluated, wherein N is a positive integer greater than or equal to 2.
Step 502 and step 504 are similar to step 402 and step 404, and are not described here again.
506. And performing quality evaluation on the resource content matched with each quality evaluation model in the resource to be evaluated through the N quality evaluation models corresponding to the resource to be evaluated to obtain N quality scores.
508. And according to the weight corresponding to each quality score, carrying out weighted summation calculation on the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated.
Step 506 and step 508 are similar to step 406 and step 408, and are not described herein again.
As an optional implementation manner, after obtaining the N quality scores corresponding to the evaluation resources, the ranking result of the resources to be evaluated may be determined according to the N quality scores based on the ranking distribution policy directly, and then the recommendation operation may be performed on the resources to be evaluated according to the ranking result. Specifically, the method may be applied to a recommendation side of internet resources (e.g., a resource recommendation server, a terminal device for resource recommendation, etc.), where the recommendation side may set a ranking distribution policy, and the ranking distribution policy may include at least a target quality score type (e.g., a title quality score, a text consistency score, etc.) adopted when ranking resources to be evaluated. According to the sorting distribution strategy, the recommending side can determine one or more target quality score types from the N quality scores, and sort the resources to be evaluated according to the actual quality scores of the resources to be evaluated, which correspond to the target quality score types. On the basis, according to the ranking result of the resources to be evaluated, the recommending side can preferentially recommend the resources to be evaluated, which are ranked in front of the ranking result, to the user (such as pushing to a terminal device used by the user, updating the recommended content on a website browsed by the user, and the like).
By implementing the method, the resource to be evaluated can be sorted and recommended and distributed quickly by utilizing the quality scores of the resource to be evaluated on multiple dimensions before the comprehensive quality evaluation of the resource to be evaluated is completed, so that the resource recommendation efficiency is improved. Meanwhile, the sequencing and distributing strategy is customizable and can be flexibly changed, so that sequencing and recommendation and distribution of resources to be evaluated are favorably realized according to the required quality evaluation dimension, and the flexibility of resource recommendation is improved. It can be understood that the method can also be applied to the quality evaluation results corresponding to the resources to be evaluated, which are obtained by performing weighted summation calculation on the N quality scores, so that more comprehensive and comprehensive quality evaluation results can be used as target quality scores used in resource sorting, and resource recommendation with higher quality is realized.
510. And analyzing the resource content of the resource to be evaluated according to the variable-level strategy to obtain a variable-level adjustment result corresponding to the resource to be evaluated, and adjusting the quality evaluation result according to the variable-level adjustment result.
Illustratively, the ranking-changing strategy at least comprises one or more of degradation of quality evaluation results of sensitive resources, degradation of quality evaluation results of poor-quality resources of the content, promotion of quality evaluation results of hot-spot resources and promotion of quality evaluation results of high-quality resources of the content. The quality evaluation result may include a quality evaluation score and a quality evaluation grade. Specifically, by analyzing the resource content of the resource to be evaluated, the server may determine whether the resource to be evaluated belongs to the resource that needs to be subjected to level change and is covered by the level change policy, and may further determine a corresponding level change adjustment result (such as a level change direction, a level change number, and the like), and then may adjust the quality evaluation result corresponding to the resource to be evaluated according to the level change direction and the level change number. For example, if the resource to be evaluated is determined to be a sensitive resource (such as a political sensitive resource, a copyright sensitive resource, a pornographic sensitive resource, and the like) by analysis, the server may determine the number of levels that the resource to be evaluated needs to be degraded according to the sensitivity degree of the resource to be evaluated, and then reduce the quality evaluation level corresponding to the resource to be evaluated by the number of levels; or after determining the target quality evaluation level to which the resource to be evaluated needs to be reduced, adjusting the quality evaluation score corresponding to the resource to be evaluated to the score range corresponding to the target quality evaluation level. For another example, if the resource to be evaluated is determined to be a poor resource (e.g., a text poor resource, a picture poor resource, a marketing poor resource, etc.) through analysis, the server may further determine a degradation adjustment result (e.g., a degradation level, a degradation score, etc.) according to the current quality evaluation result, and then may directly adjust the quality evaluation level and the quality evaluation score corresponding to the resource to be evaluated according to the degradation adjustment result.
Referring to fig. 6, fig. 6 is a schematic diagram of another resource quality evaluation framework disclosed in the embodiment of the present application, and the resource quality evaluation framework shown in fig. 6 is obtained by performing optimization on the resource quality evaluation framework shown in fig. 1. As shown in fig. 6, in order to implement the above-mentioned variable-level adjustment, the resource quality evaluation framework may further include a policy variable-level module 30 in addition to the model evaluation module 10 and the fusion module 20, and the policy variable-level module 30 may be configured to perform the above-mentioned variable-level adjustment on the quality evaluation result output by the fusion module 20 according to a variable-level policy, so as to further improve the flexibility of resource quality evaluation. It should be noted that, for resources to be evaluated of different resource types, not only the model evaluation module 10 may perform feature extraction and quality evaluation by using different quality evaluation models 11 (i.e., quality evaluation models), but also the fusion module 20 may perform fusion calculation according to different fusion calculation formulas (such as the above formula 1 and formula 2), and perform corresponding level-changing adjustment on the quality evaluation result output by the fusion module 20 according to the resource type in the policy level-changing module 30. The quality evaluation of various resources to be evaluated is realized through a uniform resource quality evaluation framework, and the efficiency and convenience of the resource quality evaluation are improved.
512. And generating a quality label corresponding to the resource to be evaluated according to the adjusted quality evaluation result and the resource type corresponding to the resource to be evaluated, wherein the quality label at least comprises one or more of a quality score label and a quality grade label.
In this embodiment of the application, when the quality evaluation result includes a quality evaluation score corresponding to a resource to be evaluated, the server may further determine a quality evaluation level corresponding to the resource to be evaluated according to the quality evaluation score and a resource type corresponding to the resource to be evaluated; when the quality evaluation result includes the quality evaluation score and the quality evaluation grade corresponding to the resource to be evaluated, the server may continue to use the quality evaluation grade. For resources to be evaluated of different resource types, the quality evaluation level can be determined according to different standards. For example, when the quality evaluation scores of the resources to be evaluated are the same, the resources to be evaluated of the image-text resource type may correspond to a higher quality evaluation level, and the resources to be evaluated of the plain text resource type or the plain picture resource type may correspond to a relatively lower quality evaluation level. According to the quality evaluation score and the quality evaluation grade corresponding to the resource to be evaluated, the server can generate a quality label corresponding to the resource to be evaluated for marking the resource in the subsequent steps.
514. And marking the resources to be evaluated according to the quality label.
Specifically, the server may mark the corresponding resource to be evaluated through the quality label, so as to bind the quality label with the resource to be evaluated, and may facilitate subsequent operations such as classification, sorting, recommendation and distribution of the resource. For example, after acquiring the resources subjected to the quality evaluation, the resource recommendation server (or the resource recommendation terminal device) may perform ranking according to the quality tags thereof, and perform recommendation operation on the resources according to the ranking result; and the resources can be classified according to the quality labels, and high-quality resources are divided to recommend distribution to users.
It can be seen that, by implementing the resource quality evaluation method described in the above embodiment, quality evaluation of various resources to be evaluated can be achieved through a unified resource quality evaluation framework, and a plurality of quality evaluation models corresponding to the resources to be evaluated are adopted, and a concise and clear comprehensive quality evaluation result is obtained and output through a weighted fusion manner, so that quality evaluation conditions of the resources to be evaluated in multiple dimensions can be integrated, and comprehensive evaluation of the resources to be evaluated is achieved; meanwhile, through strategy grading, the quality evaluation result of the resource to be evaluated can be adjusted according to actual requirements, and the flexibility of resource quality evaluation is favorably improved.
Referring to fig. 7, fig. 7 is a schematic block diagram of a resource quality evaluation apparatus according to an embodiment of the present disclosure. As shown in fig. 7, the resource quality evaluation apparatus may include a quality evaluation unit 701 and a fusion unit 702, wherein:
a quality evaluation unit 701, configured to perform quality evaluation on resource content, which is matched with each quality evaluation model, in the resource to be evaluated through N quality evaluation models corresponding to the resource to be evaluated, to obtain N quality scores, where N is a positive integer greater than or equal to 2;
and a fusion unit 702, configured to fuse the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated.
It can be seen that, by using the resource quality evaluation apparatus described in the foregoing embodiment, a plurality of quality evaluation models that are correspondingly matched can be used for performing multi-dimensional quality evaluation on the resource to be evaluated, so that the resource content of the resource to be evaluated in each dimension can be subjected to quality evaluation by using a targeted model, and a corresponding quality score is output, so as to improve the accuracy of resource quality evaluation. The obtained quality scores are fused, so that a comprehensive result after multi-dimensional quality evaluation can be obtained, the bias of evaluation results caused by evaluating the resource quality from only a single dimension (for example, evaluating only the text quality, evaluating only the picture quality and the like) can be avoided, and the reliability of resource quality evaluation is ensured. Therefore, comprehensive evaluation of resources to be evaluated can be achieved through the quality evaluation models, and accuracy and reliability of resource quality evaluation are improved.
Referring to fig. 8, fig. 8 is a schematic block diagram of another resource quality evaluation apparatus according to an embodiment of the present disclosure. The resource quality evaluation device shown in fig. 8 is optimized by the resource quality evaluation device shown in fig. 7. Compared with the resource quality evaluation apparatus shown in fig. 7, the resource quality evaluation apparatus shown in fig. 8 may further include a first determination unit 703 and a second determination unit 704, where:
a first determining unit 703, configured to determine a resource type corresponding to the resource to be evaluated before the quality evaluation unit 701 performs quality evaluation on the resource content, which is matched with each quality evaluation model, in the resource to be evaluated through N quality evaluation models corresponding to the resource to be evaluated to obtain N quality scores, where the resource type at least may include any one of a plain text resource type, a plain picture resource type, and an image-text resource type;
a second determining unit 704, configured to determine, according to the resource type corresponding to the resource to be evaluated, N quality evaluation models corresponding to the resource to be evaluated.
Exemplarily, if the resource type corresponding to the resource to be evaluated is a pure text resource type, the quality evaluation model corresponding to the resource to be evaluated at least includes one or more of a title quality model, a text quality model, and a text consistency model;
if the resource type corresponding to the resource to be evaluated is a pure picture resource type, the quality evaluation model corresponding to the resource to be evaluated at least comprises one or more of a title quality model, a picture quality model and a picture aesthetic model;
if the resource type corresponding to the resource to be evaluated is the image-text resource type, the quality evaluation model corresponding to the resource to be evaluated at least comprises one or more of a title quality model, a text consistency model, an image quality model, an image aesthetic model and an image-text consistency model.
The title quality model may be a model obtained by training according to a title in a sample resource, the text quality model may be a model obtained by training according to text content in the sample resource, the title consistency model may be a model obtained by simultaneously training according to the title and the text content in the sample resource, the picture quality model may be a model obtained by training according to picture attribute parameters in the sample resource, the picture aesthetic model may be a model obtained by training according to the picture content in the sample resource, and the picture-text consistency model may be a model obtained by training according to the title, the text content and the picture content in the sample resource.
By adopting the resource quality evaluation device described in the above embodiment, the corresponding quality evaluation model can be adopted for a variety of resources to be evaluated, and the pertinence and accuracy of resource quality evaluation are further improved.
In an embodiment, the quality evaluation unit 701 may include a feature extraction subunit and a quality evaluation calculation subunit, not shown in the figure, wherein:
a feature extraction subunit, configured to extract, from a resource to be evaluated, a first resource content that matches a first quality evaluation model, where the first quality evaluation model is any one of the N quality evaluation models;
and the quality evaluation operator unit is used for inputting the first resource content into the first quality evaluation model, performing characteristic extraction on the first resource content through the first quality evaluation model, and performing quality evaluation calculation according to a characteristic extraction result to obtain a first quality score corresponding to the first resource content.
In an embodiment, the resource quality evaluation apparatus shown in fig. 8 may further include a sorting unit 705 and a recommending unit 706, where:
a sorting unit 705, configured to perform quality evaluation on resource content, which is matched with each quality evaluation model, in the resource to be evaluated through N quality evaluation models corresponding to the resource to be evaluated in the quality evaluation unit 701 to obtain N quality scores, and determine a sorting result of the resource to be evaluated according to the N quality scores based on a sorting distribution policy;
and the recommending unit 706 is configured to perform recommending operation on the resources to be evaluated according to the sorting result.
By adopting the resource quality evaluation device described in the above embodiment, before the comprehensive quality evaluation of the resources to be evaluated is completed, the ranking and recommendation distribution of the resources to be evaluated can be quickly realized by using the quality scores of the resources to be evaluated in multiple dimensions, so that the resource recommendation efficiency is improved. Meanwhile, the sequencing and distributing strategy is customizable and can be flexibly changed, so that sequencing and recommendation and distribution of resources to be evaluated are favorably realized according to the required quality evaluation dimension, and the flexibility of resource recommendation is improved.
In an embodiment, the fusion unit 702 may be specifically configured to perform weighted summation calculation on the N quality scores according to a weight corresponding to each quality score, so as to obtain a quality evaluation result corresponding to the resource to be evaluated.
In an embodiment, the resource quality evaluation apparatus shown in fig. 8 may further include a policy ranking changing unit 707, where the policy ranking changing unit 707 may be configured to, after the fusion unit 702 obtains the quality evaluation result corresponding to the resource to be evaluated, analyze the resource content of the resource to be evaluated according to a ranking changing policy, obtain a ranking changing adjustment result corresponding to the resource to be evaluated, and adjust the quality evaluation result according to the ranking changing adjustment result.
Alternatively, after the policy ranking changing unit 707 performs the ranking changing adjustment, the ranking unit 705 and the recommending unit 706 may also rank and recommend and distribute the resources to be evaluated according to the adjusted quality evaluation result.
The ranking-changing strategy at least comprises one or more of degradation of quality evaluation results of sensitive resources, degradation of quality evaluation results of poor-quality resources of the content, upgrading of quality evaluation results of hot-spot resources and upgrading of quality evaluation results of high-quality resources of the content.
By adopting the resource quality evaluation device described in the above embodiment, a simple and clear comprehensive quality evaluation result can be obtained and output in a weighted fusion manner, so that the quality evaluation conditions of the resources to be evaluated in multiple dimensions can be integrated, and the comprehensive evaluation of the resources to be evaluated can be realized; meanwhile, through strategy grading, the quality evaluation result of the resource to be evaluated can be adjusted according to actual requirements, and the flexibility of resource quality evaluation is favorably improved.
In an embodiment, the resource quality evaluation apparatus shown in fig. 8 may further include a label generation unit 708 and a marking unit 709, where:
a tag generating unit 708, configured to generate a quality tag corresponding to the resource to be evaluated according to the adjusted quality evaluation result and the resource type corresponding to the resource to be evaluated, where the quality tag at least includes one or more of a quality score tag and a quality level tag;
the marking unit 709 is configured to mark the resource to be evaluated according to the quality label, so that the quality label can be bound to the resource to be evaluated, and subsequent operations such as classification, sorting, recommendation and distribution of the resource are facilitated.
Referring to fig. 9, fig. 9 is a schematic block diagram of an electronic device according to an embodiment of the present disclosure. As shown in fig. 9, the electronic device may include:
a memory 901 in which executable program code is stored;
a processor 902 coupled to a memory 901;
the processor 902 calls the executable program code stored in the memory 901, and may execute all or part of the steps in any resource quality evaluation method described in the above embodiments.
In addition, the present application further discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute all or part of the steps in any one of the resource quality evaluation methods described in the above embodiments.
In addition, the embodiments of the present application further disclose a computer program product, which, when running on a computer, enables the computer to execute all or part of the steps in any one of the resource quality evaluation methods described in the above embodiments.
It will be understood by those skilled in the art that all or part of the steps in the methods of the embodiments described above may be implemented by hardware instructions of a program, and the program may be stored in a computer-readable storage medium, where the storage medium includes Read-Only Memory (ROM), Random Access Memory (RAM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), One-time Programmable Read-Only Memory (OTPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Compact Disc Read-Only Memory (CD-ROM), or other Memory, such as a magnetic disk, or a combination thereof, A tape memory, or any other medium readable by a computer that can be used to carry or store data.
The resource quality evaluation method and apparatus, the electronic device, and the storage medium disclosed in the embodiments of the present application are introduced in detail, and a specific example is applied in the present application to explain the principle and the implementation manner of the present application, and the description of the embodiments is only used to help understand the method and the core idea of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Claims (12)
1. A resource quality evaluation method is characterized by comprising the following steps:
performing quality evaluation on resource content matched with each quality evaluation model in the resources to be evaluated through N quality evaluation models corresponding to the resources to be evaluated to obtain N quality scores, wherein N is a positive integer greater than or equal to 2;
and fusing the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated.
2. The method according to claim 1, wherein before performing quality evaluation on resource contents, which are matched with each quality evaluation model, in the resources to be evaluated through N quality evaluation models corresponding to the resources to be evaluated to obtain N quality scores, the method further comprises:
determining a resource type corresponding to a resource to be evaluated, wherein the resource type comprises any one of a plain text resource type, a plain picture resource type and an image-text resource type;
and determining N quality evaluation models corresponding to the resources to be evaluated according to the resource types corresponding to the resources to be evaluated.
3. The method according to claim 2, wherein if the resource type corresponding to the resource to be evaluated is a plain text resource type, the quality evaluation model corresponding to the resource to be evaluated at least comprises one or more of a title quality model, a text quality model, and a text consistency model;
if the resource type corresponding to the resource to be evaluated is a pure picture resource type, the quality evaluation model corresponding to the resource to be evaluated at least comprises one or more of a title quality model, a picture quality model and a picture aesthetic model;
if the resource type corresponding to the resource to be evaluated is an image-text resource type, the quality evaluation model corresponding to the resource to be evaluated at least comprises one or more of a title quality model, a text consistency model, an image quality model, an image aesthetic model and an image-text consistency model;
the image quality model is obtained by training according to image attribute parameters in the sample resources, the image aesthetic model is obtained by training according to image contents in the sample resources, and the image consistency model is obtained by training according to the image contents in the sample resources.
4. The method according to claim 1, wherein the quality evaluation of the resource content matching each quality evaluation model in the resource to be evaluated through N quality evaluation models corresponding to the resource to be evaluated to obtain N quality scores comprises:
extracting first resource content matched with a first quality evaluation model from resources to be evaluated, wherein the first quality evaluation model is any one of the N quality evaluation models;
inputting the first resource content into the first quality evaluation model, performing feature extraction on the first resource content through the first quality evaluation model, and performing quality evaluation calculation according to a feature extraction result to obtain a first quality score corresponding to the first resource content.
5. The method according to claim 1, wherein after the resource content matching with each quality evaluation model in the resource to be evaluated is subjected to quality evaluation through N quality evaluation models corresponding to the resource to be evaluated to obtain N quality scores, the method further comprises:
determining a sorting result of the resources to be evaluated according to the N quality scores based on a sorting distribution strategy;
and recommending the resources to be evaluated according to the sequencing result.
6. The method according to any one of claims 1 to 5, wherein the fusing the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated comprises:
and according to the weight corresponding to each quality score, carrying out weighted summation calculation on the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated.
7. The method according to any one of claims 1 to 5, wherein after the fusing the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated, the method further comprises:
and analyzing the resource content of the resource to be evaluated according to the variable-level strategy to obtain a variable-level adjustment result corresponding to the resource to be evaluated, and adjusting the quality evaluation result according to the variable-level adjustment result.
8. The method of claim 7, wherein the ranking policy at least comprises one or more of degradation of quality evaluation results of sensitive resources, degradation of quality evaluation results of poor-quality resources of the content, promotion of quality evaluation results of hot-spot resources, and promotion of quality evaluation results of high-quality resources of the content.
9. The method of claim 7, wherein after said adjusting said quality assessment result according to said level-change adjustment result, said method further comprises:
generating a quality label corresponding to the resource to be evaluated according to the adjusted quality evaluation result and the resource type corresponding to the resource to be evaluated, wherein the quality label at least comprises one or more of a quality score label and a quality grade label;
and marking the resource to be evaluated according to the quality label.
10. An apparatus for evaluating quality of image-text data, comprising:
the quality evaluation unit is used for carrying out quality evaluation on resource content matched with each quality evaluation model in the resources to be evaluated through N quality evaluation models corresponding to the resources to be evaluated to obtain N quality scores, wherein N is a positive integer greater than or equal to 2;
and the fusion unit is used for fusing the N quality scores to obtain a quality evaluation result corresponding to the resource to be evaluated.
11. An electronic device, comprising:
a memory storing executable program code;
a processor coupled with the memory;
the processor calls the executable program code stored in the memory to perform the method of any of claims 1 to 9.
12. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, wherein the computer program causes a computer to perform the method of any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011582561.9A CN112613775A (en) | 2020-12-28 | 2020-12-28 | Resource quality evaluation method and device, electronic device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011582561.9A CN112613775A (en) | 2020-12-28 | 2020-12-28 | Resource quality evaluation method and device, electronic device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112613775A true CN112613775A (en) | 2021-04-06 |
Family
ID=75248562
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011582561.9A Pending CN112613775A (en) | 2020-12-28 | 2020-12-28 | Resource quality evaluation method and device, electronic device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112613775A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116170360A (en) * | 2022-12-08 | 2023-05-26 | 中国联合网络通信集团有限公司 | Network quality evaluation method, device and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111882371A (en) * | 2019-04-15 | 2020-11-03 | 阿里巴巴集团控股有限公司 | Content information processing method, image-text content processing method, computer device, and medium |
-
2020
- 2020-12-28 CN CN202011582561.9A patent/CN112613775A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111882371A (en) * | 2019-04-15 | 2020-11-03 | 阿里巴巴集团控股有限公司 | Content information processing method, image-text content processing method, computer device, and medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116170360A (en) * | 2022-12-08 | 2023-05-26 | 中国联合网络通信集团有限公司 | Network quality evaluation method, device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8225190B1 (en) | Methods and apparatus for clustering news content | |
| CN102073699B (en) | For improving the method for Search Results, device and equipment based on user behavior | |
| US10235680B2 (en) | System and method for populating a database with user input | |
| CN106708817B (en) | Information searching method and device | |
| CN106970991B (en) | Similar application identification method and device, application search recommendation method and server | |
| EP2068276A1 (en) | Information processing device and method, program, and recording medium | |
| CN110852793A (en) | Document recommendation method and device and electronic equipment | |
| KR20160055930A (en) | Systems and methods for actively composing content for use in continuous social communication | |
| CN111177559B (en) | Text travel service recommendation method and device, electronic equipment and storage medium | |
| CN103455538B (en) | Information processing unit, information processing method and program | |
| CN107590232A (en) | A kind of resource recommendation system and method based on Network Study Environment | |
| CN111400586A (en) | Group display method, terminal, server, system and storage medium | |
| Chiny et al. | Netflix recommendation system based on TF-IDF and cosine similarity algorithms | |
| US9235655B2 (en) | Task-based design evaluation | |
| CN111209480A (en) | Method and device for determining pushed text, computer equipment and medium | |
| CN116823410B (en) | Data processing method, object processing method, recommending method and computing device | |
| CN112613775A (en) | Resource quality evaluation method and device, electronic device and storage medium | |
| CN110188277B (en) | Resource recommendation method and device | |
| CN112989053A (en) | Periodical recommendation method and device | |
| CN113158022A (en) | Service recommendation method, device, server and storage medium | |
| CN111882224A (en) | Method and device for classifying consumption scenes | |
| CN111597787A (en) | Test question generation method and system | |
| CN108595453B (en) | URL (Uniform resource locator) identifier mapping obtaining method and device | |
| CN113657766A (en) | Tourist attraction joy index metering method based on tourist multi-metadata | |
| CN114201680A (en) | Method for recommending marketing product content to user |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |