Webpage quality evaluation method, webpage quality evaluation model training method and device

Abstract

The present disclosure discloses a webpage quality evaluation method and a webpage quality evaluation model training method and device, relates to the technical field of computers, and particularly relates to the technical field of intelligent search and intelligent recommendation. The specific implementation scheme is as follows: in response to receiving a quality evaluation request of a target webpage, webpage content information of the target webpage and a target search term used to search for the target webpage are obtained. According to the webpage content information and the target search term, relevance information of the target search term and the target webpage is obtained. According to the relevance information and the webpage content information, a webpage quality evaluation result is obtained.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to the field of intelligent search and intelligent recommendation technology, specifically to a webpage quality assessment method, a training method and apparatus for a webpage quality assessment model. Background Technology

[0002] In the internet age, after receiving search terms from users, search engines retrieve relevant web pages based on those terms. The quality of these web pages directly determines the search engine's retrieval results. Therefore, accurate assessment of web page quality is of great importance to search engines. Summary of the Invention

[0003] This disclosure provides a webpage quality assessment method, a webpage quality model training method, and an apparatus.

[0004] According to one aspect of this disclosure, a webpage quality assessment method is provided, comprising: in response to receiving a quality assessment request for a target webpage, obtaining webpage content information of the target webpage and target search terms used to search for the target webpage; obtaining relevance information between the target search terms and the target webpage based on the webpage content information and the target search terms; and obtaining a webpage quality assessment result based on the relevance information and the webpage content information.

[0005] According to another aspect of this disclosure, a method for training a webpage quality assessment model is provided, comprising: obtaining sample relevance information between sample search terms and sample webpages based on the content information of sample webpages and sample search terms used to search for sample webpages; processing the sample relevance information and the content information of sample webpages using a preset model to obtain a webpage quality assessment result for the sample webpages; obtaining loss information based on the webpage quality assessment result and the tags of sample webpages according to a target loss function; and adjusting the model parameters of the preset model based on the loss information to obtain a webpage quality assessment model.

[0006] According to another aspect of this disclosure, a webpage quality assessment apparatus is provided, comprising: an acquisition module, a first relevance analysis module, and a first assessment module. The acquisition module is configured to, in response to receiving a quality assessment request for a target webpage, acquire webpage content information of the target webpage and target search terms used to search for the target webpage. The first relevance analysis module is configured to, based on the webpage content information and the target search terms, obtain relevance information between the target search terms and the target webpage. The first assessment module is configured to, based on the relevance information and the webpage content information, obtain a webpage quality assessment result.

[0007] According to another aspect of this disclosure, a training apparatus for a webpage quality assessment model is provided, comprising: a second relevance analysis module, a second assessment module, an acquisition module, and an adjustment module. The second relevance analysis module is used for...

[0008] According to another aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method as described above.

[0009] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform the method described above.

[0010] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the method described above.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0012] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0013] Figure 1 This illustration schematically shows an exemplary system architecture of a training method and apparatus for applying webpage quality assessment methods or webpage quality assessment models according to embodiments of the present disclosure;

[0014] Figure 2 A flowchart illustrating a webpage quality assessment method according to an embodiment of the present disclosure is shown schematically.

[0015] Figure 3 This illustration schematically shows a diagram of obtaining relevance information based on image information, text information, and search terms according to an embodiment of the present disclosure;

[0016] Figure 4 The illustration shows a schematic diagram of obtaining webpage quality assessment results according to an embodiment of the present disclosure;

[0017] Figure 5 The illustration shows a schematic diagram of obtaining webpage quality assessment results according to other embodiments of this disclosure;

[0018] Figure 6 A flowchart illustrating a method for training a webpage quality assessment model according to an embodiment of the present disclosure is shown schematically.

[0019] Figure 7A block diagram of a webpage quality assessment apparatus according to an embodiment of the present disclosure is shown schematically.

[0020] Figure 8 A block diagram schematically illustrates a training apparatus for a webpage quality assessment model according to an embodiment of the present disclosure; and

[0021] Figure 9 A block diagram of an electronic device suitable for implementing a webpage quality assessment method or a training method for a webpage quality assessment model, according to embodiments of the present disclosure, is shown schematically. Detailed Implementation

[0022] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0023] After receiving the search terms entered by the user, the search engine retrieves relevant web pages based on the search terms. The quality of the web pages directly determines the search engine's retrieval results.

[0024] In related technologies, webpage quality is generally evaluated based on users' browsing history or feedback. This post-hoc evaluation method has low accuracy due to the subjectivity of user browsing behavior and feedback.

[0025] In view of this, this disclosure provides a webpage quality assessment method, comprising: in response to receiving a quality assessment request for a target webpage, obtaining webpage content information of the target webpage and target search terms used to search for the target webpage; obtaining relevance information between the target search terms and the target webpage based on the webpage content information and the target search terms; and obtaining a webpage quality assessment result based on the relevance information and the webpage content information. This method objectively assesses the webpage quality based on the relevance between the search terms and the webpage content, as well as the webpage content itself, thereby improving the accuracy of the webpage quality assessment result.

[0026] Figure 1 The illustration schematically shows an exemplary system architecture of a training method and apparatus for applying web page quality assessment methods or web page quality assessment models according to embodiments of the present disclosure.

[0027] It is important to note that Figure 1The examples shown are merely illustrative of system architectures applicable to embodiments of this disclosure, intended to help those skilled in the art understand the technical content of this disclosure. They do not imply that embodiments of this disclosure cannot be used in other devices, systems, environments, or scenarios. For instance, in another embodiment, an exemplary system architecture for applying the webpage quality assessment method and apparatus may include a terminal device. However, the terminal device can implement the webpage quality assessment method and apparatus provided by embodiments of this disclosure without interacting with a server.

[0028] like Figure 1 As shown, the system architecture 100 according to this embodiment may include a database 101, a data interface 102, a relevance analysis module 103, a content quality analysis module 104, and a webpage quality assessment model 105.

[0029] Database 101 can store web page resources and the index relationship between web page resources and search terms. Web page resources corresponding to search terms can be retrieved from database 101 by calling data interface 102. Web page resources may include images, text, web page structure, web page link addresses, etc., displayed on the web page.

[0030] The relevance analysis module 103 is used to analyze the relevance between search terms and webpage content. For example, the search term could be "XX brand," and the analysis would examine whether the product brand in the webpage content is related to the brand in the search term.

[0031] The content quality analysis module 104 is used to analyze the quality of webpage content. For example, it checks whether there are watermarks on the images displayed on the webpage, and whether the text displayed on the webpage has issues such as excessive text piling up or being unclear.

[0032] For example, when a user enters the search term "vacuum cleaner," the webpage corresponding to the search term may include product images, text information such as brand, origin, and price, and may also include a ranking of recommended vacuum cleaners displayed at the bottom of the webpage, such as the top 10 vacuum cleaners in a certain region. The relevance analysis module 103 can perform relevance analysis between the search term and the image and text information, while the content quality analysis module 104 can perform quality analysis on the image and text information. The webpage quality assessment model 105 can evaluate the quality of the webpage based on the relevance analysis results and content quality analysis results from the relevance analysis module 103 and the content quality analysis module 104.

[0033] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0034] In the technical solution disclosed herein, the collection, storage, use, processing, transmission, provision, disclosure, and application of user personal information comply with the provisions of relevant laws and regulations, necessary confidentiality measures have been taken, and there is no violation of public order and good morals.

[0035] In the technical solution disclosed herein, the user's authorization or consent is obtained before acquiring or collecting the user's personal information.

[0036] Figure 2 A flowchart illustrating a webpage quality assessment method according to an embodiment of the present disclosure is shown schematically.

[0037] like Figure 2 As shown, the method includes operations S210 to S230.

[0038] In operation S210, in response to receiving a quality assessment request for the target webpage, the webpage content information of the target webpage and the target search terms used to search for the target webpage are obtained.

[0039] In operation S220, based on the webpage content information and the target search term, the relevance information between the target search term and the target webpage is obtained.

[0040] In operation S230, the webpage quality assessment results are obtained based on relevance information and webpage content information.

[0041] According to embodiments of this disclosure, webpage content information may include image information and text information displayed on the target webpage. Image information may include images, videos, and layout information for displaying the images or videos on the target webpage. Text information may include text content and text layout information on the target webpage.

[0042] According to embodiments of this disclosure, the relevance information between a target search term and a target webpage can characterize the semantic and category relevance between the target search term and the webpage content. For example, the target search term could be "water cup," and the webpage content could include "glass cup," "ceramic cup," "mug," etc. If the "glass cup," "ceramic cup," and "mug" displayed in the webpage content belong to the same product category as "water cup," it indicates that the target search term is relevant to the webpage content.

[0043] According to embodiments of this disclosure, the relevance information between the target search term and the target webpage can also characterize the relevance of the target search term itself to the search intent. By identifying the query fields of the target search term itself, it can be determined whether the target search term satisfies the search intent, such as whether the target search term is a content-related search term or a downgraded search term.

[0044] According to embodiments of this disclosure, relevance information involves multiple dimensions, such as product category relevance, brand relevance, and regional relevance, and the relevance information of each dimension is not correlated with each other. Relevance information from multiple dimensions can be converted into feature information that can be processed by a neural network model through configuration information. For example, if the product category of the target search term is related to the product category displayed on the target webpage, the corresponding relevance feature can be 1. If the product category of the target search term is not related to the product category displayed on the target webpage, the corresponding relevance feature can be 0.

[0045] According to embodiments of this disclosure, webpage content information may include image information and text information displayed on the webpage. For example: if a watermark is present in the image, the webpage image content feature can be determined to be 1. If no watermark is present in the image, the webpage image content feature can be determined to be 0. If the text is unclear, the webpage text content feature can be determined to be 1. If the text is clear, the webpage text content feature can be determined to be 0.

[0046] According to embodiments of this disclosure, different weights can be assigned to the text content features and image content features of a webpage. The webpage content features are obtained by weighted summing of the text content features and image content features with their respective weights. For example, the weight of the image content features can be 0.6, and the weight of the text content features can be 0.4. In the case where there is no watermark in the image but the text is unclear, the webpage content feature can be determined to be 0.4.

[0047] According to embodiments of this disclosure, the quality of a webpage can be evaluated by combining the relevance of the target search term to the webpage content and the webpage content information using a trained webpage quality assessment model.

[0048] According to embodiments of this disclosure, a technical means is employed to obtain relevance information between target search terms and target web pages based on web page content information and target search terms, and then to obtain web page quality assessment results based on the relevance information and web page content information. This achieves objective assessment of web page quality and improves the accuracy of web page quality assessment results.

[0049] According to embodiments of this disclosure, the above operation S220 may include the following operations:

[0050] Based on text information and target search terms, obtain text relevance information. Based on image information and target search terms, obtain image relevance information. Based on text relevance information and image relevance information, obtain overall relevance information.

[0051] According to embodiments of this disclosure, the text information may include brand information of the top 10 recommended products displayed on the target webpage. The relevance between the brand information in the target search term and the brand information of the recommended products can represent the number of recommended products whose brand information matches the brand information in the target search term. The correlation between the number of products with matching brand information and brand relevance can be configured through configuration information.

[0052] For example: if the number of products with identical brand information is 10, the brand relevance can be 3. If the number of products with identical brand information is greater than or equal to 5 and less than 10, the brand relevance can be 2. If the number of products with identical brand information is greater than 0 and less than 5, the brand relevance can be 1. If the number of products with identical brand information is 0, the brand relevance can be 0.

[0053] According to embodiments of this disclosure, image information may include object information and text information within the image. The relevance of image information to a target search term can be determined by recognizing objects in the image. Similarly, the relevance of image information to a target search term can be determined by recognizing text within the image.

[0054] For example, the target search term could be "truck," and the objects displayed in the images on the target webpage could include: "large truck," "medium truck," "small truck," and "family car." The relevance between the image information and the target search term can be determined using the image recognition results on the target webpage. For example, the relevance of "large truck" could be 5, "medium truck" could be 3, "small truck" could be 2, and "family car" could be 0.

[0055] For example, images displayed on a target webpage can also include text information, such as the image's title or other text describing the image. The relevance of the image information to the target search term can be determined using the text recognition results. For instance, if the recognized text in the image is "boat," which is completely unrelated to the target search term "truck," the relevance between the image information and the target search term can be determined to be 0.

[0056] According to embodiments of this disclosure, by performing relevance analysis on the image information and text information in the target webpage with the target search terms respectively, the dimensions of the relevance analysis are increased, effectively improving the accuracy of the relevance analysis.

[0057] According to embodiments of this disclosure, obtaining text relevance information based on text information and target search terms may include the following operations: extracting a first semantic feature of the text information and a second semantic feature of the target search term; and processing the first and second semantic features to obtain the text relevance information.

[0058] According to embodiments of this disclosure, in a target webpage, descriptions of information such as product brand, origin, and performance parameters are generally contained within a single text segment. Therefore, when extracting the first semantic features of the text information, a syntactic-semantic structure analysis can be performed first. Based on the results of the syntactic-semantic analysis, the text information can be divided into multiple clauses according to different semantics. For example, the text describing the product brand can be considered as one clause. The text describing the product's performance parameters can be considered as another clause. Then, semantic features are extracted from each clause separately.

[0059] For example: The power consumption of a XX brand vacuum cleaner is yy. Depending on the semantics, the resulting sentences could be "XX brand," "vacuum cleaner," and "power consumption is yy." Semantic analysis reveals that the brand information is "XX," the product is a "vacuum cleaner," and the product's performance parameter is "power consumption is yy." The second semantic feature could include: the brand feature "XX," the subject feature "vacuum cleaner," and the performance feature "power consumption yy."

[0060] According to embodiments of this disclosure, the target search term is generally a short text, which can be segmented first and then semantically analyzed to obtain the first semantic feature. Alternatively, semantic analysis can be performed directly without segmentation to obtain the first semantic feature. For example, the target search term could be "XX brand vacuum cleaner", and the first semantic feature could include: the brand feature "XX" and the subject feature "vacuum cleaner".

[0061] According to embodiments of this disclosure, text relevance information can be obtained by comparing a first semantic feature and a second semantic feature. For example, the first semantic feature may include: brand feature "XX" and subject feature "vacuum cleaner". The second semantic feature may include: brand feature "XX", subject feature "vacuum cleaner", and performance feature "power consumption yy". Through semantic comparison analysis, it can be determined that the first semantic feature and the second semantic feature are related, and the text relevance can be "1".

[0062] According to embodiments of this disclosure, obtaining image relevance information based on image information and target search terms may include the following operations: extracting image features from the image information; processing the image features to obtain an image recognition result; and obtaining image relevance information based on the image recognition result and the target search terms.

[0063] According to embodiments of this disclosure, image information may include object information in the image, and may also include text information in the image. The text information may be the title of the image, other descriptive text in the image, or identifying text that can be used to identify object information. For example, the identifying text on the image may be "dictionary," which can determine that the category of the object in the image is books.

[0064] For example, the target search term could be "reference book" and the image recognition result could be "dictionary". Both reference books and dictionaries belong to the category of books, indicating that the target search term and the image information are related, and the image relevance can be "1".

[0065] It should be noted that the embodiments of this disclosure can utilize image recognition models, text recognition models, semantic analysis models, etc., for image recognition, text recognition, and semantic analysis. The image recognition model, text recognition model, and semantic analysis model can all employ any model from related technologies capable of achieving image recognition, text recognition, and semantic analysis; no specific limitations are imposed here.

[0066] According to embodiments of this disclosure, operation S230 may include the following operations: processing webpage content information to obtain content quality information; and obtaining a webpage quality assessment result based on relevance information and content quality information.

[0067] The following is for reference. Figures 3-5 In conjunction with specific embodiments, Figure 2 The method shown will be further explained.

[0068] Figure 3 The illustration shows a schematic diagram of obtaining relevance information based on image information, text information, and search terms according to an embodiment of the present disclosure.

[0069] like Figure 3 As shown, in embodiment 300, the webpage content information of the target webpage may include image information 3201 and text information 3206. Image features 3202 of image information 3201, second semantic features 3207 of text information 3206, and first semantic features 3205 of target search term 3204 are extracted respectively. Image recognition result 3203 is obtained by recognizing image features 3202. Image relevance information 3208 is obtained based on image recognition result 3203 and first semantic features 3205. Text relevance information 3209 is obtained based on first semantic features 3205 and second semantic features 3207. Relevance information 3210 between the webpage content of the target webpage and the target search term is obtained based on image relevance information 3208 and text relevance information 3209.

[0070] According to embodiments of this disclosure, webpage content information can include image information and text information. Content quality information can also include image quality information and text quality information. Image quality can be reflected in several aspects, such as whether the images on the target webpage are clear, whether the images have watermarks, and whether there is occlusion between images. Text quality can be reflected in several aspects, such as whether the text on the target webpage is clear, whether the text is piled up, and the quality of the recommended products in the text.

[0071] According to embodiments of this disclosure, processing webpage content information to obtain content quality information may include the following operations: extracting image features from image information and text features from text information; processing the image features to obtain image recognition results; processing the text features to obtain text recognition results; and obtaining content quality information based on the image recognition results and text recognition results.

[0072] For example, a neural network model trained with watermarked sample images can be used to identify image features and determine whether a watermark exists on an image on a target webpage.

[0073] For example, text features can be converted into images, and then optical character recognition (OCR) models can be used to recognize the text and determine whether the characters are stacked together.

[0074] According to embodiments of this disclosure, for the text information of the top 10 recommended products on a target webpage, the number of high-quality products among the recommended products can be determined by analyzing information such as price and quality. The text content quality is then determined based on the number of high-quality products. For example: if the number of high-quality products is greater than 10, the text content quality can be 3. If the number of high-quality products is greater than or equal to 5 but less than 10, the text content quality can be 2. If the number of high-quality products is greater than 0 but less than 5, the text content quality can be 1. If the number of high-quality products is equal to 0, the text content quality can be determined to be 0.

[0075] Figure 4 The illustration shows a schematic diagram of obtaining webpage quality assessment results according to an embodiment of the present disclosure.

[0076] like Figure 4 As shown, in Example 400, image features 4302 of image information 4301 are extracted, and image features 4302 are processed to obtain image recognition result 4303. Text features 4305 of text information 4304 are extracted, and text features 4305 are processed to obtain text recognition result 4306. Based on image recognition result 4303 and text recognition result 4306, content quality information 4307 is obtained. Based on content quality information 4307 and relevance information 4308, webpage quality assessment result 4309 is obtained.

[0077] According to embodiments of this disclosure, since the quality of webpage content affects the user's experience and thus the user's feedback on the target webpage, evaluating webpage quality by combining the relevance of the target search term to the webpage content with the quality of the webpage content can indirectly evaluate webpage quality from the perspective of the user's experience of using the webpage, thereby improving the accuracy of webpage quality evaluation.

[0078] Since both the relevance of the target search term to the webpage content and the quality of the webpage content contain discrete feature information, the discrete feature information can be processed separately to improve the accuracy of the evaluation.

[0079] According to embodiments of this disclosure, obtaining a webpage quality assessment result based on relevance information and content quality information may include the following operations:

[0080] Discretized relevance information is determined from relevance information. Discretized content quality information is determined from content quality information. The discretized relevance information and discretized content quality information are processed to obtain a first evaluation result. The relevance information and content quality information are processed to obtain a second evaluation result. Based on the first and second evaluation results, the webpage quality evaluation result is obtained.

[0081] According to embodiments of this disclosure, discretized relevance information can characterize categorical relevance information. For example, the relevance of the target search term content: if the target search term is relevant to the content, the relevance is 1; if the target search term is irrelevant to the content, the relevance is 0.

[0082] According to embodiments of this disclosure, discretized content quality information can characterize the quantity of recommended products on a target webpage. For example, if the number of recommended products on a target webpage is greater than or equal to 10, it indicates high content quality, and the corresponding content quality feature can be 1. If the number of recommended products on a target webpage is less than 10, it indicates low content quality, and the corresponding content quality feature can be 0.

[0083] According to embodiments of this disclosure, since relevance information and webpage content information are multi-dimensional, a fusion model can be used to enhance the accuracy of webpage quality prediction. A webpage quality prediction model can be obtained by jointly training a logistic regression model and a decision tree model. The logistic regression model can be an LR model, and the decision tree model can be a LightGBM model. The decision tree model can be used to process discrete relevance information and discrete webpage content information, while the logistic regression model can be used to process all relevance information and webpage content information. The evaluation results obtained from each model are weighted and summed to obtain the final webpage quality evaluation result. Using a fusion model to process relevance information and webpage content information can enhance the model's memory function while improving its generalization ability.

[0084] For example, using a decision tree model to process discretized relevance and content quality information, the first evaluation result could be "0.8". Using a logistic regression model to process the relevance and content quality information, the second evaluation result could be "0.6". The average of the first and second evaluation results, "0.7", can be taken as the webpage quality evaluation result.

[0085] According to embodiments of this disclosure, obtaining a webpage quality assessment result based on a first assessment result and a second assessment result may include the following operations: determining a first weight and a second weight; and obtaining the webpage quality assessment result based on the first assessment result, the first weight, the second assessment result, and the second weight.

[0086] For example, the first weight could be 0.4, and the second weight could be 0.6. The first evaluation result could be 0.8, the second evaluation result could be 0.6, and the resulting webpage quality evaluation result could be 0.34.

[0087] Figure 5 The illustration shows a schematic diagram of obtaining webpage quality assessment results according to other embodiments of the present disclosure.

[0088] like Figure 5 As shown, in embodiment 500, a second evaluation result 5305 is obtained based on content quality information 5301 and relevance information 5303. Discretized content quality information 5302 is determined from the content quality information 5301. Discretized relevance information 5304 is then determined from the relevance information 5303. A first evaluation result 5306 is obtained based on the discretized content quality information 5302 and the discretized relevance information 5304. Based on the first weight and the second weight, and according to the first evaluation result 5306 and the second evaluation result 5305, a webpage quality evaluation result 5307 is obtained.

[0089] According to embodiments of this disclosure, by evaluating the discrete relevance information and content quality information separately and then combining the evaluation results of all relevance information and content quality information, the evaluation accuracy can be improved.

[0090] Figure 6 A flowchart illustrating a method for training a webpage quality assessment model according to an embodiment of the present disclosure is shown.

[0091] like Figure 6 As shown, in Example 600, the training method includes operations S610 to S640.

[0092] In operation S610, based on the content information of the sample webpage and the sample search terms used to search for the sample webpage, the sample relevance information between the sample search terms and the sample webpage is obtained.

[0093] In operation S620, the sample relevance information and sample webpage content information are processed using a preset model to obtain the webpage quality assessment results of the sample webpages.

[0094] In operation S630, based on the objective loss function, loss information is obtained according to the webpage quality assessment results and the labels of the sample webpages.

[0095] When operating the S640, the model parameters of the preset model are adjusted based on the loss information to obtain the webpage quality assessment model.

[0096] According to the embodiments of this disclosure, the meanings of the content information of the sample webpage, the sample search terms, and the sample relevance are the same as those of the content information of the target webpage, the target search terms, and the relevance between the target search terms and the webpage content information in the webpage quality assessment method, and will not be repeated here.

[0097] According to embodiments of this disclosure, the preset model may include two sub-models: one sub-model may be a logistic regression sub-model, and the other sub-model may be a decision tree sub-model. The logistic regression sub-model can input discretized features into the decision tree sub-model, and input both discretized and non-discrete features into the logistic regression sub-model. The weights and outputs of the two models are then used as the webpage quality evaluation result during the training process.

[0098] According to embodiments of this disclosure, the target loss function can be a cross-entropy loss function. The webpage quality assessment results during training are substituted into the cross-entropy loss function to obtain the loss value. If the loss value does not reach the convergence condition, the model parameters of the preset model are adjusted until the loss value reaches the convergence condition, thus obtaining the trained webpage quality assessment model.

[0099] According to embodiments of this disclosure, the decision tree sub-model can employ the LightGBM model. The logistic regression sub-model can employ the LR model. Using a fusion model to process relevance information and webpage content information can enhance the model's memory function while improving its generalization ability.

[0100] According to embodiments of this disclosure, a webpage quality assessment model trained using sample relevance information and sample webpage content information does not rely on posterior feedback signals such as user browsing behavior, but directly utilizes the content of the target webpage and target search terms, making the assessment results more objective.

[0101] Figure 7 A block diagram of a webpage quality assessment apparatus according to an embodiment of the present disclosure is shown schematically.

[0102] like Figure 7 As shown, in embodiment 700, the webpage quality assessment device may include: an acquisition module 710, a first relevance analysis module 720, and a first assessment module 730.

[0103] The acquisition module 710 is used to acquire the webpage content information of the target webpage and the target search terms used to search for the target webpage in response to receiving a quality assessment request for the target webpage.

[0104] The first relevance analysis module 720 is used to obtain the relevance information between the target search term and the target webpage based on the webpage content information and the target search term.

[0105] The first evaluation module 730 is used to obtain the webpage quality evaluation result based on relevance information and webpage content information.

[0106] According to embodiments of this disclosure, the first relevance analysis module 720 may include: a text relevance analysis submodule, an image relevance analysis submodule, and an acquisition submodule. The text relevance analysis submodule is used to obtain text relevance information based on text information and target search terms. The image relevance analysis submodule is used to obtain image relevance information based on image information and target search terms. The acquisition submodule is used to obtain relevance information based on the text relevance information and the image relevance information.

[0107] According to embodiments of this disclosure, the text relevance analysis submodule may include a first extraction unit and a first acquisition unit. The first extraction unit is used to extract a first semantic feature of the text information and a second semantic feature of the target search term. The first acquisition unit is used to process the first and second semantic features to obtain text relevance information.

[0108] According to embodiments of this disclosure, the image relevance analysis submodule may include: a second extraction unit, a first recognition unit, and a second acquisition unit. The second extraction unit is used to extract image features from image information. The first recognition unit is used to process the image features to obtain an image recognition result. The second acquisition unit is used to obtain image relevance information based on the image recognition result and the target search term.

[0109] According to embodiments of this disclosure, the first evaluation module may include a content quality analysis submodule and an evaluation submodule. The content quality analysis submodule is used to process webpage content information to obtain content quality information. The evaluation submodule is used to obtain a webpage quality evaluation result based on relevance information and content quality information.

[0110] According to embodiments of this disclosure, the content quality analysis submodule may include: a third extraction unit, a second recognition unit, a third recognition unit, and a third acquisition unit. The third extraction unit is used to extract image features from image information and text features from text information. The second recognition unit is used to process the image features to obtain an image recognition result. The third recognition unit is used to process the text features to obtain a text recognition result. The third acquisition unit is used to obtain content quality information based on the image recognition result and the text recognition result.

[0111] According to embodiments of this disclosure, the evaluation submodule may include: a first determining unit, a second determining unit, a first evaluating unit, a second evaluating unit, and a fourth obtaining unit. The first determining unit is used to determine discretized relevance information from relevance information. The second determining unit is used to determine discretized content quality information from content quality information. The first evaluating unit is used to process the discretized relevance information and the discretized content quality information to obtain a first evaluation result. The second evaluating unit is used to process the relevance information and the content quality information to obtain a second evaluation result. The fourth obtaining unit is used to obtain a webpage quality evaluation result based on the first evaluation result and the second evaluation result.

[0112] According to embodiments of this disclosure, the fourth obtaining unit may include a determining subunit and an obtaining subunit. The determining subunit is used to determine a first weight and a second weight. The obtaining subunit is used to obtain a webpage quality evaluation result based on a first evaluation result, the first weight, the second evaluation result, and the second weight.

[0113] Figure 8 A block diagram of a training apparatus for a webpage quality assessment model according to an embodiment of the present disclosure is shown schematically.

[0114] like Figure 8 As shown, in embodiment 800, the training device for the webpage quality assessment model may include: a second relevance analysis module 810, a second assessment module 820, an acquisition module 830, and an adjustment module.

[0115] The second relevance analysis module 810 is used to obtain the sample relevance information between the sample search terms and the sample webpages based on the content information of the sample webpages and the sample search terms used to search for the sample webpages.

[0116] The second evaluation module 820 is used to process the sample relevance information and the content information of the sample web pages using a preset model to obtain the web page quality evaluation results of the sample web pages.

[0117] The module 830 is used to obtain loss information based on the target loss function, the webpage quality assessment results of the sample webpages, and the tags of the sample webpages.

[0118] The adjustment module 840 is used to adjust the model parameters of the preset model based on the loss information to obtain the webpage quality assessment model.

[0119] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0120] According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described above.

[0121] According to embodiments of the present disclosure, a non-transitory computer-readable storage medium stores computer instructions, wherein the computer instructions are used to cause a computer to perform the method described above.

[0122] According to an embodiment of this disclosure, a computer program product includes a computer program that, when executed by a processor, implements the method described above.

[0123] Figure 9 A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0124] like Figure 9 As shown, device 900 includes a computing unit 901, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 902 or a computer program loaded from storage unit 908 into random access memory (RAM) 903. RAM 903 may also store various programs and data required for the operation of device 900. The computing unit 901, ROM 902, and RAM 903 are interconnected via bus 904. Input / output (I / O) interface 905 is also connected to bus 904.

[0125] Multiple components in device 900 are connected to I / O interface 905, including: input unit 906, such as keyboard, mouse, etc.; output unit 907, such as various types of monitors, speakers, etc.; storage unit 908, such as disk, optical disk, etc.; and communication unit 909, such as network card, modem, wireless transceiver, etc. Communication unit 909 allows device 900 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0126] The computing unit 901 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as web page quality assessment methods or web page quality assessment model training methods. For example, in some embodiments, the web page quality assessment methods or web page quality assessment model training methods can be implemented as computer software programs tangibly contained in a machine-readable medium, such as storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed on device 900 via ROM 902 and / or communication unit 909. When the computer program is loaded into RAM 903 and executed by the computing unit 901, one or more steps of the web page quality assessment methods or web page quality assessment model training methods described above can be performed. Alternatively, in other embodiments, the computing unit 901 may be configured by any other suitable means (e.g., by means of firmware) to perform a web page quality assessment method or a training method for a web page quality assessment model.

[0127] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0128] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0129] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0130] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0131] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0132] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, distributed system servers, or servers incorporating blockchain technology.

[0133] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0134] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A webpage quality assessment method, comprising: In response to receiving a quality assessment request for a target webpage, the system obtains the webpage content information of the target webpage and the target search terms used to find the target webpage; wherein, the webpage content information includes image information and text information displayed on the target webpage; Based on the webpage content information and the target search term, relevance information between the target search term and the target webpage is obtained; wherein, the relevance information characterizes the semantic and category relevance between the target search term and the webpage content; the relevance information includes: image relevance between the target search term and the image information and text relevance between the target search term and the text information; The webpage content information is processed to obtain content quality information; the content quality information includes image quality information and text quality information. Discretized correlation information is determined from the correlation information; Discretized content quality information is determined from the content quality information; The discretized relevance information and the discretized content quality information are processed to obtain a first evaluation result; The relevance information and content quality information are processed to obtain a second evaluation result; and Based on the first evaluation result and the second evaluation result, the webpage quality evaluation result is obtained.

2. The method according to claim 1, wherein, The step of obtaining the relevance information between the target search term and the target webpage based on the webpage content information and the target search term includes: Based on the text information and the target search term, text relevance information is obtained; Based on the image information and the target search term, image relevance information is obtained; and The relevance information is obtained based on the text relevance information and the image relevance information.

3. The method according to claim 2, wherein, The step of obtaining text relevance information based on the text information and the target search term includes: Extract the first semantic features of the text information and the second semantic features of the target search term; and The first semantic feature and the second semantic feature are processed to obtain the text relevance information.

4. The method according to claim 2, wherein, The step of obtaining image relevance information based on the image information and the target search term includes: Extract image features from the image information; The image features are processed to obtain the image recognition result; and Based on the image recognition results and the target search term, the image relevance information is obtained.

5. The method according to claim 1, wherein, The webpage content information includes image information and text information. The process of processing the webpage content information to obtain content quality information includes: Extract the image features of the image information and the text features of the text information; The image features are processed to obtain the image recognition result; The text features are processed to obtain the text recognition result; and The content quality information is obtained based on the image recognition results and the text recognition results.

6. The method according to claim 1, wherein, The step of obtaining the webpage quality assessment result based on the first assessment result and the second assessment result includes: Determine the first weight and the second weight; and The webpage quality assessment result is obtained based on the first assessment result, the first weight, the second assessment result, and the second weight.

7. A method for training a webpage quality assessment model, comprising: Based on the content information of the sample webpage and the sample search terms used to search for the sample webpage, sample relevance information between the sample search terms and the sample webpage is obtained; wherein, the content information of the sample webpage includes sample image information and sample text information displayed on the sample webpage; the sample relevance information characterizes the semantic relevance and category relevance between the sample search terms and the content of the sample webpage; the sample relevance information includes: the image relevance between the sample search terms and the sample image information and the text relevance between the sample search terms and the sample text information; The content information of the sample webpage is processed using a preset model to obtain sample content quality information; discrete sample relevance information is determined from the sample relevance information; discrete sample content quality information is determined from the sample content quality information; the discrete sample relevance information and the discrete sample content quality information are processed to obtain a first sample evaluation result; the sample relevance information and the sample content quality information are processed to obtain a second sample evaluation result; and the webpage quality evaluation result of the sample webpage is obtained based on the first sample evaluation result and the second sample evaluation result. Based on the objective loss function, loss information is obtained according to the webpage quality assessment results and tags of the sample webpages; and Based on the loss information, the model parameters of the preset model are adjusted to obtain the webpage quality assessment model.

8. A webpage quality assessment device, comprising: The acquisition module is used to, in response to receiving a quality assessment request for a target webpage, acquire webpage content information of the target webpage and target search terms for searching the target webpage; wherein, the webpage content information includes image information and text information displayed on the target webpage; The first relevance analysis module is used to obtain relevance information between the target search term and the target webpage based on the webpage content information and the target search term; wherein, the relevance information characterizes the semantic and category relevance between the target search term and the webpage content; the relevance information includes: image relevance between the target search term and the image information and text relevance between the target search term and the text information; and The first evaluation module is used to process the webpage content information to obtain content quality information; the content quality information includes image quality information and text quality information. Discretized correlation information is determined from the correlation information; Discretized content quality information is determined from the content quality information; The discretized relevance information and the discretized content quality information are processed to obtain a first evaluation result; The relevance information and content quality information are processed to obtain a second evaluation result; and Based on the first evaluation result and the second evaluation result, the webpage quality evaluation result is obtained.

9. The apparatus according to claim 8, wherein, The webpage content information includes image information and text information, and the first relevance analysis module includes: The text relevance analysis submodule is used to obtain text relevance information based on the text information and the target search term; The image relevance analysis submodule is used to obtain image relevance information based on the image information and the target search term; and The obtaining submodule is used to obtain the relevance information based on the text relevance information and the image relevance information.

10. The apparatus according to claim 9, wherein, The text relevance analysis submodule includes: The first extraction unit is used to extract the first semantic features of the text information and the second semantic features of the target search term; and The first obtaining unit is used to process the first semantic feature and the second semantic feature to obtain the text relevance information.

11. The apparatus according to claim 9, wherein, The image correlation analysis submodule includes: The second extraction unit is used to extract image features from the image information; The first recognition unit is used to process the image features to obtain an image recognition result; and The second obtaining unit is used to obtain the image relevance information based on the image recognition result and the target search term.

12. The apparatus according to claim 9, wherein, The webpage content information includes image information and text information, and the content quality analysis submodule includes: The third extraction unit is used to extract the image features of the image information and the text features of the text information; The second recognition unit is used to process the image features to obtain the image recognition result; The third recognition unit is used to process the text features to obtain the text recognition result; and The third obtaining unit is used to obtain the content quality information based on the image recognition result and the text recognition result.

13. The apparatus according to claim 12, wherein, The fourth acquisition unit includes: Determine sub-units for determining the first and second weights; and A sub-unit is obtained to obtain the webpage quality assessment result based on the first assessment result, the first weight, the second assessment result, and the second weight.

14. A training device for a webpage quality assessment model, comprising: The second relevance analysis module is used to obtain sample relevance information between the sample search terms and the sample webpage based on the content information of the sample webpage and the sample search terms used to search for the sample webpage; wherein, the content information of the sample webpage includes sample image information and sample text information displayed on the sample webpage; the sample relevance information characterizes the semantic relevance and category relevance between the sample search terms and the sample webpage content; the sample relevance information includes: the image relevance between the sample search terms and the sample image information and the text relevance between the sample search terms and the sample text information; The second evaluation module is used to process the content information of the sample webpage using a preset model to obtain sample content quality information; determine discretized sample relevance information from the sample relevance information; determine discretized sample content quality information from the sample content quality information; process the discretized sample relevance information and the discretized sample content quality information to obtain a first sample evaluation result; process the sample relevance information and the sample content quality information to obtain a second sample evaluation result; and obtain a webpage quality evaluation result of the sample webpage based on the first sample evaluation result and the second sample evaluation result. The acquisition module is used to obtain loss information based on the target loss function, the webpage quality assessment results of the sample webpages, and the tags of the sample webpages; and The adjustment module is used to adjust the model parameters of the preset model based on the loss information to obtain a webpage quality assessment model.

15. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

16. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-8.

17. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-8.

Images