Website information collection and publishing method and system

Abstract

The application discloses a website information collection and release method and system. The method comprises the following steps: constructing a keyword mapping table, wherein at least one keyword and at least one website corresponding to each keyword are stored in the keyword mapping table; inputting release information; screening recommended keywords; generating a title based on the recommended keywords; screening recommended websites; generating a release template based on the recommended websites; optimizing the release information based on the release template; and releasing website information based on the generated title and the optimized release information. The application analyzes the browsing, searching and operation information of historical users in the website searching and browsing process, and compares the information released by an information publisher with the analysis result, so that the information publisher is provided with recommended keywords which can improve the information detection rate and a release template which can improve the information recognition degree, and the information publisher is helped to efficiently release the website.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of website publishing, in particular to a website information collection and publishing method and system. BACKGROUND

[0002] With the development of Internet technology and information technology, various websites emerge in an endless stream. In order to promote and publicize the website, the website publisher usually adopts the way of advertisement to link and recommend in other websites. However, this way not only increases the publishing cost, but also reduces the user experience, and the effect is not ideal.

[0003] At the same time, the website query work is also the focus of major browser search engines. The accuracy and recall rate of website query are usually two aspects that cannot be guaranteed at the same time in the same retrieval technology category. In order to improve the accuracy and recall rate of website query, the existing browser search engines use technical means such as constructing a separate keyword library and an associated keyword library for each website, and improving the semantic recognition accuracy in the retrieval technology. However, all these technologies do not solve the problem from the root, that is, from the beginning of website publishing. SUMMARY

[0004] In order to solve the problems existing in the prior art, the present application provides a website information collection and publishing method, comprising:

[0005] S1: based on the historical information query record, a keyword mapping table and a website mapping table are constructed, the keyword mapping table stores at least one keyword and at least one website corresponding to each keyword; the website mapping table stores at least one website and a website document feature set corresponding to each website;

[0006] S2: input the publishing information;

[0007] S3: based on the similarity of the publishing information and the document of the website in the website mapping table, the keywords corresponding to the websites whose document similarity meets a certain condition are selected from the keyword mapping table as recommended keywords;

[0008] S4: generating a title based on the recommended keywords;

[0009] S5: screening the recommended website based on the similarity of the publishing information and the document of the website in the website mapping table;

[0010] S6: generating a publishing template based on the recommended website;

[0011] S7: optimizing the publishing information based on the publishing template;

[0012] S8: publishing the website information based on the title generated in S4 and the optimized publishing information in S7.

[0013] Each keyword in the keyword mapping table corresponds to at least one website, and each website in the website mapping table is represented by a unique website identifier.

[0014] Specifically, based on the document similarity between the published information and the websites in the website mapping table, keywords corresponding to websites whose document similarity meets specific conditions are selected from the keyword mapping table as recommended keywords, including:

[0015] S31: Obtain the document feature set of the published information based on the published information, and obtain the document similarity between the published information and the website based on the document feature set of the published information and the website document feature set in the website mapping table;

[0016] S32: Filter websites with document similarity greater than a specific value as target websites;

[0017] S33: Based on the unique identifier of the target website, find the corresponding keywords from the keyword mapping table and use them as recommended keywords.

[0018] Wherein, the set of features of the published information document is set A = (α1A1, α2A2, ..., α... m A m );

[0019] The set of website document features is set B = (β1B1, β2B2, ..., β...). n B n );

[0020] Among them, (A1, A2, ..., A m ) are the keyword vectors corresponding to the keywords appearing in the published information, (α1, α2, ..., α3) m ) are keyword vectors (A1, A2, ..., A m The frequency of the corresponding keyword in the published information, where the subscript m indicates the number of keywords appearing in the published information;

[0021] (B1, B2, ..., B) n ) are the keyword vectors corresponding to the keywords appearing in the website documents, (β1, β2, ..., β3) n ) are keyword vectors (B1, B2, ..., B) n The frequency of the corresponding keyword in the website document, where the subscript n represents the number of keywords appearing in the website document;

[0022] By calculating the vector similarity and set similarity between set A and set B, the document similarity between the published information and the website can be obtained.

[0023] The vector similarity between the set A and the set B is determined by the following method:

[0024] The vector of the set A is determined as ;

[0025] The vector of the set B is determined as ;

[0026] The cosine value between the vector of the set A and the vector of the set B is determined as the vector similarity between the set A and the set B.

[0027] The set similarity between the set A and the set B is determined by the following method:

[0028] Based on the same keyword vector appearing in the set A=(α1A1, α2A2, …, α m A m ) and the set B=(β1B1, β2B2, …, β n B n ), a set C=(γ1C1, γ2C2, …, γ j C j ) is constructed.

[0029] Wherein (C1, C2, …, C j ) is the same keyword vector appearing in the set A and the set B, subscript j is the number of the same keyword vector appearing in the set A and the set B, (γ1, γ2, …, γ j ) is the sum of the appearance frequency of the keyword corresponding to the keyword vector (C1, C2, …, C j ) in the published information and the appearance frequency in the website document.

[0030] The set similarity between the set A and the set B is determined by the following formula:

[0031] The set similarity ;

[0032] Wherein, is the vector absolute value of the vector A i , B i , C i .

[0033] Wherein, based on the similarity of the website document in the published information and the keyword mapping table, the recommended website is screened, including:

[0034] S51: obtaining the target website in S32;

[0035] S52: based on the historical information query record, screening the website with high user preference degree from the target website as the recommended website.

[0036] Wherein, based on the historical information query record, the website with high user preference degree is screened from the target website as the recommended website, comprising:

[0037] S521: For each target website, according to the operation information of the historical user, the average positive feedback vector, the positive feedback browsing time, the average negative feedback vector and the negative feedback browsing time are obtained.

[0038] S522: The positive feedback browsing time and the negative feedback browsing time of the target website are normalized and converted to obtain the normalized positive feedback browsing time coefficient and the normalized negative feedback browsing time coefficient.

[0039] S523: The screening coefficient of the target website is obtained by the normalized positive feedback browsing time coefficient, the normalized negative feedback browsing time coefficient, the average positive feedback vector and the average negative feedback vector.

[0040] S524: The target website is screened based on the screening coefficient to obtain the recommended website.

[0041] The application further provides a website information collection and release system for realizing the website information collection and release method.

[0042] The application analyzes the browsing, query and operation information of the historical user in the website query and browsing process, and compares the information published by the information publisher with the analysis result, so as to provide the recommended keyword for the information publisher to improve the information detection rate, and at the same time, provide the release template for the information publisher to improve the information recognition degree, and help the information publisher to efficiently publish the website. DETAILED DESCRIPTION

[0043] In order to further understand the technical solutions and beneficial effects of the application, the technical solutions and the beneficial effects thereof will be described in detail below.

[0044] The website information collection and release method of the application comprises the following steps S1-S7.

[0045] S1: Based on the historical information query record, a keyword mapping table and a website mapping table are constructed, wherein the website mapping table stores the website unique identifier of at least one website and the corresponding website document feature set, and the keyword mapping table stores at least one keyword, the unique identifier of at least one website corresponding to each keyword, and the correlation coefficient of each keyword and each website corresponding thereto.

[0046] The historical information query record is the query operation of the historical user in the process of website information query. A mapping relationship between keywords and websites is constructed based on the keywords selected by the historical user in the query operation and the websites queried. In one embodiment, the website mapping table and the keyword mapping table can be seen from Table 1 and Table 2 shown below.

[0047] Table 1: A website mapping table of the present application

[0048]

[0049] In the above website document feature set, (B 11 , B 21 , …, B n1 ) are respectively the keyword vectors corresponding to the keywords appearing in the documents of website 1, (β 11 , β 21 , …, β n1 ) are respectively the frequencies of the keywords corresponding to the keyword vectors (B 11 , B 21 , …, B n1 ) appearing in the documents of website 1, and subscript n1 represents the number of keywords appearing in the documents of website 1.

[0050] Similarly, (B 12 , B 22 , …, B n2 ) are respectively the keyword vectors corresponding to the keywords appearing in the documents of website 2, (β 12 , β 22 , …, β n2 ) are respectively the frequencies of the keywords corresponding to the keyword vectors (B 12 , B 22 , …, B n2 ) appearing in the documents of website 2, and subscript n2 represents the number of keywords appearing in the documents of website 2; the meanings of the symbols in the feature sets of the documents of the remaining websites are sequentially similar, and will not be described one by one here.

[0051] Table 2: A keyword mapping table of the present application

[0052]

[0053] The mapping relationship between the keywords and the websites has the following meanings: when a user is searching for information, the user is more likely to search for information of a website corresponding to a keyword, for example, when searching for website 1, the user is more likely to search for information of website 1 through keyword 1, keyword 2 or keyword 3; when searching for website 2, the user is more likely to search for information of website 2 through keyword 3 or keyword 4; when searching for website 3, the user is more likely to search for information of website 3 through keyword 2, keyword 3 or keyword 4; when searching for website 4, the user is more likely to search for information of website 4 through keyword 2; when searching for website 5, the user is more likely to search for information of website 5 through keyword 3 or keyword 4; and when searching for website 6, the user is more likely to search for information of website 6 through keyword 1.

[0054] Therefore, by constructing the mapping relationship between the keywords and the websites, for a certain type of website, a keyword that is more suitable for the user's searching habit can be found, and the keyword can be provided as a reference for a website information publisher, so that the website information publisher generates a theme of a website to be published based on the keyword, thereby improving the possibility that the website is searched by a user.

[0055] In the present application, for a certain keyword, the corresponding website can be determined based on the number of times that a historical user opens a website under the keyword and the browsing time, and the specific determination method can be: taking a website whose opening times under the keyword exceed a certain value as the corresponding website, or taking a website whose browsing time exceeds a certain value as the corresponding website; or setting weights for the opening times and the browsing time respectively, and performing a comprehensive calculation based on the weights to filter; or performing a descending order sorting on the opening times or the browsing time or the comprehensive calculation value of the opening times and the browsing time of the historical user under the keyword for a certain website, and taking a predetermined number of websites as the corresponding websites, and the above methods are easily understood by those skilled in the art, and will not be described here.

[0056] The opening times and the browsing time of the historical user under the keyword for a certain website directly represent the correlation coefficient between the certain website and the keyword, and the correlation coefficient is determined by:

[0057] Correlation coefficient = ( + ) / 2;

[0058] wherein, represents the number of times that the historical user opens a certain website corresponding to the keyword, represents the sum of the number of times that the historical user opens all websites corresponding to the keyword; represents the browsing time of the historical user on a certain website corresponding to the keyword, The total browsing time length of the historical user on all websites corresponding to the keyword.

[0059] Taking keyword 1 as an example, assuming that the historical user opens website 1 5000 times and opens website 6 20000 times when querying website information based on keyword 1, the browsing time length on website 1 is 20000 hours, and the browsing time length on website 6 is 30000 hours; then the correlation coefficient of website 1 and keyword 1 = (5000-20000) / 2=0.3, and the correlation coefficient of website 6 and keyword 1 = (20000-30000) / 2=0.7.

[0060] S2: The publisher inputs the publishing information, which is initial information intended by the publisher to be published on the website.

[0061] S3: Based on the document similarity between the publishing information and the website in the website mapping table, the keywords corresponding to the websites whose document similarity meets a specific condition are screened from the keyword mapping table as recommended keywords.

[0062] The screening of the recommended keywords is the core point of the application, including the following steps S31-S33.

[0063] S31: Obtain a publishing information document feature set based on the publishing information, and obtain the document similarity between the publishing information and the website based on the publishing information document feature set and the website document feature set in the website mapping table.

[0064] The publishing information document feature set is represented as set A=(α1A1, α2A2, …, αmAm), where (A1, A2, …, Am) are keyword vectors corresponding to keywords appearing in the publishing information, (α1, α2, …, αm) are the appearance frequencies of the keywords corresponding to the keyword vectors (A1, A2, …, Am) in the publishing information, and subscript m represents the number of keywords appearing in the publishing information. m A m

[0065] The website document feature set is represented as set B=(β1B1, β2B2, …, βnBn), where (B1, B2, …, Bn) are keyword vectors corresponding to keywords appearing in the website document, (β1, β2, …, βn) are the appearance frequencies of the keywords corresponding to the keyword vectors (B1, B2, …, Bn) in the website document, and subscript n represents the number of keywords appearing in the website document. n B n

[0066] where (A1, A2, …, Am) are keyword vectors corresponding to keywords appearing in the publishing information, (α1, α2, …, αm) are the appearance frequencies of the keywords corresponding to the keyword vectors (A1, A2, …, Am) in the publishing information, and subscript m represents the number of keywords appearing in the publishing information. m m m

[0067] (B1, B2, …, Bn) are keyword vectors corresponding to keywords appearing in the website document, (β1, β2, …, βn) are the appearance frequencies of the keywords corresponding to the keyword vectors (B1, B2, …, Bn) in the website document, and subscript n represents the number of keywords appearing in the website document. n n ​​​​​​​​​​) are keyword vectors (B1, B2, …, B n ) are the frequencies of the corresponding keywords in the website document, and subscript n represents the number of keywords appearing in the website document.

[0068] Then, the vector similarity and the set similarity of set A and set B are calculated to obtain the similarity of the publishing information and the website document.

[0069] In the present application, the vector similarity is determined as follows: the vector of set A is determined as the vector of set B is determined as The cosine value of the vector of set A and the vector of set B is determined as the vector similarity of set A and set B.

[0070] In the present application, the set similarity is determined as follows: based on the same keyword vectors appearing in set A=(α1A1, α2A2, …, α m A m ) and set B=(β1B1, β2B2, …, β n B n ), set C=(γ1C1, γ2C2, …, γ j C j ) is constructed; wherein, (C1, C2, …, C j ) are the same keyword vectors appearing in set A and set B, subscript j is the number of the same keyword vectors appearing in set A and set B, (γ1, γ2, …, γ j ) are the sum of the frequencies of the corresponding keywords in the publishing information and in the website document, respectively, of the keyword vectors (C1, C2, …, C j ).

[0071] The set similarity of set A and set B is determined by the following formula:

[0072] Set similarity ;

[0073] Wherein, are the vector absolute values of vectors A i , B i , C i .

[0074] S32: screening the website with the document similarity greater than a specific value as a target website; when screening, a minimum threshold number of target websites can be determined: when the number of screened target websites is less than the threshold, the specific value of the document similarity can be appropriately reduced to ensure that the required number of target websites can be screened; when the number of screened target websites is too much to exceed the threshold, the specific value of the document similarity can be appropriately increased to ensure that the number of screened target websites is controlled within a reasonable range.

[0075] S33: based on the unique identifier of the target website, finding the corresponding keyword from the keyword mapping table as the recommended keyword.

[0076] Therefore, in the process of determining the recommended keyword, the present application considers that the number of websites corresponding to each keyword is not one, so two mapping tables are set; when determining the recommended keyword, the corresponding website is first determined based on the website document feature set, and then the recommended keyword is determined from the keyword mapping table based on the unique identifier of the website; in the above implementation process, in the keyword mapping table, for each keyword, only the unique identifier of the corresponding website needs to be stored, and the website document feature set does not need to be stored; since the number of bytes occupied by the unique identifier of the website is much smaller than the number of bytes occupied by the website document feature set, by constructing two mapping tables, for each specific website, only the website document feature set needs to be stored once in the website mapping table, and multiple storage of the website document feature set for each corresponding keyword in the keyword mapping table is not required, which greatly saves the storage space and resource consumption.

[0077] In the preferred embodiment of the present application, when providing the recommended keywords to the publisher, the recommended coefficients of the recommended keywords are also provided to the publisher to more clearly understand the degree of fit of each recommended keyword to the published information in the query habits of the public users.

[0078] The determination method of the recommended coefficient includes the following steps S34-S37.

[0079] S34: determining all target websites screened in S32 as a target website set.

[0080] S35: determining all website sets corresponding to the recommended keywords as a sub-website set; and determining the intersection of the sub-website set of each recommended keyword and the target website set as a sub-target website set.

[0081] S36: for each recommended keyword, the similarity coefficient, the total correlation coefficient and the keyword weight of the recommended keyword are determined respectively.

[0082] The similarity coefficient is determined by the following method: obtaining the similarity of the publishing information and the documents of all websites in the sub-target website set, taking the average value, and obtaining the similarity coefficient; the similarity coefficient directly indicates the similarity between the sub-target website set of the recommended keyword and the publishing information, and the higher the similarity, the closer the similarity coefficient is to 1.

[0083] The total correlation coefficient is determined by the following method: obtaining the correlation coefficient of all websites in the sub-target website set, taking the average value, and obtaining the total correlation coefficient; the total correlation coefficient directly indicates the correlation between the websites in the sub-target website set and the recommended keyword, that is, the user's tendency and willingness to query the related websites in the sub-target website set through the recommended keyword, more specifically, in all websites corresponding to the recommended keyword, the user's tendency and willingness to query the websites in the sub-target website set through the recommended keyword; the higher the tendency and willingness of the user to query the websites in the sub-target website set through the recommended keyword, the closer the total correlation coefficient is to 1.

[0084] The keyword weight is determined by the following method: the ratio of the total number of websites in the sub-target website set to the total number of websites in the target website set is determined as the keyword weight; the keyword weight directly indicates the generalization degree of the recommended keyword to the target website set, and in the technical aspect, it belongs to a supplement and strengthening of the total correlation coefficient; as described above, the total correlation coefficient only represents the tendency and willingness of the user to query the websites in the sub-target website set through the recommended keyword from all websites corresponding to the recommended keyword, and cannot represent the tendency and willingness of the user to query the websites in the target website set through the recommended keyword from the entire website set (which does not belong to any keyword), and cannot directly represent the importance of the recommended keyword in the entire recommended keyword set; in other words, in actual operation, there is more than one recommended keyword corresponding to each publishing information, and different recommended keywords represent the key information in the publishing information from different angles, and these representation angles have different importance; therefore, the keyword weight is introduced at the same time, the higher the keyword weight, the greater the overlap between the website set corresponding to the recommended keyword and the target website set, and the more it can be explained that the recommended keyword itself is specially used to represent the key information in the target website set, and the higher the fit degree with the publishing information; the closer the keyword weight is to 1, the more important the representation angle of the recommended keyword to the key information in the publishing information.

[0085] S37: The recommendation coefficient of each recommended keyword is determined by the following formula:

[0086] The recommendation coefficient = the similarity coefficient The total correlation coefficient The keyword weight.

[0087] Still taking Table 1 and Table 2 as examples, assuming that the target websites screened based on step S32 are website 1, website 2, website 3 and website 4, and the document similarity of the publishing information with website 1, website 2, website 3 and website 4 is 0.6, 0.8, 0.9 and 0.8 respectively, then:

[0088] The target website set is (website 1, website 2, website 3, website 4);

[0089] The recommended keywords involve keyword 1, keyword 2, keyword 3 and keyword 4.

[0090] The sub-target website set of keyword 1 is (website 1), the sub-target website set of keyword 2 is (website 1, website 3, website 4), the sub-target website set of keyword 3 is (website 1, website 2, website 3), and the sub-target website set of keyword 4 is (website 2, website 3).

[0091] For keyword 1: similarity coefficient = 0.6, total correlation coefficient = 0.3, keyword weight = 1 / 4 = 0.25, recommended coefficient = 0.6 0.3 0.25 = 0.045;

[0092] For keyword 2: similarity coefficient = (0.6+0.9+0.8) / 3 = 0.76, total correlation coefficient = (0.1+0.3+0.6) / 3 = 0.33, keyword weight = 3 / 4 = 0.75, recommended coefficient = 0.76 0.33 0.75 = 0.188;

[0093] For keyword 3: similarity coefficient = (0.6+0.8+0.9) / 3 = 0.76, total correlation coefficient = (0.6+0.1+0.2) / 3 = 0.3, keyword weight = 3 / 4 = 0.75, recommended coefficient = 0.76 0.3 0.75 = 0.171;

[0094] For keyword 4: similarity coefficient = (0.8+0.9) / 2 = 0.85, total correlation coefficient = (0.5+0.3) / 2 = 0.4, keyword weight = 2 / 4 = 0.5, recommended coefficient = 0.85 0.4 0.5 = 0.170.

[0095] Due to the similarity coefficient, the total correlation coefficient and the keyword weight determination method provided by the application, the value range of the three is 0 to 1, and the closer to 1, the more the recommendation degree of the recommended keyword can be represented from different angles; therefore, when the similarity coefficient, the total correlation coefficient and the keyword weight are integrated to determine the recommendation coefficient, the three are directly multiplied, without normalization processing, without considering the difference in data dimension between the three, and without considering the difference in the representation direction of the recommendation degree caused by the numerical size of the three, so that the determination of the recommendation coefficient is more convenient and the data processing is simpler.

[0096] S4: The publisher generates the title of the website based on the provided recommended keyword.

[0097] Specifically, the publisher can balance the content theme and the recommendation weight of the recommended keyword, and select the keyword that is most suitable for the required published information to generate the title of the website.

[0098] Therefore, the application matches the website with high similarity to the published information provided by the publisher as the target website, and finds the keyword that can improve the probability of the target website being queried based on the operation of the historical user, and provides the keyword to the publisher as the recommended keyword, thereby improving the probability of the published information being queried by the user subsequently.

[0099] S5: Filtering the recommended website based on the document similarity between the published information and the website in the keyword mapping table.

[0100] Specifically, the recommended website is further filtered from the target website obtained in the above step S32, and the website that better meets the user's preference is filtered from the target website as the recommended website based on the user's preference.

[0101] A specific method for filtering the recommended website includes the following steps S51-S54.

[0102] S51: For each target website, the average positive feedback vector, the positive feedback browsing time, the average negative feedback vector and the negative feedback browsing time are obtained according to the operation information of the historical user.

[0103] For a specific target website, the average positive feedback vector is the average of all positive feedback vectors of the target website, the positive feedback vector is obtained by the positive feedback operation of the historical user in browsing the website, the positive feedback operation can be understood as an operation that can represent the approval of the target website, such as like, forward, collect, share and positive comment. The positive feedback browsing time is the total browsing time related to the positive feedback operation.

[0104] Correspondingly, for a specific target website, the average negative feedback vector is the mean of all negative feedback vectors of the target website, the negative feedback vector is obtained through the negative feedback operation of the historical user in browsing the website, the negative feedback operation can be understood as the operation that can represent the disapproval of the target website, such as opposing (thumb down button), shielding and negative comments, etc. The negative feedback browsing time is the total browsing time related to the negative feedback operation.

[0105] If there is no operation involved in the website browsing process or both positive feedback operation and negative feedback operation are involved, the browsing is not counted.

[0106] For example, assuming that a specific target website is browsed 10000 times, 5000 browsing records involve positive feedback operation and 2000 browsing records involve negative feedback operation in the 10000 browsing processes; in the positive feedback operation, 1000 likes, 2000 forwards, 3000 collections, 4000 shares and 1500 positive comments are involved, at the same time, the browsing time involved in the positive feedback operation is 3000 hours, then the average positive feedback vector is Z= (1000 / 5000, 2000 / 5000, 3000 / 5000, 4000 / 5000, 1500 / 5000)= (0.2, 0.4, 0.6, 0.8, 0.3), and the positive feedback browsing time is 3000 hours; in the negative feedback operation, 1000 opposing, 500 shielding and 1500 negative comments are involved, at the same time, the browsing time involved in the negative feedback operation is 500 hours, then the average negative feedback vector is F= (1000 / 2000, 500 / 2000, 1500 / 2000)= (0.5, 0.25, 0.75), and the negative feedback browsing time is 500 hours.

[0107] S52: Normalize the positive feedback browsing time and the negative feedback browsing time of the target website to obtain the normalized positive feedback browsing time coefficient and the normalized negative feedback browsing time coefficient respectively.

[0108] For a specific target website, the calculation formula of the normalized positive feedback browsing time coefficient and the normalized negative feedback browsing time coefficient of the target website as the current target website is:

[0109] J= (L 当 -L min ) / (L max -L min );

[0110] K= (S 当 -S min ) / (S max -S min ).

[0111] wherein, J represents a normalized positive feedback browsing time coefficient of the current target website, L 当 represents a positive feedback browsing time of the current target website, L max represents a maximum positive feedback browsing time of all target websites participating in statistics, L min represents a minimum positive feedback browsing time of all target websites participating in statistics;

[0112] K represents a normalized negative feedback browsing time coefficient of the current target website, S 当 represents a negative feedback browsing time of the current target website, S max represents a maximum negative feedback browsing time of all target websites participating in statistics, S min represents a minimum negative feedback browsing time of all target websites participating in statistics.

[0113] S53: determining the screening coefficient by the following formula:

[0114] X=(J / )-(K / );

[0115] wherein, X is the screening coefficient of the target website, respectively, and p and q are the vector dimensions of the average positive feedback vector E and the average negative feedback vector F.

[0116] Therefore, the finally determined screening coefficient not only considers the positive feedback operation and the negative feedback operation of the historical user to the target website, but also considers the operation time of the positive feedback operation and the negative feedback operation of the historical user. And on this basis, the application considers that when a user approves a website, he will usually stay for a long time, and when he does not approve a website, he will close as soon as possible; therefore, if the positive feedback browsing time and the negative feedback browsing time are directly applied to the calculation formula of the screening coefficient, the result will inevitably be that the negative feedback browsing time is significantly less than the positive feedback browsing time, thereby weakening the decision effect of the negative feedback operation on the screening coefficient. Therefore, the application proposes to normalize the positive feedback browsing time and the negative feedback browsing time, so that the normalized positive feedback browsing time coefficient and the normalized negative feedback browsing time coefficient are both between 0 and 1. At the same time, the ratio of the average positive feedback vector to the square root of its vector dimension, and the ratio of the vector absolute value of the average negative feedback vector to the square root of its vector dimension are also between 0 and 1. In this way, the influence of the browsing time and the operation on the screening coefficient is just at the same level, which can provide a more accurate and objective screening coefficient.

[0117] S54: screening the target website based on the screening coefficient to obtain the recommended website.

[0118] S6: generating a publishing template based on the screened recommended website; specifically, a multi-document summarization method can be used for the recommended website to segment and cluster the documents of the recommended website, and the nearest core segment to the cluster center is selected from each segmented cluster, and the publishing template is obtained after combining multiple core segments; the publishing template at this time represents a generally recognized publishing template by historical users, and providing the publishing template to the information publisher helps to generate a webpage information with high recognition.

[0119] S7: the information publisher optimizes the publishing information based on the publishing template.

[0120] S8: publishing website information based on the title generated in S4 and the optimized publishing information in S7.

[0121] Therefore, the application analyzes the browsing, searching and operation information of historical users in the website searching and browsing process, combines big data technology and semantic recognition technology, and compares the information published by the information publisher with the analysis result, to provide the information publisher with recommended keywords that can improve the information detection rate, and at the same time, provide the information publisher with a publishing template that can improve the information recognition degree, to help the information publisher to efficiently publish the website.

[0122] Although the application has been described by the above preferred embodiments, it is not intended to limit the protection scope of the application, and any person skilled in the art can make various changes and modifications to the above embodiments without departing from the spirit and scope of the application, which are still within the protection scope of the application, therefore, the protection scope of the application is defined by the claims.

Claims

1. A method for collecting and distributing information on a website, characterized by, The method comprises the following steps: S1: constructing a keyword mapping table and a website mapping table based on historical information query records, the keyword mapping table storing at least one keyword and at least one website corresponding to each keyword, and the website mapping table storing at least one website and a website document feature set corresponding to each website; S2: inputting publishing information; S3: screening keywords corresponding to websites whose document similarity with the publishing information meets a specific condition from the keyword mapping table based on the document similarity of the publishing information with the websites in the website mapping table, and taking the keywords as recommended keywords, comprising: S31: obtaining a publishing information document feature set based on the publishing information, and obtaining the document similarity of the publishing information with the websites based on the publishing information document feature set and the website document feature set in the website mapping table; The publishing information document feature set is set A=(α1A1, α2A2, …, αnAn), and the publishing information document feature set is set B=(β1B1, β2B2, …, βmBm). m A m ) The website document feature set is set B=(β1B1, β2B2, …, β n B n ) wherein (A1, A2, …, A m ) are keyword vectors corresponding to keywords appearing in the published information, (a1, a2, …, a m ) are frequencies of the keywords corresponding to the keyword vectors (A1, A2, …, A m ) appearing in the published information, and subscript m represents the number of keywords appearing in the published information; (B1, B2, …, B n ) are keyword vectors corresponding to keywords appearing in the website document, (b1, b2, …, b n ) are frequencies of the keywords corresponding to the keyword vectors (B1, B2, …, B n ) appearing in the website document, and subscript n represents the number of keywords appearing in the website document. The document similarity of the publishing information with the websites is obtained by calculating the vector similarity and the set similarity of set A and set B; S32: screening websites whose document similarity is greater than a specific value as target websites; S33: finding keywords corresponding to the target websites from the keyword mapping table based on the unique identifiers of the target websites, and taking the keywords as recommended keywords; S4: generating a title based on the recommended keywords; S5: screening recommended websites based on the document similarity of the publishing information with the websites in the website mapping table, comprising: S51: obtaining the target websites in S32; and S52: screening websites with high user preference from the target websites based on historical information query records, and taking the websites as recommended websites; S6: generating a publishing template based on the recommended websites; S7: optimizing the publishing information based on the publishing template; S8: publishing website information based on the title generated in S4 and the publishing information optimized in S7.

2. The website information collection and distribution method of claim 1, wherein, Each keyword in the keyword mapping table corresponds to at least one website, and at least one website in the website mapping table is represented by a unique website identifier.

3. The website information collection and distribution method of claim 1, wherein, The method for determining the vector similarity of set A and set B is as follows: Will The vector that is determined to be set A; Will The vector is determined to be part of set B; The cosine value of the vector of set A and the vector of set B is determined as the vector similarity of set A and set B.

4. The website information collection and distribution method of claim 1, wherein, The method for determining the set similarity of set A and set B is as follows: Based on the same keyword vectors appearing in set A = (a1A1, a2A2, …, a m A m ) and set B = (b1B1, b2B2, …, b n B n ), a set C = (g1C1, g2C2, …, g j C j ) is constructed. wherein (C1, C2, …, C j ) is the same keyword vector of set A and set B, subscript j is the number of the same keyword vector of set A and set B, (γ1, γ2, …, γ j ) is the sum of the occurrence frequency of the keyword corresponding to the keyword vector (C1, C2, …, C j ) in the published information and the occurrence frequency in the website document, respectively. The set similarity of set A and set B is determined by the following formula: Set similarity ; wherein, are the vector absolute values of vectors A i , B i , C i , respectively.

5. The website information collection and distribution method of claim 1, wherein, The websites with high user preference are screened from the target websites based on historical information query records, and the websites are taken as recommended websites, comprising: S521: for each target website, obtaining its average positive feedback vector, positive feedback browsing time, average negative feedback vector and negative feedback browsing time according to the operation information of historical users; S522: normalizing and converting the positive feedback browsing time and the negative feedback browsing time of the target website to obtain a normalized positive feedback browsing time coefficient and a normalized negative feedback browsing time coefficient, respectively; S523: obtaining a screening coefficient of the target website by using the normalized positive feedback browsing time coefficient, the normalized negative feedback browsing time coefficient, the average positive feedback vector and the average negative feedback vector; S524: screening the target websites based on the screening coefficient to obtain recommended websites.

6. A website information collection and distribution system, characterized by, The method is used for implementing the website information collection and publishing method in any one of claims 1-5.

Images