Personalized recommendation preview method based on webpage search

Abstract

According to the personalized recommendation preview method based on webpage search, a first display area and a second display area are provided to form a search page, the first display area is used for displaying a webpage searched based on search information input by a user, and the second display area is used for displaying the search information input by the user. According to the method, firstly, the recommended user is roughly screened out for the current user through the browsing duration and the browsing times of the user on the webpage, then on the basis of the browsing duration and the browsing times, the operation condition of the user on the webpage is further considered, and the webpage is recommended to the user through a finer screening index. And the webpage which can best meet the requirements of the user is recommended for the user, so that the recommendation accuracy is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of web page search recommendation, and in particular to a personalized recommendation preview method based on web page search. Background Art

[0002] The rapid development of the Internet and the increase of massive network resources have brought many conveniences to users' work and life. How to accurately and comprehensively obtain the information needed by users from massive network resources is the goal that major search engines strive to achieve. Existing web page searches provide various algorithms to provide users with the required resources. Most of these algorithms are aimed at improving the accuracy of searches based on keywords and users' historical search information; however, the existing search methods are only single-mindedly targeted at users' current search needs, and the web pages provided only involve a single search need. However, in different scenarios such as actual life, work, and study, when users search for network resources each time, they usually do not have only one search need. Often, based on the development of the search work, other related searches are required. If users can be provided with targeted searches for current keywords and related searches based on current keywords at the same time, the user-friendliness of the search page can be greatly improved, and the search efficiency can also be greatly improved. Summary of the invention

[0003] In order to solve the shortcomings of the prior art, the present invention provides a personalized recommendation preview method based on web page search, comprising the following steps: Step S1: obtaining keywords from the search information input by the current user, performing web page search based on the keywords, and displaying the searched web pages in the first display area of ​​the search page; Step S2: Determine the web page tag, which is the category to which the web page belongs; Step S3: obtaining a current user web page set, wherein the current user web page set is a set of all web pages involving the web page tags determined in step S2 and browsed by the current user within a first predetermined time period; Step S4: determining associated users based on the current user web page set; wherein the associated users are users whose web page sets involving the web page tags determined in step S2 and browsed within the second predetermined time period overlap with the current user web page set; Step S5: determining the operation similarity between the current user and the associated user based on the historical operation set of the current user and the historical operation set of the associated user; Step S6: based on the operation similarity, filter the recommended users from the associated users; Step S7: Based on the historical operation set of the current user and the historical operation set of the recommending user, a recommended web page is generated, and the recommended web page is displayed in the second display area of ​​the search page.

[0004] The historical operation set of the current user includes each web page in the current user web page set and the web page weight corresponding to each web page; The historical operation set of the associated user includes each web page in the associated user web page set and the web page weight corresponding to each web page; The historical operation set of the recommended user includes each web page in the recommended user web page set and the web page weight corresponding to each web page; The associated user web page set is a set of all web pages related to the web page tag determined in step S2 and browsed by the associated user within the second predetermined time period; The recommended user web page set is a set of all web pages involving the web page tags determined in step S2 that are recommended for the user to browse within the second predetermined time period; The weight of a web page is determined by the browsing time and number of views.

[0005] The similarity between the operations of the current user and the associated user is determined by the following formula: ; Where W represents the operation similarity between the current user and the associated user; A represents the current user's web page collection, including web pages a1, a2, a3...a n , web pages a1, a2, a3...a n The web page weights are λ1, λ2, λ3...λ n ; B represents the set of web pages associated with the user, including web pages b1, b2, b3, ... b m , web pages b1, b2, b3...b m The web page weights are β1, β2, β3...β m ; n and m represent the number of web pages in the current user web page set and the associated user web page set, respectively; A∩B means that there are overlapping web page sets between the current user web page set and the associated user web page set; tλ and tβ represent the corresponding web page weights of the web pages in the overlapping web page set in the historical operation set of the current user and the historical operation set of the associated user respectively; It means obtaining the overlapping web page set of the current user's web page set and the associated user's web page set, and summing up the corresponding web page weights of all web pages in the overlapping web page set in the current user's historical operation set and the corresponding web page weights in the associated user's historical operation set.

[0006] Wherein, the first predetermined time period includes a predetermined time period before the current user performs the current search operation and / or a specific time period after the current user performs the current search operation; The second predetermined time period is a predetermined time period before the current user performs the current search operation.

[0007] The following steps are also included: after the current user performs the search operation, at every predetermined interval, the web page tag in step S2 is updated according to the web page browsing situation of the user on the search page, and steps S3 to S7 are re-executed according to the updated web page tag.

[0008] After the current user performs the current search operation, before performing a web page tag update operation according to the web page browsing situation of the user on the search page, the web page tag is determined by the keyword obtained in step S1.

[0009] The method further includes the following steps: after the current user performs the current search operation, at predetermined intervals, the display ratio of the first display area and the second display area on the search page is updated according to the web page browsing situation of the user on the search page.

[0010] The present invention firstly roughly screens out recommended users for the current user based on the browsing time and number of times the user browses the web page, and then further considers the user's operation of the web page based on the browsing time and number of times, and recommends the web page that best meets the user's needs through more refined screening indicators, thereby greatly improving the accuracy of the recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The figure is a flow chart of the logic flow of the personalized recommendation preview method based on web page search of the present invention. DETAILED DESCRIPTION

[0012] In order to have a further understanding of the technical solution and beneficial effects of the present invention, the technical solution and beneficial effects produced by the present invention are described in detail below with reference to the accompanying drawings.

[0013] The present invention aims to provide users with search convenience in the following application scenarios: in most search scenarios, users may have more than one search requirement. As information collection work continues to unfold, users may need to constantly adjust keywords, etc., to obtain more comprehensive information: in daily search scenarios, taking the search for "tops" as an example, after the user has browsed enough links to tops, he may need to browse the shoes, pants, etc. that match them; taking the search for a certain movie as an example, after the user types in the movie title and browses enough movie reviews and movie information related to the movie, he may also need to browse the corresponding theater information; in addition to daily search scenarios, when retrieving information in study and work, as the query work unfolds, after browsing and studying some web pages, the user may have new problems and need to understand and learn new knowledge; therefore, the user needs to continuously perform related searches. In most related search scenarios, related web pages mostly belong to the same category. For example, search pages for tops, shoes, pants, etc. basically belong to the "shopping category"; search pages for movie reviews, movie information, theaters, etc. basically belong to the "entertainment" category; and when searching in study and work, the information retrieved in one search work basically belongs to the same subject.

[0014] Therefore, the original intention of the present invention is to, in addition to providing a collection of web pages for the search information typed by the user, recommend other web pages related to the search information to the user based on the user's historical operation information, help the user to diverge and expand search ideas, provide more comprehensive customized information, and continuously update the search page based on the user's actual search operations.

[0015] Based on the above invention concept, the search page of the present invention provides a first display area and a second display area; wherein the first display area is used to display web pages searched based on the search information entered by the user; the second display area is used to display web pages recommended to the user after helping the user to perform an associated search based on the search information entered by the user; the search method of the web pages in the first display area and the second display area is combined with Figure 1 As shown, it is described in detail below.

[0016] 1. First display area In the present invention, a conventional web page search is performed based on the search information currently input by the user, and the searched web pages are displayed in the first display area of ​​the search page.

[0017] Specifically, the search information currently input by the user may be a specific keyword or a search sentence. The search sentence may be segmented by a word segmentation tool such as query, and after obtaining the keywords, a web page search may be performed.

[0018] In the present invention, the webpage displayed in the first display area can be searched based on existing conventional search technology.

[0019] 2. Second display area The web page displayed in the second display area is searched by the following method.

[0020] 1. Determine the web page tags Web page tags represent the categories to which web pages belong, such as shopping, technology, news, air ticket booking, etc. Web page tags can be constructed in advance before the invention is implemented, and the vertical domains of web pages are matched with corresponding tags, and the corresponding relationship between the two is stored in a mapping table of the server.

[0021] In the present invention, after the user enters the search information, the web page tag can be determined based on the keywords obtained from the search information.

[0022] Specifically, the probability of each keyword belonging to each tag can be estimated based on the Bayesian classification method, the probability of each keyword belonging to each tag is summed up, and the tag with the largest probability value is determined as the webpage tag.

[0023] For example, if keyword 1, keyword 2, and keyword 3 are obtained for the search information typed by the user, and there are N preset web page tags, the probability sets of keyword 1 to keyword 3 belonging to the N web page tags are respectively:

[0024] Sum the above three probability sets, and we get the following probability set: ; The label corresponding to the maximum numerical probability in the probability set can be determined as the web page label.

[0025] 2. Get the current user's web page collection The current user web page set is a set of all web pages involving the determined web page tag browsed by the current user within the first predetermined time period.

[0026] Specifically, the first predetermined time period may be a predetermined time period before executing the current search operation. The so-called "before executing the current search operation" refers to a time period of a specific length before the user types in the search information. For example, after the user types in "jeans", based on step 1 above, the web page label is obtained as "shopping", more specifically "shopping-clothing", then a collection of all web pages related to "shopping-clothing" browsed by the user within a predetermined time (e.g., two months) before executing the current search operation is obtained and determined as the current user web page collection.

[0027] Specifically, the set of all web pages involving the determined web page tag browsed by the user in the first predetermined time period can be obtained through the user's login name or identity identification information for identifying the user. If the user does not browse any web page involving the determined web page tag in the first predetermined time period, the above step 1 can be re-executed to select the web page tag corresponding to the second highest probability value from the probability set as the determined web page tag, and so on.

[0028] 3. Determine associated users Determine the associated user based on the current user web page set; wherein the associated user is a user whose web page set involving the determined web page tag browsed within the second predetermined time period overlaps with the current user web page set; correspondingly, the web page set involving the determined web page tag browsed by the associated user within the second predetermined time period can be determined as the associated user web page set. The number of associated users is more than two.

[0029] Specifically, the second predetermined time period may be a predetermined time period before executing the current search operation. Similar to the above, “before executing the current search operation” refers to a time period of a specific length before the user types in the search information. The length of the first predetermined time period may be the same as or different from the length of the second predetermined time period.

[0030] For example: assuming that the current user has browsed a web page 1 under the category of "Shopping - Clothing" within two months before performing this search operation, and another user has also browsed a web page 1 under the category of "Shopping - Clothing" within two months before the current user performs this search operation, then this user can be determined as an associated user of the current user.

[0031] In actual implementation, some specific screening mechanisms can be set for associated users, including but not limited to: (1) determining a web page set that overlaps between the associated user web page set and the current user web page set; the ratio of the number of web pages in the overlapping web page set to the number of web pages in the associated user web page set should be greater than a predetermined value; (2) determining a web page set that overlaps with the web page set of the associated user and the web page set of the current user; the ratio of the total browsing time of the associated user on the web pages in the overlapping web page set to the total browsing time of the associated user on the web pages in the associated user web page set should be greater than a predetermined value; (3) Determine the web page set that overlaps between the associated user web page set and the current user web page set; the ratio of the number of web pages in the overlapping web page set to the total number of web pages in the associated user web page set and the current user web page set should be greater than a predetermined value.

[0032] In this way, it is ensured that the selected related users have a certain relevance, while reducing the data volume and the subsequent work burden.

[0033] 4. Filter recommended users Based on the historical operation set of the current user and the historical operation set of the associated users, the operation similarity between the current user and the associated users is determined; based on the operation similarity, the recommended users are filtered from the associated users.

[0034] The historical operation set of the current user includes each web page in the current user web page set and the web page weight and operation vector corresponding to each web page; its storage form is (λ1, a1, ;λ2,a2, ;λ3,a3, ; ...; λ n , a n , ), where a1, a2, a3, ... a n Represents web pages, λ1, λ2, λ3...λ n Respectively represent web pages a1, a2, a3...a n The web page weight, , , … Respectively represent the current user's web pages a1, a2, a3...a n The meaning, determination method and use of the operation vector will be described in detail below.

[0035] The historical operation set of the associated user includes each web page in the associated user web page set and the web page weight and operation vector corresponding to each web page; the storage form is (β1, b1, ; β2, b2, ; β3, b3, ; ...; β m 、b m , ), where b1, b2, b3, ... b m Represents web pages, β1, β2, β3...β m Respectively represent web pages b1, b2, b3...b m The web page weight, , , … Respectively represent the associated users' views on web pages b1, b2, b3...b m The meaning, determination method and use of the operation vector will be described in detail below.

[0036] Among them, the weight of a web page is determined by browsing time and browsing times. A better way to determine it is as follows: Q=Q inital + (x1+T ime / C) / x2+(y1+C) / y2; Among them, Q is the weight of the web page (representing the above λ1, λ2, λ3...λ n and β1, β2, β3, ... β m ), Q inital is to initialize the web page weight, x1, x2, y1 and y2 are adjustment factors, T ime is the total browsing time of the corresponding user (current user or associated user) on the corresponding webpage, and C is the total number of browsing times of the corresponding user (current user or associated user) on the corresponding webpage.

[0037] By introducing adjustment factors x1, x2, y1 and y2, a buffer is provided for the influence of the average browsing time and total browsing times on the weight of the web page, thereby avoiding large fluctuations in the weight of the web page due to changes in the average browsing time and total browsing times, and enhancing the stability of the recommendation preview method.

[0038] In the present invention, the operation similarity between the current user and the associated user is determined by the following formula: ; Where W represents the operation similarity between the current user and the associated user; A represents the current user's web page collection, including web pages a1, a2, a3...a n ; B represents the set of web pages associated with the user, including web pages b1, b2, b3, ... b m ; n and m represent the number of web pages in the current user web page set and the associated user web page set, respectively; A∩B means that there are overlapping web page sets between the current user web page set and the associated user web page set; tλ and tβ represent the corresponding web page weights of the web pages in the overlapping web page set in the historical operation set of the current user and the historical operation set of the associated user respectively; It means obtaining the overlapping web page sets of the current user's web page set and the associated user's web page set, and summing up the corresponding web page weights of all web pages in the overlapping web page set in the current user's historical operation set and the corresponding web page weights in the associated user's historical operation set.

[0039] For example, suppose the current user's historical operation set is (0.3, web page 1, ; 0.2, web page 2, ; 0.5, web page 3, ; 0.1, web page 4, ; 0.6, web page 5, ; 0.7, web page 6, ), the historical operation set of a certain associated user is (0.5, web page 3, ; 0.2, web page 4, ; 0.6, web page 5, ; 0.8, web page 7, ; 0.5, web page 8, ; 0.9, web page 9, ); then in the historical operation set of the current user, the sum of the webpage weights of all webpages = 0.3 + 0.2 + 0.5 + 0.1 + 0.6 + 0.7 = 2.4; in the historical operation set of this associated user, the sum of the webpage weights of all webpages = 0.5 + 0.2 + 0.6 + 0.8 + 0.5 + 0.9 = 3.5; the sum of the webpage weights of the overlapping webpage sets in the current user's webpage set and the associated user's webpage set, i.e., webpage 3, webpage 4, and webpage 5 = (0.5 + 0.1 + 0.6) + (0.5 + 0.2 + 0.6) = 2.5; the operation similarity between the current user and this associated user = 2.5 / (2.4 + 3.5 - 2.5) = 0.74.

[0040] The meaning of the above formula is: determine the overlapping web page sets in the current user's web page set and the associated user's web page set, and solve the ratio of the sum of the web page weights in the overlapping web page set to the sum of the web page weights of the non-overlapping web page sets in the current user's web page set and the associated user's web page set; because the web page weight integrates the number of views and average browsing time information of the relevant web pages, the ratio value obtained through the above calculation process can intuitively reflect the similarity of the preferences of the current user and the associated user when browsing the web pages under the same web page tag: the higher the ratio, the more the current user and the associated user are inclined to browse the same web pages when browsing the web pages under the same web page tag, and the closer the preferences for different web pages are; based on this associated user, subsequent web page recommendations will be able to recommend web pages that better meet the needs of the current user.

[0041] Therefore, in the present invention, the operation similarity between the current user and the associated users is determined based on the historical operation set of the current user and the historical operation set of the associated users; based on the operation similarity, the principle of screening recommended users from the associated users is: according to the order of operation similarity from low to high, the associated users with low operation similarity are eliminated in sequence, and several associated users with the highest operation similarity are left for the subsequent generation of recommended web pages.

[0042] 5. Generate recommended web pages Based on the historical operation set of the current user and the historical operation set of the recommending user, a recommended web page is generated, and the recommended web page is displayed in the second display area of ​​the search page.

[0043] The storage form of the historical operation set of the current user has been described above. The recommended users are screened from the associated users. Therefore, the storage form of the historical operation set of the recommended users is the same as the storage form of the historical operation set of the associated users described above.

[0044] In the present invention, based on the operation status of the current user and the recommended user on the relevant web page in the historical operation set, that is, the operation vector, and combined with the web page weight of the relevant web page, a recommended web page is generated. A specific implementation method is as follows: (1) Based on the operation vector of each web page in the historical operation set of the current user, obtain the first absolute value of each web page in the historical operation set of the current user; (2) based on the operation vector of each web page in the historical operation set of the recommending user, obtaining the second absolute value of each web page in the historical operation set of the recommending user; (3) determining a recommended web page set, the recommended web page set including all web pages in the current user web page set and the recommended user web page set; and recommending that the set of all web pages involving the determined web page tag browsed by the user within the second predetermined time period be the recommended user web page set; (4) The recommendation weight value of each web page in the recommended web page set is determined by the following formula: ; Among them, T is the recommended weight value of the web page; λ i is the webpage weight corresponding to the webpage in the historical operation set of the current user, J1 is the first absolute value of the operation vector corresponding to the webpage in the historical operation set of the current user, if there is no relevant webpage in the historical operation set of the current user, then λ i and J1 are both 0; β j is the webpage weight corresponding to the webpage in the historical operation set of the recommended user, J2 is the second absolute value of the operation vector corresponding to the webpage in the historical operation set of the recommended user, if there is no relevant webpage in the historical operation set of the recommended user, then β j and J2 are both set to 0.

[0045] (5) Based on the recommendation weight value, a number of web pages are selected from the recommended web page set as the final recommended web pages, and displayed in the second display area of ​​the search page.

[0046] In the present invention, the operation vector is used to represent what kind of operation the user has performed on the web page during the process of browsing the web page, such as whether to like, comment, forward, collect, etc. The dimension of the operation vector is the number of types of the above operation types, and the value of each dimension can be 1 or 0 to indicate whether the user has performed the relevant operation during the process of browsing the web page. For example: Assume that the operation vector has four dimensions, representing the like, comment, forward, and collect operations respectively. If a user has liked and commented on a web page, but has not forwarded or collected it, then the operation vector of the user on the web page is (1, 1, 0, 0), and the absolute value of the operation vector is: .

[0047] For the generation step of the recommended web pages, a specific example is given as follows: Assume that the current user's historical operation set involves web pages 1, 2, 3, 4, and 5, and their respective web page weights are 0.1, 0.2, 0.3, 0.4, and 0.5, respectively. The current user performs the following operations on each web page while browsing web pages 1, 2, 3, 4, and 5: Web page 1: Like, comment, forward, and collect; Web page 2: Like and comment; Web page 3: Like, comment, and forward; Web page 4: Like, comment, and favorite; Web page 5: forward, collect; The historical operation set of a certain recommended user involves web page 3, web page 4, web page 6, web page 7, and web page 8, and their respective web page weights are 0.9, 0.4, 0.6, 0.7, and 0.8. The recommended user performs the following operations on each web page while browsing web page 3, web page 4, web page 6, web page 7, and web page 8: Web page 3: No operation; Web page 4: Like and comment; Web page 6: Comments, forwarding; Web page 7: Like, comment, and forward; Web page 8: forward and collect.

[0048] The storage forms of the current user's historical operation set and the recommended user's historical operation set are as follows: (0.1, web page 1, (1, 1, 1, 1); 0.2, web page 2, (1, 1, 0, 0); 0.3, web page 3, (1, 1, 1, 0); 0.4, web page 4, (1, 1, 0, 1); 0.5, web page 5, (0, 0, 1, 1)); (0.9, web page 3, (0, 0, 0, 0); 0.4, web page 4, (1, 1, 0, 0); 0.6, web page 6, (0, 1, 1, 0); 0.7, web page 7, (1, 1, 1, 0); 0.8, web page 8, (0, 0, 1, 1)).

[0049] The recommended weights of each web page in the recommended web page set are as follows:

[0050] Afterwards, in descending order of recommendation weight values, several web pages with high recommendation weight values ​​are selected from the recommended web page set as final recommended web pages, and displayed in the second display area of ​​the search page.

[0051] It should be noted that, taking web page 3 as an example, since the recommended user did not perform any operation on the web page, the corresponding “β j The value of "J2" is 0. However, it can be seen from the above that in the historical operation set of the recommended user, the webpage weight corresponding to webpage 3 is as high as 0.9, which means that the recommended user browsed webpage 3 for a long time or browsed it many times. Webpage 3 should be a webpage that is more in line with the needs of the recommended user and should also be a webpage suitable for recommendation to the current user. However, the second absolute value of webpage 3 is 0, resulting in that when calculating the recommendation weight value of the webpage, the webpage weight of the webpage in the historical operation set of the recommended user is not taken into account at all. This may result in the omission of webpages that can provide important information.

[0052] Therefore, in a more preferred embodiment of the present invention, another indicator - the absolute value of the public - is introduced to optimize the recommendation weight value of each web page.

[0053] The details are as follows: T = λ i J1+β j J2; + (λ i +β j ) J / 2 Among them, T, λ i , β j , J1 and J2 have the same meanings as above; J is the public absolute value of the web page, which is determined by averaging the operation vectors of all historical operations of the web page before the current user performs the current search operation (i.e. before typing the search information) to obtain the public operation vector. The absolute value of the public operation vector is the public absolute value. For example, before the current user types the search information, a web page has been browsed 2000 times. In the 2000 browsing records, it has been liked 2000 times, commented 1500 times, forwarded 1000 times, and collected 500 times. Then the public operation vector of the web page is (2000 / 2000, 1500 / 2000, 1000 / 2000, 500 / 2000) = (1, 0.75, 0.5, 0.25).

[0054] After the recommended web pages are screened, they should be checked for duplication with the web pages displayed in the first display area, and the recommended web pages that are the same as the web pages displayed in the first display area should be deleted. This is easy to understand for those skilled in the art and will not be elaborated in detail in the present invention.

[0055] At this point, the web pages displayed in the first display area and the second display area have been screened out: since the first display area searches for web pages based on the user's search information, and the second display area performs associated searches and recommendations based on web page tags, it is just an alternative provided to the user; at the very beginning of the search, the web pages in the first display area should be more in line with the user's needs; therefore, the initial proportion of the first display area in the entire search page should be greater than the proportion of the second display area in the entire search page.

[0056] That is, the present invention first filters out recommended users whose preferences are similar to those of the current user from the associated users. This screening process only considers the browsing time and number of views of the current user and the associated users on the web page, which is equivalent to a coarse screening process; then, based on the operations of the current user and the recommended user on the web page, the recommended web pages are screened out. This screening process, on the basis of the browsing time and number of views, further considers the likes, comments, forwarding, collection and other operations of the current user and the recommended user on the web page. At the same time, the likes, comments, forwarding, collection and other operations of the general users on the web page are introduced. Through more refined screening indicators, the web pages that best meet the needs of users are recommended to users, which greatly improves the accuracy of the recommendations.

[0057] 3. Update In the existing conventional search technology, for the same keyword, the web pages given by the existing search engine mostly involve many different categories to provide users with more comprehensive web page information; however, since the second display area of ​​the present invention recommends web pages based on web page tags, the recommended web pages given only involve one web page tag; under the premise of inaccurate judgment of the category to which the keyword belongs, the recommended web pages may not be what the user needs. For example, for the keyword "A Brief History of Time", it may represent the book "A Brief History of Time" or the movie "A Brief History of Time". When the user needs to search for the book "A Brief History of Time" and some physics knowledge related to the book, and the second display area displays web pages related to the movie "A Brief History of Time", the web pages recommended by the second display area at this time have no value to the user. Therefore, it is necessary to update the recommended web pages in the second display area based on the user's browsing situation on the search page.

[0058] A specific update method is as follows: 1. After the current user performs the search operation, the web page label is updated at predetermined intervals according to the web page browsing situation of the user on the search page.

[0059] In the present invention, the predetermined period can be a time period, such as updating the web page label every five minutes, or it can be a period of the number of times the user browses the web page, such as updating the web page label every five times the user browses the web page; it can also be a period of the user's total browsing time, such as updating the web page label every time the user browses the web page for 20 minutes.

[0060] The updating principles of web page tags may include the following methods: (1) Obtain the web pages browsed by the current user within the predetermined period and the web page tags corresponding to each web page, count the number of web pages corresponding to each web page tag, and take the web page tag with the largest number of web pages as the updated web page tag; (2) Obtain the web pages browsed by the current user within the predetermined period and the web page tags corresponding to each web page, count the total browsing time of the current user on all web pages corresponding to each web page tag, and take the web page tag with the largest total browsing time as the updated web page tag.

[0061] 2. Based on the updated web page tags, the current user web page set, associated users, and recommended users are re-determined in turn, and recommended web pages are generated.

[0062] When determining the current user web page set, it is necessary to obtain all web page sets involving the determined web page tags browsed by the current user within the first predetermined time period; when the current user web page set is first obtained, as described above, the first predetermined time period is the predetermined time period before performing the current search operation (in this case, the first predetermined time period is the same as the second predetermined time period); and after the user performs the current search operation, the user performs a corresponding browsing operation on the web pages on the search page. Therefore, when re-determining the current user web page set, the first predetermined time period may include the following situations: (1) The first predetermined time period includes the predetermined time period before the current user performs the current search operation and the specific time period after the current user performs the current search operation; in this case, the current user's historical browsing behavior before performing the current search operation and the browsing behavior during the current search operation will be considered; (2) The first predetermined time period only includes a specific time period after the current user performs the current search operation. In this case, only the browsing behavior of the current user during the current search operation is considered.

[0063] However, when determining associated users, it is necessary to obtain users whose web page sets involving the determined web page tags and which overlap with the current user's web page set browsed within the second predetermined time period; the second predetermined time period here only includes the predetermined time period before the current user performs this search operation.

[0064] 3. Update the display ratio of the first display area and the second display area As described above, the proportion of the initial first display area in the entire search page should be greater than the proportion of the second display area in the entire search page.

[0065] However, as the user's search progresses, the web pages in the second display area are constantly updated based on the user's browsing behavior during the search operation, and the web pages in the second display area will become more and more in line with the user's needs; therefore, similar to the rules for updating web page tags, at every predetermined interval, it is also necessary to update the display ratio of the first display area and the second display area on the search page based on the user's web browsing behavior on the search page.

[0066] The scheduled update period here may be the same as or different from the scheduled update period of the web page tag.

[0067] Similarly, similar to the web page tag update, the predetermined period here can also be a time period, such as updating every five minutes; it can also be a period of the number of times the user browses the web page, such as updating once every five times the user browses the web page; it can also be a period of the user's total browsing time, such as updating once every 20 minutes when the user browses the web page.

[0068] The updating principle of the display ratio of the first display area and the second display area on the search page may include the following methods: (1) obtaining the number of web pages browsed by the current user in the first display area and the second display area within the predetermined period, and determining the display ratio of the first display area and the second display area on the search page based on the ratio of the number of web pages browsed by the current user in the first display area to the number of web pages browsed in the second display area; (2) Obtaining a time duration for which the current user browses web pages in the first display area and the second display area during the predetermined period, and determining a display ratio of the first display area and the second display area on the search page based on a ratio of a total time duration for which the current user browses web pages in the first display area to a total time duration for which the current user browses web pages in the second display area.

[0069] Therefore, during the user's search process, the present invention continuously adjusts the display of the search page and the recommended web pages based on the user's browsing of the web pages, continuously adapts to the user's needs, and can provide the user with a more humane experience.

[0070] Although the present invention has been described using the above preferred embodiments, they are not intended to limit the scope of protection of the present invention. Any person skilled in the art may make various changes and modifications to the above embodiments without departing from the spirit and scope of the present invention. The scope of protection of the present invention shall be based on the definition of the claims.

Claims

1. A personalized recommendation preview method based on web page search, characterized in that: The steps include: Step S1: obtaining keywords from the search information input by the current user, performing web page search based on the keywords, and displaying the searched web pages in the first display area of ​​the search page; Step S2: Determine the web page tag, which is the category to which the web page belongs; Step S3: obtaining a current user web page set, wherein the current user web page set is a set of all web pages involving the web page tags determined in step S2 and browsed by the current user within a first predetermined time period; Step S4: determining associated users based on the current user web page set; wherein the associated users are users whose web page sets involving the web page tags determined in step S2 and browsed within the second predetermined time period overlap with the current user web page set; Step S5: determining the operation similarity between the current user and the associated user based on the historical operation set of the current user and the historical operation set of the associated user; Step S6: based on the operation similarity, filter the recommended users from the associated users; Step S7: Based on the historical operation set of the current user and the historical operation set of the recommending user, a recommended web page is generated, and the recommended web page is displayed in the second display area of ​​the search page.

2. The personalized recommendation preview method based on web page search according to claim 1, characterized in that: The historical operation set of the current user includes each web page in the current user's web page set and the web page weight corresponding to each web page; The historical operation set of the associated user includes each web page in the associated user web page set and the web page weight corresponding to each web page; The historical operation set of the recommended user includes each web page in the recommended user web page set and the web page weight corresponding to each web page; The associated user web page set is a set of all web pages related to the web page tag determined in step S2 and browsed by the associated user within the second predetermined time period; The recommended user web page set is a set of all web pages involving the web page tags determined in step S2 that are recommended for the user to browse within the second predetermined time period; The weight of a web page is determined by the browsing time and number of views.

3. The personalized recommendation preview method based on web page search according to claim 2, characterized in that: The similarity between the operations of the current user and the associated users is determined by the following formula: ； Where W represents the operation similarity between the current user and the associated user; A represents the current user's web page collection, including web pages a1, a2, a3...a n , web pages a1, a2, a3...a n The web page weights are λ1, λ2, λ3...λ n ; B represents the set of web pages associated with the user, including web pages b1, b2, b3, ... b m , web pages b1, b2, b3...b m The web page weights are β1, β2, β3...β m ; n and m represent the number of web pages in the current user web page set and the associated user web page set, respectively; A∩B means that there are overlapping web page sets between the current user web page set and the associated user web page set; tλ and tβ represent the corresponding web page weights of the web pages in the overlapping web page set in the historical operation set of the current user and the historical operation set of the associated user respectively; It means obtaining the overlapping web page set of the current user's web page set and the associated user's web page set, and summing up the corresponding web page weights of all web pages in the overlapping web page set in the current user's historical operation set and the corresponding web page weights in the associated user's historical operation set.

4. The personalized recommendation preview method based on web page search according to claim 1, characterized in that: The first predetermined time period includes a predetermined time period before the current user performs the current search operation and / or a specific time period after the current user performs the current search operation; The second predetermined time period is a predetermined time period before the current user performs the current search operation.

5. The personalized recommendation preview method based on web page search according to claim 4, characterized in that: The method further includes the following steps: after the current user performs the search operation, at predetermined intervals, the web page tag in step S2 is updated according to the web page browsing status of the user on the search page, and steps S3 to S7 are re-executed according to the updated web page tag.

6. The personalized recommendation preview method based on web page search according to claim 5, characterized in that: After the current user performs the current search operation, before performing a web page tag update operation according to the web page browsing situation of the user on the search page, the web page tag is determined by the keyword obtained in step S1.

7. The personalized recommendation preview method based on web page search according to claim 4, characterized in that: The method further includes the following steps: after the current user performs the current search operation, at predetermined intervals, the display ratio of the first display area and the second display area on the search page is updated according to the web page browsing situation of the user on the search page.

Images