A social platform-oriented tagged information recommendation method

Abstract

The social platform-oriented tagged information recommendation method belongs to the technical field of Internet, and mainly comprises the following steps: establishing an information tag system, establishing a mapping between information and information tags, constructing a demander tag system based on the information tag system, establishing a mapping between demanders and demander tags, matching the demander carrying tags with the information carrying tags, extracting the information with the highest adaptation degree to the demander, and performing information recommendation. The method described in the application realizes personalized recommendation of information demanders based on tags, describes information in the form of tags, realizes tagging of information, grades and weights the tags, improves the management efficiency of information owners, more accurately describes the potential demands of demanders, and improves the experience of demanders.

Description

Technical Field

[0001] This invention belongs to the field of Internet technology, and in particular relates to a tag-based information recommendation method for social platforms. Background Technology

[0002] With the development of internet technology, social media platforms have become important tools for people to obtain information and engage in social interaction. For information owners, such as e-commerce businesses and news writers, social media platforms have great potential and value. However, these information owners are often limited to pushing information to information seekers without understanding their needs, thus failing to meet their personalized information requirements. Therefore, constructing an information recommendation method based on social media platforms that meets the needs of different information seekers is of significant importance. Summary of the Invention

[0003] To address the aforementioned issues, this invention proposes a tag-based information recommendation method for social platforms.

[0004] The technical solution adopted in this invention is as follows:

[0005] A tag-based information recommendation method for social platforms, comprising the following steps:

[0006] The first step is to establish an information tagging system to describe all information, including a set of information-specific attribute tags that provide candidate tags for specific attributes of the information and a set of information audience attribute tags that provide candidate tags for the audience attributes of the information; obtain the URL of the information owner's information and assign it an ID; generate an information tag mapping list based on the information tagging system, and establish a mapping between information IDs and information-specific attribute tags and information audience attribute tags through this list;

[0007] The second step involves establishing a demander tagging system based on the information tagging system. This includes a dynamic demander tagging set that provides candidate tags to describe the demander's potential needs and a static demander tagging set that provides candidate tags to describe the demander's basic personal characteristics. A static tagging mapping list is generated based on this system, and a mapping between demander IDs and static demander tags is established through this list. Interaction data between demanders and information is acquired, and the interaction behavior of demanders is described and analyzed. A dynamic tagging set that best meets the demander's potential needs is generated, and a mapping relationship between the demander ID and this dynamic tagging set is established and updated in real time.

[0008] The third step is to match the dynamic tag group of the demander with the specific attribute tag of the information to obtain a set of information IDs. Then, the audience attribute tag corresponding to the obtained information ID is matched with the static tag of the demander to obtain a set of information IDs suitable for recommendation. Finally, the URL corresponding to the information ID is added to the information recommendation queue for information recommendation.

[0009] Furthermore, the set of information-specific attribute labels in the first step can be described as follows: ST = {st1, st2, ..., st...} i The specific attribute tag set ST of this system consists of a set of specific attribute tag classes, where each tag class st represents a specific attribute tag, st = {stag1, stag2, ..., stag...} j}, where any specific attribute tag stag =<category,range,index,weight,parent> The details are as follows:

[0010] (1) Category represents the category name of the specific attribute tag class to which the tag belongs. All specific attribute tags in the specific attribute tag class have the same category.

[0011] (2) range represents the range of values ​​that the label can take;

[0012] (3) The index represents the index of the tag in its tag class;

[0013] (4) weight represents the weight of the tag. The weight represents the level of detail of the tag. The smaller the value range, the greater the weight of the tag.

[0014] (5) parent represents the index of the parent tag of the tag. If the value range of a tag A covers the value range of another tag B, then the tag A is called the parent tag of the tag B.

[0015] The label weight, or W for short, is calculated as follows: Where i represents the level of the specific attribute tag within its class, k represents the number of parent tags with child tags in level i-1, and m j The number of child tags contained in the j-th parent tag among the k parent tags obtained is represented by c, where c is a constant.

[0016] Furthermore, the set of information audience attribute tags in the first step can be described as follows: TT = {tt1, tt2, ..., tt} kThe audience attribute tag set TT of this system consists of a set of audience attribute tag tree classes tt, where each tag class tt represents a type of audience attribute tag, tt = {ttag1, ttag2, ..., ttag...} p}, where any audience attribute tag ttag =<category,range,index> The details are as follows:

[0017] (1) Category represents the category name of the audience attribute tag class to which the tag belongs. All audience attribute tags in the audience attribute tag class have the same category.

[0018] (2) range represents the range of values ​​that the label can take;

[0019] (3) index represents the index of the tag in its tag class.

[0020] Furthermore, the information in the first step is described by the following binary tuple: I =<id,url> The id is assigned after the information is imported into the information tag system. The id serves as a unique identifier for the information, and a mapping between the information's id and information tags is established based on the information's id. The url is the path of the information, used to obtain the specific content of the information.

[0021] Furthermore, the information tag mapping list in the first step can be defined according to the following model: L = {id, SR, TR}, where id corresponds to the unique identifier id of information I, SR is the mapping between information and specific attribute tags, and TR is the mapping between information and audience attribute tags, as detailed below:

[0022] SR = {sr1, sr2, ..., sr} n The mapping SR between information and specific attribute labels consists of a set of single-mapping tuples sr, where sr =<category,index> The category corresponds to the category name in the information-specific attribute tag stag, and the index corresponds to the tag number in the specific attribute tag stag;

[0023] TR = {tr1, tr2, ..., tr mThe mapping between information and audience attribute tags (TR) consists of a set of single-attribute tuples (tr), where ta =<category,index,importance> The category corresponds to the category name in the information audience attribute tag ttag, the index corresponds to the tag number in the audience attribute tag ttag, and the importance represents the importance of the tag to the information that has established a mapping relationship with it. The value range is 1 to m, and the importance in any mapping tr of the information and audience attribute tag is different. The smaller the value of importance, the higher the importance of the tag.

[0024] Furthermore, the process of establishing the mapping between information IDs and specific information attribute tags and information audience attribute tags, as described in the first step, consists of the following four steps:

[0025] (1) Use the unique identifier id of the information I selected by the information owner as the content of id in the information label mapping list L;

[0026] (2) The information owner’s selection of a specific attribute tag corresponds to a mapping tuple sr in the specific attribute tag mapping SR. For any specific attribute tag selection process of the information owner: the category name of the tag class to which the specific attribute tag stag selected by the information owner belongs is used as the content of the category in the mapping tuple sr in the specific attribute tag mapping SR, and the tag number index of the specific attribute tag stag selected by the information owner is used as the content of the index in the mapping tuple sr in the specific attribute tag mapping SR.

[0027] (3) Each selection of an audience attribute tag by the information owner corresponds to a mapping tuple tr in a specific attribute tag mapping TR. For any selection of an audience tag by the information owner: the category name of the tag class to which the selected audience attribute tag ttag belongs is used as the content of the category in the mapping tuple tr in the audience attribute tag mapping TR; the tag number index of the selected audience attribute tag ttag is used as the content of the index in the mapping tuple tr in the audience attribute tag mapping TR; and the information owner's editing of the importance element of the mapping tuple tr is recorded.

[0028] (4) After the information owner determines all the specific attribute tags and audience attribute tags that need to be added, the importance of each tr in the audience attribute tag mapping TR is converted into an importance score, generating a complete information tag mapping list L. GThe importance score, abbreviated as S, is calculated as follows: Among them, M target The number of tuples in the audience attributes representing the information, where c is a constant.

[0029] Furthermore, the second step of establishing a demander tag system based on the information tag system includes a dynamic demander tag set and a static demander tag set. The demander tag system has the same content as the information tag system. Specifically, the dynamic demander tag set and the static demander tag set have the same content as the information-specific attribute tag set and the information audience attribute tag set, respectively.

[0030] Furthermore, the static tag mapping list in the second step can be defined according to the following model: JL = {id, JR}, where id is the unique identifier of the requester who fills in the list, and JR is the mapping between the requester and the static tag, as detailed below:

[0031] JR = {jr1, jr2, ..., jr} h The mapping JR between demanders and static tags consists of a set of single-mapping tuples jr, where jr =<category,index> The category corresponds to the category name of the static tag class jt to which the static tag belongs, and the index corresponds to the tag number index in the static tag jtag.

[0032] Furthermore, the process of establishing the mapping between the demander ID and the demander's static tags in the second step can be divided into the following three steps:

[0033] (1) Extract the IDs of users who have established a mapping relationship with static tags;

[0034] (2) The user needs to select at least one static tag jtag from each tag class jt in the user's static tag set JT. However, the user can select multiple static tags from the static tag classes that describe the user's preferences. Each selection of a static tag by the user corresponds to a mapping tuple jr in the static tag mapping JR. The process for any static tag selection by the user is as follows: extract the category name of the tag class containing the static tag jtag selected by the user as the content of the category in the mapping tuple jr in the static tag mapping JR, and extract the tag index of the static tag jtag selected by the user as the content of the index in the mapping tuple jr in the static tag mapping JR.

[0035] (3) After the user confirms all the static tags that need to be added, a complete static tag mapping list JL is generated. G .

[0036] Furthermore, the interaction behavior of the user in the second step is defined according to the following model: B =<stagGroup,type,time,num> Where stagGroup is the specific attribute tag group that the information corresponding to this behavior needs to carry, which has been mapped to the information tag. stagGroup = {stag1, stag2, ..., stag...} y}, where type represents the type parameter of the behavior, time represents the time parameter of the behavior, and num represents the number of behaviors of this type, with num defaulting to 1; the time parameter time is calculated as follows: Where Δh represents the difference between the time when the behavior occurs and the current time, in hours, and c is a constant.

[0037] Furthermore, the second step of generating dynamic tag components that best meet the potential needs of users is divided into the following three steps:

[0038] (1) The behavior described by the model is preprocessed according to the following rules: (I) The specific attribute tag group stagGroup carried by the information that has been mapped to the information tag corresponding to behavior B is a group of specific attribute tags stag. Behaviors with the same behavior type parameter type, behavior time parameter time, etc. are regarded as the same type of behavior; (II) A group of behaviors B1, B2, ..., B that belong to the same type of behavior are classified as B1, B2, ..., B2, B3, B4, B5, B6, B7, B8, B9, B1, B2, ..., B9, B1, B2 ... z Merge to obtain behavior class B M =<stagGroup,type,time,num> The stagGroup, type, and time are the common parts of this type of behavior, and num becomes the number of all specific behaviors contained in this behavior class;

[0039] (2) After preprocessing, a set of behavior classes B is obtained. M 1. B M 2、…、B M p The following three steps are then taken:

[0040] (I) Extract all the tag classes st1, st2, ..., st involved in these behavior classes. o Then, each tag class st is analyzed and calculated separately in turn;

[0041] (II) In a specific attribute tag class st, to obtain the specific attribute tag stag that best describes the potential needs of the user in the tag tree, the following three steps are required:

[0042] ① All behavior classes B containing the specific attribute tag stag in the tag class st.M =<stagGroup,type,time,num> Filtered out, for behavior type B M Filter the specific attribute tags (stags) within the specific attribute tag group (stagGroup) and retain only the specific attribute tags (stags) belonging to the current tag class (st). The result is B. MS =<{stag1, stag2,…}, type, time, num>;

[0043] ② Starting from the bottom layer of the tag class `st`, calculate the demand level of each specific attribute tag. Quantify the demand level using the demand score `BW`. The calculation method for demand score `BW` is as follows: Where m represents the B applied to the specific attribute tag stag. MS There are m, where n represents the number of sub-tags of the specific attribute tag stag that are calculated to meet the demand for entering this layer, and W represents the weight of the specific attribute tag stag.

[0044] ③ After calculating the demand weight (BW) for each specific attribute tag (stag) in this layer, for two specific attribute tags in this layer whose demand weights (BW) are relatively close, it indicates that the demanders have a common demand for them. Therefore, it is necessary to proceed to the next layer to continue the calculation and obtain the non-zero behavior weight values ​​(BW1, BW2, ..., BW) in this layer. r The maximum value BW in max Calculate the difference between each value and the maximum value, set a threshold V, and if the difference is less than V, proceed to calculate the demand level BW of the next higher level. Repeat the above calculation process until all non-zero demand levels BW1, BW2, ..., BW of a certain level are calculated. x If there is one and only one value less than V, then the specific attribute tag stag' corresponding to the demand degree is obtained;

[0045] (III) For tag classes st1, st2, ..., st o After calculating each tag individually, a set of specific attribute tags stag'1, stag'2, ..., stag' is obtained. o ;

[0046] (3) The resulting set of specific attribute tags stag'1, stag'2, ..., stag' o Dynamic tags with the same content corresponding to each specific attribute tag are added to the user's dynamic tag group, resulting in the user's dynamic tag group dtag1, dtag2, ..., dtag. o .

[0047] Furthermore, the process of establishing the mapping relationship between the demander and the dynamic tag group in the second step involves the following model of the dynamic tag mapping list: DL = {id, DR}, where id is the unique identifier of the demander, and DR is the mapping between the demander and the dynamic tag, as detailed below:

[0048] DR = {dr1, dr2, ..., dr} o The mapping DR between demanders and dynamic tags consists of a set of single mapping tuples dr, where dr =<category,index> The category corresponds to the category name of the dynamic tag class dt to which the dynamic tag dtag belongs, and the index corresponds to the tag number index in the dynamic tag dtag.

[0049] Furthermore, the process of establishing the mapping relationship between the requester ID and the dynamic tag group in the second step consists of the following three steps:

[0050] (1) Extract the IDs of users who have established a mapping relationship with dynamic tag groups;

[0051] (2) The dynamic tag groups dtag1, dtag2, ..., dtag o Each dynamic tag dtag is processed sequentially. Each processing corresponds to a mapping tuple dr in the dynamic tag mapping DR. For any processing process: extract the category name of the tag class to which the dynamic tag dtag belongs in this processing as the content of the category in the mapping tuple dr in the dynamic tag mapping DR, and extract the tag number index of the dynamic tag dtag as the content of the index in the mapping tuple dr in the dynamic tag mapping DR.

[0052] (3) After all dynamic tags in the dynamic tag group have been processed, a complete dynamic tag mapping list DL is generated. G .

[0053] Furthermore, the process of matching the dynamic tag group of the demander with the specific attribute tags of the information in the third step is divided into the following three steps:

[0054] (1) Obtain the mapping list L between information and information tags. G ={id,SR G TR G}, where SR G ={sr G 1, sr G 2, ..., sr G n} represents the mapping between information and specific attribute tags, SR G There are n mapping tuples sr GThat is, this information is mapped to n specific attribute labels, and the mapping tuple sr G =<category,index> ; Retrieve the generated mapping list of users and dynamic tag groups (DL) G ={id,DR G}, where DR G ={dr G 1, dr G 2, ..., dr G o} represents the mapping between demanders and dynamic tags, DR G There are o mapping tuples dr G That is, a mapping was established between this and o dynamic tags, and the mapping tuple dr G =<category,index> ;

[0055] (2) A specific attribute tag is mapped to a piece of information. The frequency of its occurrence in the dynamic tags of the dynamic tag group is represented by fr. The information is then mapped to the specific attribute tag SR in turn. G Each single-mapped tuple sr G Obtain the mapping between demanders and dynamic tags (DR) G In sequence with DR G Each single-mapped tuple in dr G Comparison. If DR G There is a dr G The content is related to the sr G If the content of the two elements is the same (i.e., the content of the category and index elements is the same), then the frequency fr is incremented by one and added to the next singly mapped tuple sr. G The comparison process continues until the specific attribute tag mapping (SR) is completed. G Each single-mapped tuple sr G The comparison is then used to calculate a dynamic similarity, which characterizes the degree of similarity between the information and the potential needs of the users. Where o represents the mapping between the demander and the dynamic tag DR G The number of mapped tuples;

[0056] (3) Following the steps above, calculate the dynamic similarity DS1, DS2, ..., DS of all information. u A threshold σ is set. If the dynamic similarity DS of a certain piece of information is greater than σ, the ID of that information is retained. This process continues until the dynamic similarity of each piece of information is processed, and finally a set of IDs of information is obtained.

[0057] Furthermore, the process of matching the audience attribute tags corresponding to the obtained information ID with the static tags of the users in the third step is divided into the following three steps:

[0058] (1) Obtain the mapping list L between information and information tags. G ={id,SR G TR G}, where TR G ={tr G 1, tr G 2, ..., tr G m} represents the mapping between information and audience attribute tags, TR G There are n mapping tuples tr G This means that the information is mapped to m audience attribute tags, and the mapping tuple tr G =<category,index,score> ; Retrieve the generated list of users and static tags JL G ={id, JR G}, where JR G ={jr G 1, jr G 2, ..., jr G h} represents the mapping between users and static tags, JR G There are h mapping tuples jr G That is, a mapping was established with h static tags, and the mapping tuple jr G =<category,index> ;

[0059] (2) The occurrence of the audience attribute tags mapped to a piece of information in the static tags mapped to the demander is described by the static weight value jw. The information is then mapped to the audience attribute tags TR in sequence. G Each single-mapped tuple tr in G Get the mapping between users and static tags JR G In turn with JR G Each single-mapped tuple jr G Comparison. If JR G There is a JR in it. G The content and the tr G If the content is the same, meaning the content of the element category and the element index are the same, then the static weight value SW is added to the audience attribute tag single mapping tuple tr. G The value of the element's score is then passed to the next audience attribute tag single-map tuple tr. G The comparison process continues until the audience attribute tag mapping (JR) is complete. G Each single-mapped tuple jr G The comparison is then performed, and finally, a static similarity JS = is calculated to characterize the degree of similarity between the information and the basic personal characteristics of the user. Among them, score sum Map the static label to TR for this information G All mapping tuples in tr G The sum of the scores of the elements;

[0060] (3) Following the steps above, calculate the static similarity JS1, JS2, ..., JS of all the obtained information. e Set threshold If the static similarity of a certain piece of information is greater than JS The ID of this information is retained until the static similarity of each piece of information is processed, and finally a set of IDs of information suitable for recommendation is obtained.

[0061] Beneficial effects:

[0062] This invention categorizes information attributes into specific attributes and audience attributes, constructing corresponding tag sets for each. This provides information owners with a means to describe information from both the information itself and the audience's perspective using tags, facilitating multi-faceted analysis and classification. Simultaneously, based on the specific attribute tags and audience attribute tags, this invention constructs dynamic and static tag sets for users, ensuring a content-based correspondence between information tags and user tags. This establishes a connection between information tags and user tags, improving system resource utilization efficiency and reducing tag redundancy to some extent. Compared to traditional information recommendation methods, this invention's tag-based approach achieves personalized recommendations for information users, and by grading and weighting tags, it improves the information owner's management efficiency, more accurately describes the potential needs of users, and enhances the user experience. Attached image description:

[0063] Figure 1 This is a block diagram of the overall system structure of the present invention.

[0064] Figure 2 This is a schematic diagram illustrating the structure of the information-specific attribute tag set ST of the present invention.

[0065] Figure 3 This is a schematic diagram of the information-specific attribute tag class st tag hierarchy of the present invention.

[0066] Figure 4 This is a schematic diagram illustrating the composition of the information audience attribute tag set TT of the present invention.

[0067] Figure 5 This is a schematic diagram illustrating the process of establishing a mapping between the information ID and specific information attributes and information audience attribute tags in this invention.

[0068] Figure 6This is a schematic diagram illustrating the relationship between the information tagging system and the demander tagging system of the present invention.

[0069] Figure 7 This is a schematic diagram illustrating the process of establishing a mapping between the demander ID and the demander static tag for this invention.

[0070] Figure 8 This is a schematic diagram illustrating the process of solving the specific attribute tag (stag) that best meets the potential needs of users according to the present invention.

[0071] Figure 9 This is a schematic diagram illustrating the process of establishing a mapping between the demander ID and the demander dynamic tag group in this invention.

[0072] Figure 10 This is a flowchart of the dynamic similarity DS calculation process of the present invention.

[0073] Figure 11 This is a flowchart of the static similarity JS calculation process of the present invention. Detailed Implementation

[0074] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with specific embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the scope of the present invention.

[0075] Example 1: Overall Structure of a Tag-Based Information Recommendation System for Social Platforms

[0076] The overall structural block diagram of a tag-based information recommendation system for social platforms according to the present invention is as follows: Figure 1As shown, the entire system comprises six modules: an information module, a tag library module, an information tag module, a user tag module, a tag matching module, and an information push module. The information module indirectly obtains the specific content of the information by acquiring its URL and assigns an ID as its unique identifier. Simultaneously, this module provides a front-end platform for users to interact with the information and records their interaction behavior. The tag library module provides information tags and user tags. The information tag module contains a set of information-specific attribute tags and a set of information audience attribute tags. This module provides information owners with an information tag mapping list, allowing them to add tags to their information. The user tag module contains a set of static user tags and a set of dynamic user tags. It provides information users with a static tag mapping list, allowing them to add static tags to themselves. Simultaneously, this module obtains user interaction data from the interaction module, describes and analyzes this data, generates a set of information-specific attribute tags, and then queries the user dynamic tag set to find dynamic tags with the same content as this set of information-specific attribute tags, using these as the user's dynamic tags. The tag matching module is responsible for matching the user's request tags with the information tags of the information, extracting the ID of the matched information, and outputting it to the information push module. The information push module receives the information ID from the tag matching module, then retrieves the corresponding information URL from the information module based on the information ID and pushes it to the user who requested the information.

[0077] Example 2: Composition of the Information-Specific Attribute Tag Set ST

[0078] The structure of the information-specific attribute tag set ST is as follows: Figure 2 As shown, the set of information-specific attribute labels ST = {st1, st2, ..., st...} i The specific attribute tag set ST of this system consists of a set of specific attribute tag classes, where each tag class st represents a specific attribute tag, st = {stag1, stag2, ..., stag...} j}, where any specific attribute tag stag =<category,range,index,weight,parent> The details are as follows:

[0079] (1) Category represents the category name of the specific attribute tag class to which the tag belongs. All specific attribute tags in the specific attribute tag class have the same category.

[0080] (2) range represents the range of values ​​that the label can take;

[0081] (3) The index represents the index of the tag in its tag class;

[0082] (4) weight represents the weight of the tag. The weight represents the level of detail of the tag. The smaller the range of values, the greater the weight of the tag. For example, "computer" is more detailed than "electronic products" and represents a smaller range than "electronic products". Therefore, "computer" has a greater weight than "electronic products".

[0083] (5) parent represents the index of the parent tag of the tag. If the value range of a tag A covers the value range of another tag B, then the tag A can be called the parent tag of tag B. The parent tag can be regarded as the commonality extraction of the child tags on the one hand, and on the other hand, it can provide a description method when the child tags cannot describe the information. For example, the beverage type of "stay-up-night water" cannot be described by tags such as carbonated beverage, tea beverage, protein beverage, coffee beverage, etc., so it can be described by their common parent tag - beverage.

[0084] Example 3: Hierarchy of information-specific attribute tag class st tag

[0085] The parent element in the specific attribute tag stag divides all specific attribute tags stag in the tag class st into multiple levels, such as Figure 3 As shown.

[0086] Example 4: Composition of the Information Audience Attribute Tag Set TT

[0087] The composition of the information audience attribute tag set TT is as follows: Figure 4 As shown, the information audience attribute tag set TT = {tt1, tt2, ..., tt} k The audience attribute tag set TT of this system consists of a set of audience attribute tag tree classes tt, where each tag class tt represents a type of audience attribute tag, tt = {ttag1, ttag2, ..., ttag...} p}, where any audience attribute tag ttag =<category,range,index> The details are as follows:

[0088] (1) Category represents the category name of the audience attribute tag class to which the tag belongs. All audience attribute tags in the audience attribute tag class have the same category.

[0089] (2) range represents the range of values ​​that the label can take;

[0090] (3) index represents the index of the tag in its tag class.

[0091] Example 5: The process of establishing a mapping between information ID, specific information attributes, and information audience attribute tags.

[0092] The process of establishing a mapping between the information ID and specific information attributes and information audience attribute tags is as follows: Figure 5 As shown, it consists of the following four steps:

[0093] (1) Use the unique identifier id of the information I selected by the information owner as the content of id in the information label mapping list L;

[0094] (2) The information owner’s selection of a specific attribute tag corresponds to a mapping tuple sr in the specific attribute tag mapping SR. For any specific attribute tag selection process of the information owner: the category name of the tag class to which the specific attribute tag stag selected by the information owner belongs is used as the content of the category in the mapping tuple sr in the specific attribute tag mapping SR, and the tag number index of the specific attribute tag stag selected by the information owner is used as the content of the index in the mapping tuple sr in the specific attribute tag mapping SR.

[0095] (3) Each selection of an audience attribute tag by the information owner corresponds to a mapping tuple tr in a specific attribute tag mapping TR. For any selection of an audience tag by the information owner: the category name of the tag class to which the selected audience attribute tag ttag belongs is used as the content of the category in the mapping tuple tr in the audience attribute tag mapping TR; the tag number index of the selected audience attribute tag ttag is used as the content of the index in the mapping tuple tr in the audience attribute tag mapping TR; and the information owner's editing of the importance element of the mapping tuple tr is recorded.

[0096] (4) After the information owner determines all the specific attribute tags and audience attribute tags that need to be added, the importance of each tr in the audience attribute tag mapping TR is converted into an importance score, generating a complete information tag mapping list L. G .

[0097] Example 6: Relationship between Information Tagging System and Demand Tagging System

[0098] The information tagging system and the demander tagging system mentioned above are as follows: Figure 6 As shown, the demander tagging system and the information tagging system have the same content. Specifically, the demander dynamic tag set and the demander static tag set have the same content as the information specific attribute tag set and the information audience attribute tag set, respectively.

[0099] Example 7: The process of establishing a mapping between demander ID and demander static tags

[0100] The process of establishing the mapping between the demander ID and the demander static tag is as follows: Figure 7 As shown, it consists of the following three steps:

[0101] (1) Extract the IDs of users who have established a mapping relationship with static tags;

[0102] (2) The user needs to select at least one static tag jtag from each tag class jt in the user's static tag set JT. However, the user can select multiple static tags from the static tag classes that describe the user's preferences. Each selection of a static tag by the user corresponds to a mapping tuple jr in the static tag mapping JR. The process for any static tag selection by the user is as follows: extract the category name of the tag class containing the static tag jtag selected by the user as the content of the category in the mapping tuple jr in the static tag mapping JR, and extract the tag index of the static tag jtag selected by the user as the content of the index in the mapping tuple jr in the static tag mapping JR.

[0103] (3) After the user confirms all the static tags that need to be added, a complete static tag mapping list JL is generated. G ;

[0104] Example 8: The process of solving for the specific attribute tag (stag) that best meets the potential needs of the user.

[0105] The process of solving for the specific attribute tag (stag) that best meets the potential needs of the user is as follows: Figure 8 As shown, it consists of the following three steps:

[0106] ① All behavior classes B containing the specific attribute tag stag in the tag class st. M =<stagGroup,type,time,num> Filtered out, for behavior type B M Filter the specific attribute tags (stags) within the specific attribute tag group (stagGroup) and retain only the specific attribute tags (stags) belonging to the current tag class (st). The result is B. MS =<{stag1, stag2,…}, type, time, num>;

[0107] ② Starting from the bottom layer of the tag class `st`, calculate the demand level of each specific attribute tag. Quantify the demand level using the demand score `BW`. The calculation method for demand score `BW` is as follows: Where m represents the B applied to the specific attribute tag stag. MSThere are m, where n represents the number of sub-tags of the specific attribute tag stag that are calculated to meet the demand for entering this layer, and W represents the weight of the specific attribute tag stag.

[0108] ③ After calculating the demand weight (BW) for each specific attribute tag (stag) in this layer, for two specific attribute tags with similar demand weights (BW), it can be assumed that the demanders have a common demand for them. Therefore, it is necessary to proceed to the next layer for further calculation. Obtain the non-zero behavior weight values ​​BW1, BW2, ..., BW in this layer. r The maximum value BW in max Calculate the difference between each value and the maximum value. Set a threshold V. If the difference is less than V, proceed to calculate the demand level BW of the next higher level. Repeat the above calculation process until all non-zero demand levels BW1, BW2, ..., BW of a certain level are calculated. x If there is one and only one value less than V, then the specific attribute tag stag' corresponding to the demand degree is obtained;

[0109] Example 9: The process of establishing a mapping between demander ID and demander dynamic tag group

[0110] The process of establishing a mapping between the demander ID and the demander dynamic tag group is as follows: Figure 9 As shown, it consists of the following three steps:

[0111] (1) Extract the IDs of users who have established a mapping relationship with dynamic tag groups;

[0112] (2) The dynamic tag groups dtag1, dtag2, ..., dtag o Each dynamic tag dtag is processed sequentially. Each processing corresponds to a mapping tuple dr in the dynamic tag mapping DR. For any processing process: extract the category name of the tag class to which the dynamic tag dtag belongs in this processing as the content of the category in the mapping tuple dr in the dynamic tag mapping DR, and extract the tag number index of the dynamic tag dtag as the content of the index in the mapping tuple dr in the dynamic tag mapping DR.

[0113] (3) After all dynamic tags in the dynamic tag group have been processed, a complete dynamic tag mapping list DL is generated. G ;

[0114] Example 10: Dynamic Similarity DS Calculation Process

[0115] The calculation process of the dynamic similarity DS is as follows: Figure 10 As shown, it consists of the following three steps:

[0116] (1) Obtain the mapping list L between information and information tags.G ={id,SR G TR G}, where SR G ={sr G 1, sr G 2, ..., sr G n} represents the mapping between information and specific attribute tags, SR G There are n mapping tuples sr G That is, this information is mapped to n specific attribute labels, and the mapping tuple sr G =<category,index> ; Retrieve the generated mapping list of users and dynamic tag groups (DL) G ={id,DR G}, where DR G ={dr G 1, dr G 2, ..., dr G o} represents the mapping between demanders and dynamic tags, DR G There are o mapping tuples dr G That is, a mapping was established between this and o dynamic tags, and the mapping tuple dr G =<category,index> ;

[0117] (2) A specific attribute tag is mapped to a piece of information. The frequency of its occurrence in the dynamic tags of the dynamic tag group is represented by fr. The information is then mapped to the specific attribute tag SR in turn. G Each single-mapped tuple sr G Obtain the mapping between demanders and dynamic tags (DR) G In sequence with DR G Each single-mapped tuple in dr G Comparison. If DR G There is a dr G The content is related to the sr G If the content of the two elements is the same (i.e., the content of the category and index elements is the same), then the frequency fr is incremented by one and added to the next singly mapped tuple sr. G The comparison process continues until the specific attribute tag mapping (SR) is completed. G Each single-mapped tuple sr G The comparison is then used to calculate a dynamic similarity, which characterizes the degree of similarity between the information and the potential needs of the users. Where o represents the mapping between the demander and the dynamic tag DR G The number of mapped tuples;

[0118] (3) Following the steps above, calculate the dynamic similarity DS1, DS2, ..., DS of all information. u Set a threshold σ. If the dynamic similarity DS of a certain piece of information is greater than σ, then retain the ID of that information. Continue this process until the dynamic similarity of each piece of information is processed, and finally obtain a set of IDs for the information.

[0119] Example 11 Static Similarity JS Calculation Process

[0120] The calculation process of the static similarity JS is as follows: Figure 11 As shown, it consists of the following four steps:

[0121] (1) Obtain the mapping list L between information and information tags. G ={id,SR G TR G}, where TR G ={tr G 1, tr G 2, ..., tr G m} represents the mapping between information and audience attribute tags, TR G There are n mapping tuples tr G This means that the information is mapped to m audience attribute tags, and the mapping tuple tr G =<category,index,score> ; Retrieve the generated list of users and static tags JL G ={id, JR G}, where JR G ={jr G 1, jr G 2, ..., jr G h} represents the mapping between users and static tags, JR G There are h mapping tuples jr G That is, a mapping was established with h static tags, and the mapping tuple jr G =<category,index> ;

[0122] (2) The occurrence of the audience attribute tags mapped to a piece of information in the static tags mapped to the demander is described by the static weight value jw. The information is then mapped to the audience attribute tags TR in sequence. G Each single-mapped tuple tr in G Get the mapping between users and static tags JR G In turn with JR G Each single-mapped tuple jr G Comparison. If JR G There is a JR in it. GThe content and the tr G If the content is the same, meaning the content of the element category and the element index are the same, then the static weight value SW is added to the audience attribute tag single mapping tuple tr. G The value of the element's score is then passed to the next audience attribute tag single-map tuple tr. G The comparison process continues until the audience attribute tag mapping (JR) is complete. G Each single-mapped tuple jr G The comparison is then performed, and finally, a static similarity score is calculated to characterize the degree of similarity between the information and the basic personal characteristics of the user. Among them, score sum Map the static label to TR for this information G All mapping tuples in tr G The sum of the scores of the elements;

[0123] (3) Following the steps above, calculate the static similarity JS1, JS2, ..., JS of all the obtained information. e Set threshold If the static similarity of a certain piece of information is greater than JS The ID of this information is retained until the static similarity of each piece of information is processed, and finally a set of IDs of information suitable for recommendation is obtained.

Claims

1. A tag-based information recommendation method for social platforms, characterized in that, Includes the following steps: The first step is to establish an information tagging system to describe all information, including a set of information-specific attribute tags that provide candidate tags for specific attributes describing the information and a set of information-audience attribute tags that provide candidate tags for audience attributes describing the information; obtain the URL of the information owner's information and assign it an ID; An information tag mapping list is generated based on the information tag system, and the mapping between information ID and specific information attribute tags and information audience attribute tags is established through this list; The second step is to establish a demander tagging system based on the content of the information tagging system. This includes a dynamic demander tagging set that provides candidate tags to describe the potential needs of demanders and a static demander tagging set that provides candidate tags to describe the basic personal characteristics of demanders. The demander tagging system has the same content as the information tagging system. Specifically, the dynamic demander tagging set and the static demander tagging set have the same content as the information-specific attribute tagging set and the information audience attribute tagging set, respectively. A static tag mapping list is generated based on the demander tag system, and a mapping between demander ID and demander static tags is established through this list; Acquire data on the interaction behavior between users and information, describe and analyze the interaction behavior of users, generate dynamic tag groups that best meet the potential needs of users, establish a mapping relationship between user IDs and dynamic tag groups, and update in real time; The third step is to match the dynamic tag group of the demander with the specific attribute tag of the information to obtain a set of information IDs. Then, match the audience attribute tag corresponding to the obtained information ID with the static tag of the demander to obtain a set of information IDs suitable for recommendation. Finally, add the URL corresponding to the information ID to the information recommendation queue for information recommendation. The information-specific attribute tag set and the information audience attribute tag set mentioned in the first step are described in the following two ways: The set of information-specific attribute labels ST = {st1, st2, ..., st...} i }, where i represents the number of elements in the set. The specific attribute tag set ST of this system consists of a set of specific attribute tag classes. Each tag class st represents a specific attribute tag, st={stag1, stag2, ..., stag...} j }, j represents the number of elements in the set, where any specific attribute tag stag=<category,range,index,weight,parent> The details are as follows: (1) Category represents the category name of the specific attribute tag class to which the tag belongs. All specific attribute tags in the specific attribute tag class have the same category. (2) range represents the range of values ​​that the label can take; (3) The index represents the index of the tag within its tag class; (4) The weight represents the weight of the tag, and the weight represents the level of detail of the tag. The tag weight is calculated in the following way: Where i represents the level of the specific attribute tag within its class of specific attribute tags, and k represents the number of parent tags that have child tags in level i-1. The number of child tags contained in the j-th parent tag among the k parent tags obtained, where c is a constant; (5) parent represents the index of the parent tag of this tag; The information audience attribute tag set TT={tt1, tt2, ..., tt} k }, where k represents the number of elements in the set. The audience attribute tag set TT of this system consists of a set of audience attribute tag tree classes tt, where each tag class tt represents a type of audience attribute tag, tt={ttag1, ttag2, ..., ttag...} p }, where p represents the number of elements in the set, and any audience attribute tag ttag=<category,range,index> The details are as follows: (1) Category represents the category name of the audience attribute tag class to which the tag belongs. All audience attribute tags in the audience attribute tag class have the same category. (2) range represents the range of values ​​that the label can take; (3) The index represents the index of the tag within its tag class; The interaction behavior of the user described in the second step is defined according to the following model: B =<stagGroup,type,time,num> Where stagGroup is the specific attribute tag group that the information corresponding to this behavior needs to carry, which has been mapped to the information tag. stagGroup={stag1, stag2, ..., stag...} y }, where type represents the type parameter of the behavior, and time represents the time parameter of the behavior. The time parameter time is calculated as follows: ,in, The difference between the time when the behavior occurred and the current time is represented in hours, c is a constant, and num represents the number of this type of behavior, with a value of 1; The second step, generating dynamic tagging components that best meet the potential needs of users, consists of the following three steps: (1) The behavior described by model B is preprocessed according to the following rules: (I) The specific attribute tag group stagGroup and stagGroup of the information corresponding to behavior B that has been mapped with the information tag are a group of specific attribute tags stag. Behaviors with the same behavior type parameter type, behavior time parameter time, etc. are regarded as the same type of behavior; (II) A group of behaviors B1, B2, ..., B that belong to the same type of behavior are classified as B1, B2, ..., B2, B3, B4, B5, B6, B7, B8, B9, B1, B2, ..., B9, B1, B2 ...2, B9, B1, B2, B2, B9, B1, B2, B2, B9, B1, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B2, B z Merge, where z represents the number of elements in the set, to obtain behavior class B. M =<stagGroup,type,time,num> The stagGroup, type, and time are the common parts of this type of behavior, and num becomes the number of all specific behaviors contained in this behavior class; (2) After preprocessing, a set of behavior classes B can be obtained. M 1. B M 2、…、B M p The following three steps can be taken: (I) Extract all the tag classes st1, st2, ..., st involved in these behavior classes. o Then, each tag class st is analyzed and calculated separately in turn; (II) In a specific attribute tag class st, to obtain the specific attribute tag stag that best describes the potential needs of the user in the tag tree, the following three steps are required: ① All behavior classes B containing the specific attribute tag stag in the tag class st. M =<stagGroup,type,time,num> Filtered out, for behavior type B M Filter the specific attribute tags (stags) within the specific attribute tag group (stagGroup) and retain only the specific attribute tags (stags) belonging to the current tag class (st). The result is B. MS = <{stag1, stag2,…}, type, time, num>; ② Starting from the bottom layer of the tag class `st`, calculate the demand level of each specific attribute tag; quantify the demand level of the user using the demand score `BW`. The calculation method for demand score `BW` is as follows: Where m represents the B applied to the specific attribute tag stag. MS There are m, where n represents the number of sub-tags of the specific attribute tag stag that are calculated to meet the demand for entering this layer, and W represents the weight of the specific attribute tag stag. ③ After calculating the demand weight (BW) for each specific attribute tag (stag) in this layer, for two specific attribute tags in this layer whose demand weights (BW) are relatively close, it can be assumed that the demanders have a common demand for them. Therefore, it is necessary to proceed to the next layer for further calculation; obtain the non-zero behavior weight values ​​(BW1, BW2, ..., BW) in this layer. r The maximum value BW in max Let r represent the number of elements in the set. Calculate the difference between each value and the maximum value. Set a threshold V. If the difference is less than V, proceed to calculate the next higher level of demand BW. Repeat the above calculation process until all non-zero demand values ​​BW1, BW2, ..., BW are calculated at a certain level. x If there is one and only one value less than V, we obtain a specific attribute tag stag' corresponding to the demand degree, where x represents the demand degree quantity; (III) For tag classes st1, st2, ..., st o After calculating each tag individually, a set of specific attribute tags stag'1, stag'2, ..., stag' is obtained. o 'o' indicates the number of tags; (3) The resulting set of specific attribute tags stag'1, stag'2, ..., stag' o Dynamic tags with the same content corresponding to each specific attribute tag are added to the user's dynamic tag group, resulting in the user's dynamic tag group dtag1, dtag2, ..., dtag. o .

2. The tag-based information recommendation method for social platforms according to claim 1, characterized in that, The information tag mapping list mentioned in the first step can be defined according to the following model: L={id, SR, TR}, where id corresponds to the unique identifier id of information I, SR is the mapping between information and specific attribute tags, and TR is the mapping between information and audience attribute tags, as detailed below: SR = {sr1, sr2, ..., sr} n }, where n represents the number of elements in the set, and the mapping SR between information and specific attribute labels consists of a set of single-mapping tuples sr, where sr =<category,index> The category corresponds to the category name in the information-specific attribute tag stag, and the index corresponds to the tag number in the specific attribute tag stag; TR = {tr1, tr2, ..., tr m }, where m represents the number of elements in the set, and the mapping TR between information and audience attribute tags consists of a set of single-attribute tuples tr, where tr =<category,index,importance> The category corresponds to the category name in the information audience attribute tag ttag, the index corresponds to the tag number in the audience attribute tag ttag, and the importance represents the importance of the tag to the information that has established a mapping relationship with it. The value range is 1 to m, and the importance in any mapping tr of the information and the audience attribute tag is different. The smaller the value of importance, the higher the importance of the tag.

3. The tag-based information recommendation method for social platforms according to claim 1, characterized in that, The process of establishing the mapping between information IDs and specific information attribute tags and information audience attribute tags, as described in the first step, consists of the following four steps: (1) Use the unique identifier id of the information I selected by the information owner as the content of id in the information tag mapping list L; (2) The information owner’s selection of a specific attribute tag corresponds to a mapping tuple sr in the specific attribute tag mapping SR. For any specific attribute tag selection process of the information owner: the category name of the tag class to which the specific attribute tag stag selected by the information owner belongs is used as the content of the category in the mapping tuple sr in the specific attribute tag mapping SR, and the tag number index of the specific attribute tag stag selected by the information owner is used as the content of the index in the mapping tuple sr in the specific attribute tag mapping SR. (3) Each selection of an audience attribute tag by the information owner corresponds to a mapping tuple tr in a specific attribute tag mapping TR. For any selection of an audience tag by the information owner: the category name of the tag class to which the audience attribute tag ttag selected by the information owner belongs is used as the content of the category in the mapping tuple tr in the audience attribute tag mapping TR; the tag number index of the audience attribute tag ttag selected by the information owner is used as the content of the index in the mapping tuple tr in the audience attribute tag mapping TR; and the information owner's editing of the importance element of the mapping tuple tr is recorded. (4) After the information owner determines all the specific attribute tags and audience attribute tags that need to be added, the importance of each tr in the audience attribute tag mapping TR is converted into an importance score. The importance score, abbreviated as S, is calculated as follows: ,in, The number of tuples in the audience attributes representing the information, where c is a constant, ultimately generates a complete list of information tag mappings L. G .

4. The tag-based information recommendation method for social platforms according to claim 1, characterized in that, The process of establishing the mapping between demander IDs and static tags in the second step involves the following dynamic tag mapping list model: JL={id, JR}, where id is the unique identifier of the demander filling in the list, and JR is the mapping between the demander and the static tag, as detailed below: JR = {jr1, jr2, ..., jr} h }, where h represents the number of elements in the set, and the mapping JR between demanders and static tags consists of a set of single-mapping tuples jr, where jr =<category,index> The category corresponds to the category name of the static tag class jt to which the static tag belongs, and the index corresponds to the tag number index in the static tag jtag. The process of establishing a mapping between requester IDs and requester static tags can be divided into the following three steps: (1) Extract the IDs of users who have established a mapping relationship with static tags; (2) The user needs to select at least one static tag jtag in each tag class jt in the user's static tag set JT. However, the user can select multiple static tags in the static tag class describing the user's preferences. The user's selection of a static tag corresponds to a mapping tuple jr in the static tag mapping JR. For any static tag selection by the user, the category name of the tag class where the user selects the static tag jtag is located is extracted as the content of the category in the mapping tuple jr in the static tag mapping JR. The tag number index of the user selects the static tag jtag is extracted as the content of the index in the mapping tuple jr in the static tag mapping JR. (3) After the user determines all the static tags that need to be added, generate a complete static tag mapping list JL. G .

5. The tag-based information recommendation method for social platforms according to claim 1, characterized in that, The process of establishing the mapping relationship between the requester's ID and the dynamic tag group described in the second step involves the following model of the dynamic tag mapping list: DL={id, DR}, where id is the unique identifier of the requester, and DR is the mapping between the requester and the dynamic tag, as detailed below: DR = {dr1, dr2, ..., dr} o The mapping DR between demanders and dynamic tags consists of a set of single mapping tuples dr, where dr =<category,index> The category corresponds to the category name of the dynamic tag class dt to which the dynamic tag dtag belongs, and the index corresponds to the tag number index in the dynamic tag dtag. The process of establishing the mapping relationship between the requester ID and the dynamic tag group consists of the following three steps: (1) Extract the IDs of users who have established a mapping relationship with the dynamic tag group; (2) Create dynamic tag groups dtag1, dtag2, ..., dtag o Each dynamic tag dtag is processed sequentially. Each processing corresponds to a mapping tuple dr in the dynamic tag mapping DR. For any processing process: extract the category name of the tag class to which the dynamic tag dtag belongs in this processing as the content of the category in the mapping tuple dr in the dynamic tag mapping DR, and extract the tag number index of the dynamic tag dtag as the content of the index in the mapping tuple dr in the dynamic tag mapping DR. (3) After all dynamic tags in the dynamic tag group have been processed, a complete dynamic tag mapping list DL is generated. G .

6. The tag-based information recommendation method for social platforms according to claim 1, characterized in that, The process of matching the dynamic tag set of the demander with the specific attribute tags of the information, as described in the third step, is divided into the following three steps: (1) Obtain the mapping list L of information and information tags G ={id,SR G TR G }, where SR G ={sr G 1, sr G 2, ..., sr G n } is a mapping between information and specific attribute tags, SR G There are n mapping tuples sr G That is, this information is mapped to n specific attribute labels, and the mapping tuple sr G =<category,index> ; Retrieve the generated mapping list of users and dynamic tag groups (DL) G ={id,DR G }, where DR G ={dr G 1, dr G 2, ..., dr G o } represents the mapping between demanders and dynamic tags, DR G There are o mapping tuples dr G That is, a mapping is established between the demander and o dynamic tags, and the mapping tuple dr G =<category,index> ; (2) A specific attribute tag is mapped to a piece of information. The frequency of its occurrence in the dynamic tags of the dynamic tag group is represented by fr. The information is then mapped to the specific attribute tag SR in turn. G Each single-mapped tuple sr G Obtain the mapping between demanders and dynamic tags (DR) G In sequence with DR G Each single-mapped tuple in dr G Compare; if DR G There is a dr G The content is related to the sr G If the content of the two elements is the same (i.e., the content of the category and index elements is the same), then the frequency fr is incremented by one and added to the next singly mapped tuple sr. G The comparison process continues until the specific attribute tag mapping (SR) is completed. G Each single-mapped tuple sr G The comparison is then used to calculate the dynamic similarity DS, which characterizes the degree of similarity between the information and the potential needs of the users. Where o is the mapping between demanders and dynamic tags (DR) G The number of mapped tuples; (3) Following the steps above, calculate the dynamic similarity DS1, DS2, ..., DS of all information. u A threshold σ is set. If the dynamic similarity DS of a certain piece of information is greater than σ, the ID of that information is retained. This process continues until the dynamic similarity of each piece of information is processed, and finally a set of IDs of information is obtained.

7. The tag-based information recommendation method for social platforms according to claim 1, characterized in that, The process of matching the audience attribute tags corresponding to the obtained information IDs with the static tags of the users, as described in the third step, consists of the following three steps: (1) Obtain the mapping list L of information and information tags G ={id,SR G TR G }, where TR G ={tr G 1, tr G 2, ..., tr G m } represents the mapping between information and audience attribute tags, m represents the number of elements in the set, and TR G There are n mapping tuples tr G This means that the information is mapped to m audience attribute tags, and the mapping tuple tr G =<category,index,score> ; Retrieve the generated list of users and static tags JL G ={id, JR G }, where JR G ={jr G 1, jr G 2, ..., jr G h } represents the mapping between demanders and static tags, h represents the number of elements in the set, JR G There are h mapping tuples jr G That is, a mapping is established between the demander and h static tags, and the mapping tuple jr G =<category,index> ; (2) The occurrence of the audience attribute tags mapped to a piece of information in the static tags mapped to the demander is described by the static weight value jw. The information is then mapped to the audience attribute tags TR in sequence. G Each single-mapped tuple tr in G Get the mapping between users and static tags JR G In turn with JR G Each single-mapped tuple jr G Compare; if JR G There is a JR in it. G The content and the tr G If the content is the same, meaning the content of the element category and the element index are the same, then the static weight value SW is added to the audience attribute tag single mapping tuple tr. G The value of the element's score is then passed to the next audience attribute tag single-map tuple tr. G The comparison process continues until the audience attribute tag mapping is complete. G Each single-mapped tuple jr G The comparison is then performed, and finally, a static similarity JS is calculated to characterize the degree of similarity between the information and the basic personal characteristics of the user. ,in Map the static label to TR for this information G All mapping tuples in tr G The sum of the scores of the elements; (3) Following the steps above, calculate the static similarity JS1, JS2, ..., JS of all the obtained information. e Set a threshold φ. If the static similarity JS of a certain piece of information is greater than φ, then retain the ID of that information. Continue this process until the static similarity of each piece of information has been processed, and finally obtain a set of IDs of information suitable for recommendation.

Images