Information recommendation method and system

Abstract

The present invention relates to an information recommendation method and system, wherein the method comprises: obtaining a first tag of a first user and a second tag of a second user, wherein the second user and the first user are in a supply-demand relationship; the tags include tag type and tag content; filtering multiple second / first users by matching the first / second tag of a target recommended user with the second / first tags of multiple second / first users to obtain multiple second / first users; sorting the multiple second / first users in descending order of matching degree; and generating recommended information based on a preset number of ranked second / first users and pushing it to the target recommended user. The present invention has a wide range of applications and can not only accurately provide recommended information to users, saving users' query time, but also fully utilize the rich information and data of the supply and demand platform, thereby effectively improving platform user stickiness.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet data processing, and in particular to an information recommendation method and system. Background Art

[0002] In today's information age, various professional and comprehensive information platforms are able to provide the information needed by both supply and demand users. For example, professional recruitment platforms offer a wide range of job postings and job search information to both recruiters (on the supply side) and job seekers (on the demand side). Some comprehensive websites provide a wealth of advertising bidding information and advertising demand information. Taking the recruitment market as an example, the vast majority of job seekers and recruiters choose to find suitable positions on online recruitment platforms. Typically, job seekers and recruiters register on recruitment websites or recruitment apps. Job seekers fill out a resume, which includes personal information and the position they are seeking, while recruiters fill out a job posting, which includes company information, the specific position being sought, and the job requirements. Because recruitment platforms aggregate a vast amount of information, relying solely on manual search by job seekers and recruiters to find suitable positions or individuals within the vast amount of information would be both time-consuming and extremely difficult. Therefore, to increase the success rate of job applications and recruitment on recruitment platforms and improve efficiency for both job seekers and recruiters, some recruitment platforms have launched job recommendation services, which use algorithms to recommend job information to job seekers. For example, Chinese invention patent application number 201811208036.3, titled "A Method and System for Job Recommendation," provides a method that uses a deep learning algorithm to predict a data matrix generated from user access data, and then generates job recommendation data based on the prediction results and persona data. Chinese invention patent application number 201710947915.7, titled "Job Recommendation Processing Method and Apparatus," provides another method that extracts features from a job seeker's resume to obtain job seeker feature information, extracts features from resumes submitted to a recruitment project to obtain job feature information, and then matches these two features to obtain relevant job recommendations based on the degree of match. There are other methods for job recommendation, which will not be discussed here. Analysis of existing job recommendation methods reveals that the recommended positions are often suboptimal, meaning they fail to meet the job seeker's requirements in one or more aspects. This can occur for a variety of reasons, including the job seeker's omission of certain information in their resume. For example, job seekers typically don't directly list salary requirements on their resumes, despite the fact that this is a highly valued metric for job seekers. Because job seekers don't include this information in their resumes, the job postings aren't factored in when calculating job offers, resulting in recommended positions with salaries that don't meet their expectations. Another reason job seekers are dissatisfied with recommended positions is that the recommendation algorithm can't accurately understand the applicant's resume, resulting in inaccurate or even irrelevant job recommendations.Furthermore, most existing job recommendation methods are unable to detect inaccurate recommendations in a timely manner, nor can they make timely revisions, resulting in inaccurate job information being sent to job seekers. As a result, the job information pushed becomes worthless and annoying spam for job seekers, seriously affecting their enthusiasm for using the recruitment platform. Summary of the Invention

[0003] In response to the technical problems existing in the prior art, the present invention proposes an information recommendation method and system, which can accurately understand the needs of target recommendation users and provide target recommendation users with required information with high matching degree.

[0004] In order to solve the above technical problem, according to one aspect of the present invention, the present invention provides an information recommendation method, which includes the following steps: obtaining a first tag of a first user and a second tag of a second user, wherein the second user and the first user are in a supply-demand relationship; the tag includes a tag type and a tag content; filtering multiple second / first users by matching the first / second tag of a target recommended user with the second / first tags of multiple second / first users to obtain multiple second / first users; sorting the multiple second / first users in order from high to low according to the degree of matching; and generating recommendation information based on a preset number of second / first users ranked first and pushing it to the target recommended user.

[0005] According to another aspect of the present invention, the present invention provides an information recommendation system, which includes a label acquisition module, a filtering module, a sorting module and a push module, wherein the label acquisition module is configured to obtain a first label of a first user and a second label of a second user based on the demand information of the first / second user, wherein the second user and the first user are in a supply and demand relationship; the label includes a label type and a label content; the filtering module is configured to filter multiple second / first users by matching the first / second label of a target recommended user with the second / first labels of multiple second / first users to obtain multiple second / first users; the sorting module is configured to sort the multiple second / first users obtained after filtering in order of matching degree from high to low; the push module is configured to generate recommendation information based on a preset number of second / first users ranked first and push it to the target recommended user.

[0006] The information recommendation method and system provided by the present invention can accurately understand the needs of target recommendation users, and can not only accurately provide recommended information to users, saving users' query time, but also make full use of the rich information and data of platforms that provide supply and demand information, such as recruitment platforms, effectively improving user stickiness. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Below, the preferred embodiments of the present invention will be further described in detail with reference to the accompanying drawings, in which:

[0008] Figure 1 is a principle block diagram of a job information recommendation system according to an embodiment of the present invention;

[0009] Figure 2 is a flowchart of a recommended method according to one embodiment of the present invention;

[0010] Figure 3 is a principle block diagram of a label acquisition module according to an embodiment of the present invention;

[0011] Figure 4A is a flow chart of a label generation method according to one embodiment of the present invention;

[0012] Figure 4B is a flowchart of obtaining keywords and corresponding semantic tags according to an embodiment of the present invention;

[0013] Figure 4C is a schematic diagram of a syntax tree according to an embodiment of the present invention;

[0014] Figure 4D is a schematic diagram of a directed acyclic graph according to an embodiment of the present invention;

[0015] Figure 5 is a principle block diagram of a sorting module according to one embodiment of the present invention;

[0016] Figure 6 is a principle block diagram of a weight calculation module according to an embodiment of the present invention; and

[0017] Figure 7 It is a principle block diagram of a push module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0019] In the detailed description that follows, reference may be made to the various drawings that form part of this application and illustrate specific embodiments of the present application. In the drawings, similar reference numerals describe substantially similar components in different figures. Each specific embodiment of the present application is described below in sufficient detail to enable a person of ordinary skill in the art to implement the technical solutions of the present application. It should be understood that other embodiments may be utilized or that structural, logical, or electrical changes may be made to the embodiments of the present application.

[0020] The present invention provides an information recommendation method for recommending demanders to suppliers, or vice versa, based on the demand information obtained from each of the two parties. For example, in the recruitment market, the suppliers are job suppliers, i.e., recruiters; the demanders are job seekers, i.e., job seekers. In the advertising and investment market, those who provide advertising channels are suppliers, and those who publish advertisements are demanders. In the financial capital market, those who provide funding are suppliers, and those seeking financing are demanders, and so on. The methods and systems provided by the present invention are applicable to various fields with the above-mentioned relationships. The following uses the recruitment market as an example to describe the methods and systems provided by the present invention in detail. To facilitate the technical solution, in the following embodiments, the first user is a job seeker, the second user is a recruiter, and the target recommended user is the first user, i.e., recommending recruiters to job seekers. Of course, the target recommended user can also be a recruiter, i.e., recommending job seekers to recruiters. The methods of the present invention are applicable to recruitment platforms, such as recruitment websites and recruitment apps. Typically, job seekers upload their resumes to recruitment platforms, or fill out their resumes according to the platform's format requirements and perform searches, viewing, and other operations on the recruitment platforms. Similarly, a recruiter sends recruitment information to a recruitment platform, or fills out recruitment information according to the platform's format requirements and searches and views job search information. The platform obtains the job seeker's requirements based on the resume uploaded or filled out by the job seeker; and obtains the recruiter's requirements based on the recruitment information uploaded or filled out by the recruiter. The present invention recommends job seekers to recruiters, or recommends recruiters to job seekers, based on the respective requirements of the recruiter and job seeker, as well as their search and viewing behavior data on the recruitment platform. This recommendation is highly accurate and well-matched, saving time for job seekers and / or recruiters, and increasing both the job seeker's job search success rate and the recruiter's recruitment success rate.

[0021] In order to clearly illustrate the solutions of the present invention, the present invention defines the specific meanings of the following terms:

[0022] Keywords: Phrases with diverse expressions, such as phrases consisting of two nouns, such as "car sales", "vehicle sales", "car promotion" and "vehicle promotion".

[0023] Segmentation unit: During the semantic extraction process, the segmented words obtained after the sentence is segmented, which are phrases composed of single words, double words, or multiple words, such as "I", "I", "Java", and "engineer" in the sentence "I am a Java engineer" after segmentation.

[0024] Word unit: A word collected in a dictionary that has independent semantics and cannot be further divided semantically, such as "car" and "sales".

[0025] Semantic tag: A standardized version of a keyword, such as using the semantic tag "car sales" as a standardized version of the keywords "car sales", "vehicle sales", "car promotion", and "vehicle promotion".

[0026] Tag: A phrase that expresses a specific content, including the tag type and tag content.

[0027] Prefix word: the first of two or more word units that make up a keyword, such as "car" in "car sales", "vehicle" in "vehicle sales", etc.

[0028] Suffix: The last word in two or more word units that make up a keyword, such as "sales" in "vehicle sales", "purchase" in "car purchase", etc.

[0029] Figure 1 is a principle block diagram of a job information recommendation system according to an embodiment of the present invention; Figure 2 In the present invention, the target recommendation user is a job seeker, and the job information recommendation system includes a tag acquisition module 1, a filtering module 2, a sorting module 3 and a pushing module 4.

[0030] Step S1, obtain the first label of the job seeker and the second label of the recruiter. Figure 1 The recruitment platform also includes a data collection module 7, which collects various data, including job seekers' resumes and recruiters' recruitment information. The label acquisition module 1 in the system uses the job seekers' resumes and recruiters' recruitment information as the required information of the two, and obtains the job seekers' first label and the recruiters' second label from them. The first and second labels are only used to distinguish and do not represent the quantity. Figure 3 , which is a principle block diagram of the label acquisition module 1, the label acquisition module 1 includes a semantic recognition unit 11, a first matching unit 12, a label generation unit 13, a label definition unit 14 and a label calculation module 15. Figure 4A The flowchart of label generation shown is described as follows:

[0031] Step S11, extract multiple keywords from the target recommendation user's demand information and obtain corresponding semantic tags. Specifically, the semantic recognition unit 11 reads all the text content in the job seeker's resume from the data collection module 7, performs semantic recognition on it, and thus obtains multiple keywords, such as "Java engineer", "software development", "proficient in Java", "C++ development", etc., or "Chinese teacher", "teaching Chinese in elementary school", "Chinese tutoring", "part-time composition professor" or "clothing sales", "women's clothing shopping guide", etc. An implementation example of obtaining the keywords and obtaining the corresponding semantic tags is as follows: Figure 4B As shown:

[0032] Step S110, the demand information of the target recommendation user is segmented in units of sentences to obtain multiple segmentation units. The segmentation unit refers to: in the process of semantic extraction, the segmentation unit is obtained after the sentence is segmented, which is a phrase composed of single words, double words or multiple words, such as "I", "I", "Java", and "engineer" in the sentence "I am a Java engineer" after segmentation. In this step, any existing segmentation method can be used for segmentation. For example, a mechanical segmentation method based on a dictionary, such as a forward maximum matching method, a reverse maximum matching method or a two-way matching method. Or it can be a statistical segmentation method, which determines the probability of a word forming a word by calculating the joint frequency of occurrence of a word and its adjacent words in the context. In one embodiment, a hierarchical hidden Markov model (HHMM) integrates lexical analysis tasks such as Chinese word segmentation, segmentation and disambiguation, unregistered word recognition, and part-of-speech tagging into a relatively unified model, thereby achieving synonym replacement, Chinese word segmentation, and part-of-speech tagging for a string in a sentence, thereby segmenting the sentence.

[0033] Step S120, analyzing whether the multiple word segmentation units constitute keywords. In which, the word unit dictionary is queried, and in response to the word segmentation unit in the sentence being queried in the word unit dictionary, the word segmentation unit is determined to be a word unit. In which, the word unit dictionary includes word units commonly used in the recruitment field. Such as the word units "software" and "hardware" in the industry dimension, the word units "engineer", "sales", and "customer service" in the function dimension, the word units "Java" and "floristry" in the skill dimension, and the word units "Japanese" and "English" in the language dimension, etc.

[0034] When multiple word units are obtained through query, the multiple word units are arranged and combined to obtain multiple phrases. A phrase dictionary is searched, and when the phrase is found in the phrase dictionary, the phrase is determined to be a keyword.

[0035] For example, after word segmentation, the sentence is "I am a Java and C++ engineer." By searching the word unit dictionary, we find the word units "Java," "C++," and "engineer" in the sentence, thus confirming that the word units of the sentence are "Java," "C++," and "engineer." By permuting and combining these three word units, we can get phrases such as "Java," "C++," "engineer," "Java engineer," "C++ engineer," "JavaC++," and "JavaC++ engineer." Then, by searching the phrase dictionary, we find "Java," "C++," "engineer," "Java engineer," and "C++ engineer," thus confirming that the keywords are "Java," "C++," "engineer," "Java engineer," and "C++ engineer." However, the phrases "JavaC++" and "JavaC++ engineer" do not appear in the phrase dictionary, so these two phrases are excluded.

[0036] After the above processing, it can be determined that the sentence "I am a Java and C++ engineer" contains keywords, so the next step of grammatical analysis can be carried out. If there are no keywords in the sentence, the next step of analysis will not be carried out.

[0037] Step S130 : In response to the multiple word segmentation units forming one or more keywords, extracting sentences containing the keywords.

[0038] Step S140, grammatically analyze the sentence containing the keyword to obtain a grammatical tree. In one embodiment, a grammatical tree is generated for each sentence, and a grammatical tree is provided with a root node Root, and each segmentation unit in the sentence is a node, wherein the root node Root points to the segmentation unit as the predicate (whose part of speech is usually a verb), and then the segmentation unit as the predicate points to other segmentation units, and two nodes with a pointing relationship constitute a certain grammatical relationship. For example, the segmentation unit with the part of speech of a verb points to the segmentation unit as the subject, and these two segmentation units constitute a subject-predicate relationship (nsubj); the segmentation unit with the part of speech of a verb points to the segmentation unit as the object of the verb, and these two segmentation units constitute a verb-object relationship (dobj). In this step, first, according to the order of the segmentation units in a sentence, starting from the beginning of the sentence, the pointing relationship and grammatical relationship of the two segmentation units are obtained in sequence according to the preset grammatical rules; then a grammatical tree is established based on the pointing relationship and grammatical relationship of the segmentation units. In one embodiment, a neural network grammatical relationship analysis model is used to calculate the pointing relationship and grammatical relationship of the two segmentation units in each sentence. In this network model, the transfer analysis method is used to obtain the pointing relationship and grammatical relationship between two word units. For example, the sentence "I am familiar with Java and C++ development." generates a grammatical tree as follows Figure 4C shown.

[0039] Step S150, extracting effective keywords and corresponding semantic labels from the syntax tree. It can be seen from the syntax tree that there are two word segmentation units with multiple grammatical relationships in a syntax tree. However, when understanding the intention of the target recommendation user demand information, some word segmentation units and their grammatical relationships are meaningless and may even interfere with the understanding of the intention of the target file. Therefore, the word segmentation unit pairs in unnecessary grammatical relationships are removed from the syntax tree. The remaining word segmentation units are reconnected according to the mutual connection relationship to obtain a directed acyclic graph, such as Figure 4D As shown, it is based on Figure 4C The directed acyclic graph of the grammatical tree shown in FIG. 1 is a directed acyclic graph, wherein all the segmentation units connected step by step on each path from the start node to the end node in the directed acyclic graph are determined as a language group. Figure 4D The directed acyclic graph shown has three language groups, namely "familiar" + "development" + "C++", "familiar" + "development" + "Java", and "familiar" + "I". Query whether all the word units that constitute the keyword are in the same language group. When all the word units that constitute the keyword are in the same language group, it is a valid keyword, otherwise it is not a valid keyword. For example, the prefix word "development" and the suffix word "Java" in the keyword "development Java" extracted from "I am familiar with Java and C++ development." are in the same language group, and the prefix word "development" and the suffix word "C++" in the keyword "development C++" are in the same language group, thus determining that "development Java" and "development C++" are valid keywords.

[0040] Then the valid keywords are standardized to obtain corresponding semantic tags. The present invention provides a configuration file, which includes a prefix table and a suffix table, in which a plurality of prefixes / suffixes constituting keywords are recorded, and each prefix / suffix has a corresponding standardized version constituting a semantic tag. For example, the prefix "socket" is a standardized version of the prefix "socket". The standardized suffix of the suffix "selling" is "sales". According to the prefix and suffix of the valid keywords, the prefix table and the suffix table are queried respectively, and the prefix and suffix of the valid keywords are mapped to standard prefixes and standard suffixes. For example, the keyword "real estate promotion" is mapped to "real estate sales", "Java development" is mapped to "Java R&D engineer", "business specialist" is mapped to "business personnel", and so on.

[0041] Step S12, using the knowledge graph library to match the corresponding knowledge node for each semantic tag. In this step, the first matching unit 12 uses the knowledge graph library to determine multiple nodes for each semantic tag. Among them, the knowledge graph library includes multiple nodes, and each node is associated with one or more other nodes. The node attribute in this embodiment is the type of node content. For example, the attribute association of the node "Java" and the node "software" is an industry inclusion relationship, that is, from the perspective of the industry, "software" includes "Java". The node "Java" has a skill inclusion relationship with the node "Hibernate", the node "J2EE", the node "Struts", and the node "Spring", that is, "Java" contains "J2EE", "Hibernate" and "Spring" respectively.

[0042] After the first matching unit 12 inputs the semantic tags into the knowledge graph, it can obtain one or more corresponding nodes. For example, outputting the keyword "Hibernate development" into the knowledge graph can obtain an "engineer" node for the job attribute; a "Hibernate" node and a "Java" node for the skill attribute; and a "software" node for the industry attribute.

[0043] Step S13, generates one or more first labels based on the one or more nodes obtained by matching. Specifically, the label generation unit 13 reads and matches these nodes, and generates a label for every two nodes. For example, when the function node "engineer", the skill node "Java" and the industry node "software" are obtained by the keyword "Java development", two labels {career direction: "software engineer"} and {career direction: "Java engineer"} can be obtained by combining them in pairs. Since the two labels belong to the same category, a more accurate label {career direction: "Java engineer"} is obtained after merging. For another example, the node combination obtained by matching the semantic labels corresponding to the keywords "software development" and "proficient in C++" is combined to obtain the label {"career direction": "C++ engineer"}.

[0044] Step S14 calculates one or more first tags based on the preset options in the requirement information. In one embodiment, when a job seeker's resume is generated according to the platform device's format, the platform includes certain options in the resume template, such as "Employment Start and End Dates," "Education," "Type of Intended Company," and "Salary Range," to more accurately understand the job seeker's job-seeking intentions or to identify information that recruiters value. When filling out a resume on the platform, the job seeker must fill in or select the appropriate information for each of these options. For example, "Employment Start and End Dates" requires two specific time points, and the job seeker must fill in two specific time points; "Education" includes options such as "High School, Technical Secondary School, Junior College, Bachelor's Degree," and the job seeker must select the appropriate option. Each option corresponds to a tag, and the tag calculation module 15 calculates or determines the tag content based on the options. For example, for the option "Employment Start and End Dates," the tag {"Working Hours": "3.5"} is obtained, and for the option "Education," the tag {"Education Requirement": "Bachelor's Degree"} is obtained.

[0045] Step S15, determine one or more main tags and one or more sub-tags among the multiple first tags. Since a target recommendation user has multiple tags, in order to facilitate use during filtering, the tag definition unit 14 determines one or more main tags among the multiple first tags. For example, the tag that expresses what to do is used as the main tag, that is, the tag of the type "career direction" is used as the main tag, and the tags that express the basic conditions, requirements, supplementary conditions, etc. of doing things are used as sub-tags. For example, tags of the types "education", "salary", "foreign language", etc. are used as sub-tags. Among them, there can be one or more main tags. For example, when the job seeker's demand information includes the willingness for different career directions in the same industry, multiple main tags of the type "career direction" can be obtained according to their demand information.

[0046] Similarly, multiple second labels of multiple recruiters are obtained according to the above steps.

[0047] Of course, when the aforementioned first label and second label are generated by other systems, it is only necessary to obtain them from other systems based on the identity of the target recommendation user.

[0048] Step S2, filtering multiple recruiters. The filtering module 2 uses the first tag of the target recommended user (a specific job seeker) to match the second tags of multiple recruiters, thereby filtering out recruiters that do not match the first tag. Finally, the remaining ones are recruiters that match the first tag of the target recommended user. When filtering, the main tag is first used to filter multiple recruiters to determine multiple recruiters that match the job seeker's career direction, and then the various sub-tags are used to filter multiple recruiters again. Each filtering operation may obtain one or more recruiters that meet the requirements. Of course, it is also possible that no recruiter is obtained, and multiple recruiters are finally obtained after the filtering operation.

[0049] Step S3, sorting the multiple recruiters. The sorting module 3 includes a single-dimension sorting unit and a total sorting unit. In one embodiment, Figure 5 As shown, the sorting module 3 includes multiple single-dimensional sorting units 311, 312, and 31n, and each single-dimensional sorting unit 311, 312, and 31n calculates a sorting value of a dimension. In one embodiment, the dimension corresponds to the label type, and each label type is used as a dimension to sort multiple recruits to obtain a sorting value Vi. For example, after filtering, a total of 20 recruits are obtained. Corresponding to the dimension of "salary", the value in the label of type "salary" is used as the base value, and compared with the salaries provided by the 20 recruits respectively. The 20 recruits are sorted from small to large according to the difference value, and each recruiter obtains a sorting value V in the "salary" dimension. m When sorting in the dimension of "Distance from Home (D)", we use the job seeker's address as the coordinate and calculate the distance D between each recruiter and the job seeker's address. Then we sort the 20 recruiters in the order of distance D from small to large. Each recruiter gets a ranking value V in the dimension of "Distance from Home (D)". D The total ranking unit 321 calculates the final ranking V of each recruiter based on the weight of the target recommended user in each dimension and the ranking value of each recruiter in each dimension. Among them, vi is the ranking value of the i-th dimension, and qi is the weight of the target recommended user in the i-th dimension.

[0050] In order to obtain more accurate and more targeted recommended user information, the present invention needs to dynamically update the dimension weights according to the user's behavior, updated demand information, etc. Therefore, in one embodiment of the present invention, the system further includes a weight calculation module 6, such as Figure 6As shown, the weight calculation module 6 includes a preference value calculation unit 61, a weight calculation unit 62 and a weight adjustment unit 63, wherein the preference value calculation unit 61 and the weight adjustment unit 63 are connected to the data collection module 7, and the preference value calculation unit 61 analyzes the demand information and behavior data of the target recommended user, and determines the preference value L of a job seeker in each dimension based on the analysis results. The present invention can use any existing analysis method to determine the preference value of each dimension. For example, a statistical information entropy scheme is used to perform information entropy statistics on the location of the job posted by the user. The more times the user posts to a single area, the lower the information entropy (the higher the preference). On the contrary, if the area where the user posts is messy, the higher the information entropy (the lower the preference); or through a preset simple probability statistical model, plus the user's actual posting data, the user's real preference is fitted; or a deep network neural model can be used. Through continuous training of samples, a high-precision prediction of the preference value of a dimension can be obtained. The weight calculation unit 62 determines the weight of the target recommended user in the dimension based on the percentage of the preference value of each dimension to the overall dimension preference value. Where li is the preference value of the i-th dimension, and m is the number of dimensions. The weight adjustment unit 63 is connected to the data collection module 7 and the preference value calculation unit 61. It detects in real time or periodically whether there are changes in the demand information of the target recommendation user and whether there are changes in the behavior data of the target recommendation user. When the target recommendation user updates the demand information and / or collects new behavior data, it sends an instruction to the preference value calculation unit 61 to recalculate the preference value of each dimension, and then the weight calculation unit 62 recalculates the weight. Therefore, the present invention can update the weight in real time or periodically.

[0051] Step S4, generates recommendation information based on a preset number of recruiters ranked in front and pushes it to the target recommendation user. In this step, the push module 4 generates recommendation information for a preset number (such as 20 or 10, or a number determined by a percentage) of recruiters ranked in front. The recommendation information includes the name of the recruiting company, the name of the recruiting position, the link to the recruitment information page posted by the recruiter, the location of the position in the recruitment information, the salary range of the position, etc. It is then pushed to the job seeker through a pop-up window or email. In a specific embodiment, if Figure 7As shown, the push module 4 includes a recommendation information generation unit 41, a comparison unit 42, and a push unit 43. The recommendation information generation unit 41, according to a configuration, obtains corresponding information of a certain number of top-ranked recruiters to generate recommendation information. For example, information such as the recruiting company name, the job title, the link to the recruiter's job information page, the job location, and the salary of the position may be recorded in a table. To avoid frequently pushing the same information and causing trouble for job seekers, in one embodiment, after obtaining the recommendation information, it is sent to the comparison unit 42. The comparison unit 42 compares the currently generated recommendation information with the historical recommendation information. Only when the currently generated recommendation information differs from the historical recommendation information will the push unit 43 push the currently generated recommendation information to the target recommendation user. In a further embodiment, the push module 4 also includes a push information filtering unit 44. As shown in the figure, the comparison unit 42 compares the currently generated recommendation information with the historical recommendation information to determine which has been pushed and which has not. The push information filtering unit 44 filters out the information that has been recommended in the past from the currently generated recommendation information. In this case, the recommended information only includes recruiters who have not been pushed. For the sake of distinction, the recommended information obtained after filtering is referred to as "pending recommendation information." Push unit 43 then pushes this pending recommendation information to the target recommending user. In some embodiments, when a job seeker logs into the recruitment platform, the system detects the user's login operation and pushes the currently generated recommended information or pending recommendation information to the user in a pop-up window. This window remains on the page for a certain period of time, such as five seconds. Operation buttons, such as Save and Close, are provided on the pop-up window. When the user chooses to save, the information is saved in table format.

[0052] Optionally, the process further includes step S5, monitoring the user's actions on the recommended recruiters, such as whether they clicked to view, when they viewed them, whether they submitted their resume, etc. In step S6, the dimensions used in the final ranking and / or their weights are adjusted based on these actions. By monitoring the user's actions on the recommended information, the present invention can modify the dimensions and their weights used in the ranking calculation, thereby ensuring that the next recommendation is more in line with the user's intent.

[0053] Similarly, when the target recommendation user is a recruiter, the same process is used to recommend job seekers to the recruiter. The same process can be applied to other supply-demand relationships involving first and second users. Therefore, the information recommendation method and system provided by the present invention can not only accurately provide recommended information to users, saving their query time, but also fully utilize the rich information and data of platforms that provide supply-demand information, such as recruitment platforms, thereby effectively improving platform user stickiness.

[0054] The above embodiments are only used to illustrate the present invention, and are not intended to limit the present invention. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the scope of the present invention. Therefore, all equivalent technical solutions should also fall within the scope of the present invention.

Claims

1. A method for recommending job seeker information and recruiter information, comprising: Obtain the job seeker's or recruiter's demand information. The job seeker's demand information is a resume, and the recruiter's demand information is a recruitment information. Extract multiple keywords from the demand information according to semantics and obtain corresponding demand semantic tags. Obtaining a preset recruitment domain knowledge graph library, wherein the attributes of the knowledge nodes in the recruitment domain knowledge graph library include function attributes, skill attributes, and industry attributes, wherein nodes having a skill inclusion relationship or an industry inclusion relationship between the node attributes are connected to each other; Matching the obtained demand semantic label with the knowledge nodes in the recruitment domain knowledge graph library to obtain multiple knowledge nodes associated with the demand semantic label, and combining the multiple knowledge nodes in pairs to generate a combined label. If the demand semantic label comes from the job seeker's resume, the combined label is used as the job seeker's job label; if the demand semantic label comes from the recruiter's recruitment information, the combined label is used as the recruiter's recruitment label; When recommending recruiter information to a target job seeker, multiple recruiters are filtered by matching the target job seeker's job search tag with the recruitment tags of multiple recruiters to obtain multiple recruiters that match the job search tags; the matching values ​​between the recruitment tags of the multiple matched recruiters and the target job seeker's job search tag are calculated, and the recruiters are sorted from high to low according to the matching values. Recommended information is generated based on a preset number of recruiters ranked first and pushed to the target job seeker; When recommending job seeker information to a target recruiter, multiple job seekers are filtered by matching the target recruiter's recruitment tag with the job seeker tags of multiple job seekers to obtain multiple job seekers that match the recruitment tags; the matching values ​​between the job seeker tags of the multiple matched job seekers and the recruitment tag of the target recruiter are calculated, and the matching values ​​are sorted from high to low. Recommendation information is generated based on a preset number of job seekers ranked first and pushed to the target recruiter.

2. The method for recommending job seeker information and recruiter information according to claim 1, wherein the step of calculating the matching values ​​between the recruitment tags of the plurality of matched recruiters and the job search tags of the target job seeker comprises: Determining a plurality of preset dimensions, wherein the preset dimensions include at least one of the following: distance, salary, industry, function, and skill; Calculate the matching values ​​between the recruitment tags of multiple recruiters and the job search tags of target job seekers under various preset dimensions; Obtaining a job applicant's resume and behavioral data, determining the job applicant's preference values ​​in multiple preset dimensions based on the resume and behavioral data, and determining the job applicant's weights in the multiple preset dimensions based on the percentage of the preference value of each preset dimension in the total; Based on the matching values ​​and weights of the job seekers in various preset dimensions, the matching values ​​between the recruitment tags of the matched multiple recruiters and the job seeking tag of the target job seeker are obtained.

3. The method for recommending job seeker information and recruiter information according to claim 2, wherein the step of calculating the matching values ​​between the job-seeking tags of the matched plurality of job seekers and the recruiting tags of the target recruiter comprises: Calculate the matching values ​​between the recruitment tags of multiple recruiters and the job search tags of target job seekers under various preset dimensions; Obtaining recruitment information and behavioral data of a recruiter, determining preference values ​​of the recruiter in a plurality of preset dimensions based on the recruitment information and behavioral data, and determining the weight of the recruiter in the preset dimension based on the percentage of the preference value of each preset dimension in the total; Based on the matching values ​​and weights of the recruiters in various preset dimensions, the matching values ​​between the job-seeking tags of the matched multiple job seekers and the recruitment tags of the target recruiter are obtained.

4. The method for recommending job seeker information and recruiter information according to claim 2 or 3, wherein the behavioral data includes browsing data of the job seeker or recruiter when the job seeker or recruiter is online.

5. The method for recommending job seeker information and recruiter information according to claim 1, wherein after obtaining the job seeker's or recruiter's demand information, the method further comprises: One or more job search tags are calculated based on one or more preset option contents in the job application resume, and one or more recruitment tags are calculated based on one or more preset option contents in the recruitment information, where each of the preset options corresponds to a tag type, and the data calculated based on the option content is used as the tag content.

6. The method for recommending job seeker information and recruiter information according to claim 5, further comprising: Determining one or more main tags and one or more sub-tags among the plurality of job-seeking tags, and determining one or more main tags and one or more sub-tags among the plurality of recruitment tags; The filtering of the plurality of recruiters includes: filtering the plurality of recruiters using a main tag in the job search tag to obtain a recruiter set, and filtering users in the recruiter set using a subtag in the job search tag; The filtering of the plurality of job seekers includes: using a main tag in the recruitment tag to filter the plurality of job seekers to obtain a job seeker set, and using a sub-tag in the recruitment tag to filter users in the job seeker set.

7. The method for recommending job seeker information and recruiter information according to claim 1, wherein: When it is detected that a job seeker is online, the recruiter recommendation information is pushed to the job seeker by means of a pop-up window; when it is detected that a recruiter is online, the job seeker recommendation information is pushed to the recruiter by means of a pop-up window.

8. A job seeker information and recruiter information recommendation system, comprising: A tag acquisition module is configured to acquire the demand information of job seekers or recruiters, where the demand information of job seekers is a resume and the demand information of recruiters is a recruitment information, and extract multiple keywords from the demand information according to semantics and obtain corresponding demand semantic tags; The tag acquisition module is further configured to acquire a preset recruitment domain knowledge graph library, wherein the attributes of the knowledge nodes in the recruitment domain knowledge graph library include function attributes, skill attributes, and industry attributes, wherein nodes having a skill inclusion relationship or an industry inclusion relationship between the node attributes are interconnected; The label acquisition module is further configured to match the obtained demand semantic label with knowledge nodes in the recruitment domain knowledge graph library to obtain multiple knowledge nodes associated with the demand semantic label, and to combine the multiple knowledge nodes in pairs to generate a combined label, wherein if the demand semantic label comes from the job seeker's resume, the combined label is used as the job seeker's job label; if the demand semantic label comes from the recruiter's recruitment information, the combined label is used as the recruiter's recruitment label; a filtering module connected to the tag acquisition module and configured to filter multiple recruiters by matching the target job seeker's job search tag with the recruitment tags of multiple recruiters to obtain multiple recruiters matching the job search tags when recommending recruiter information to the target job seeker; a ranking module, connected to the filtering module and configured to calculate the matching values ​​between the recruitment tags of the plurality of matching recruiters and the job-seeking tags of the target job seeker, and to sort them from high to low according to the matching values; and a push module, connected to the ranking module and configured to generate recommendation information based on a preset number of recruiters ranked first and push it to the target job seeker; The filtering module is further configured to filter the multiple job seekers by matching the recruitment tag of the target recruiter with the job search tags of the multiple job seekers when recommending job seeker information to the target recruiter, so as to obtain multiple job seekers matching the recruitment tags; The sorting module is further configured to calculate the matching values ​​between the job-seeking tags of the matched multiple job seekers and the recruitment tags of the target recruiter, and sort them from high to low according to the matching values; and the pushing module is further configured to generate recommendation information based on a preset number of job seekers ranked first and push it to the target recruiter.

9. The job seeker information and recruiter information recommendation system according to claim 8, wherein the ranking module is specifically configured as follows: Determining a plurality of preset dimensions, wherein the preset dimensions include at least one of the following: distance, salary, industry, function, and skill; Calculate the matching values ​​between the recruitment tags of multiple recruiters and the job search tags of target job seekers under various preset dimensions; Obtaining a job applicant's resume and behavioral data, determining the job applicant's preference values ​​in multiple preset dimensions based on the resume and behavioral data, and determining the job applicant's weights in the multiple preset dimensions based on the percentage of the preference value of each preset dimension in the total; Based on the matching values ​​and weights of the job seekers in various preset dimensions, the matching values ​​between the recruitment tags of the matched multiple recruiters and the job seeking tag of the target job seeker are obtained.

10. The job seeker information and recruiter information recommendation system according to claim 9, wherein the ranking module is specifically configured as follows: Calculate the matching values ​​between the recruitment tags of multiple recruiters and the job search tags of target job seekers under various preset dimensions; Obtaining recruitment information and behavioral data of a recruiter, determining preference values ​​of the recruiter in a plurality of preset dimensions based on the recruitment information and behavioral data, and determining the weight of the recruiter in the preset dimension based on the percentage of the preference value of each preset dimension in the total; Based on the matching values ​​and weights of the recruiters in various preset dimensions, the matching values ​​between the job-seeking tags of the matched multiple job seekers and the recruitment tags of the target recruiter are obtained.

Images