Object matching method and device, electronic equipment and storage medium

Abstract

The invention discloses an object matching method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring a first to-be-matched object with a first effective object feature and a second to-be-matched object with a second effective object feature; inputting the first to-be-matched object and the second to-be-matched object into a pre-training object matching network comprising a first feature processing structure of the first to-be-matched object and a second feature processing structure of the second to-be-matched object; vectorizing the first effective object feature into a first vector through a first feature processing structure, and vectorizing the second effective object feature into a second vector through a second feature processing structure; and determining a matching result of the first to-be-matched object and the second to-be-matched object according to the matching degree of the first vector and the second vector. According to the technical scheme, the understanding degree of the BERT-based semantic expression model on the to-be-matched object in a recruitment field object matching scene with a large data size is improved, and the object matching accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an object matching method, device, electronic device and storage medium. Background Art

[0002] Resumes and job matching accurately matches job seekers with recruiters by analyzing the browsing, submission, and reply behavior data of both parties accumulated on online recruitment platforms. Similar resume matching accurately identifies and recommends job seekers with similar characteristics by analyzing the browsing, submission, and reply behaviors of recruiters to resumes. Similar job matching accurately identifies and recommends jobs with similar characteristics by analyzing the browsing, submission, and reply behaviors of job seekers to positions. The job seeker's resume (hereinafter referred to as CV) records in detail the job search intention, work experience, educational background and other information in text form; the recruiter's job description (hereinafter referred to as JD) also comprehensively records the job title, job category, job skills required, job responsibilities and job requirements in text form.

[0003] At present, the implementation schemes for matching resumes with positions, similar resumes or similar positions usually use BERT (Bidirectional Encoder Representation from Transformers) and its derivative models: directly input personal resumes and / or job texts into the BERT model, and use its self-attention mechanism to obtain the vector representation of the text as a whole and each word. In practice, the overall vector can be used or the word vector can be average-pooled before use. Summary of the invention

[0004] The present invention provides an object matching method, device, electronic device and storage medium, so as to provide a method for applying a BERT-based semantic expression model to resume and job matching, similar resume matching or similar job matching, thereby improving the BERT-based semantic expression model's understanding of matching objects in object matching scenarios in the recruitment field with large data volumes, and improving the accuracy of object matching.

[0005] According to one aspect of the present invention, there is provided an object matching method, the method comprising:

[0006] Acquire a first to-be-matched object and a second to-be-matched object, wherein the first to-be-matched object has a first valid object feature, and the second to-be-matched object has a second valid object feature;

[0007] Inputting the first to-be-matched object and the second to-be-matched object into a pre-trained object matching network, wherein the pre-trained object matching network includes a first feature processing structure of the first to-be-matched object and a second feature processing structure of the second to-be-matched object;

[0008] Vectorizing the first valid object feature by the first feature processing structure to obtain a first vector, and vectorizing the second valid object feature by the second feature processing structure to obtain a second vector;

[0009] A matching result between the first to-be-matched object and the second to-be-matched object is determined according to a matching degree between the first vector and the second vector.

[0010] According to another aspect of the present invention, there is provided an object matching device, the device comprising:

[0011] An object acquisition module, used to acquire a first to-be-matched object and a second to-be-matched object, wherein the first to-be-matched object has a first valid object feature, and the second to-be-matched object has a second valid object feature;

[0012] An object input module, used for inputting the first to-be-matched object and the second to-be-matched object into a pre-trained object matching network, wherein the pre-trained object matching network includes a first feature processing structure of the first to-be-matched object and a second feature processing structure of the second to-be-matched object;

[0013] A vector acquisition module, configured to vectorize the first valid object feature through the first feature processing structure to obtain a first vector, and vectorize the second valid object feature through the second feature processing structure to obtain a second vector;

[0014] The object matching result determination module is used to determine the matching result between the first to-be-matched object and the second to-be-matched object according to the matching degree between the first vector and the second vector.

[0015] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0016] at least one processor; and

[0017] a memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the object matching method described in any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the object matching method described in any embodiment of the present invention when executed.

[0020] The technical solution of the embodiment of the present invention is as follows: by obtaining a first object to be matched and a second object to be matched, the first object to be matched has a first valid object feature, and the second object to be matched has a second valid object feature; the first object to be matched and the second object to be matched are input into a pre-trained object matching network, the pre-trained object matching network includes a first feature processing structure of the first object to be matched and a second feature processing structure of the second object to be matched; the first valid object feature is vectorized by the first feature processing structure to obtain a first vector, and the second valid object feature is vectorized by the second feature processing structure to obtain a second vector; according to the degree of matching between the first vector and the second vector, the matching result between the first object to be matched and the second object to be matched is determined, and by using the object to be matched containing the valid object feature as the input of the BERT model, a method for applying a BERT model to the fields of resume and job matching, similar resume matching or similar job matching is provided, which improves the understanding of the BERT-based semantic expression model in the object matching scenario in the recruitment field with a large amount of data, and improves the accuracy of object matching.

[0021] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1a A flowchart of an object matching method provided in Embodiment 1 of the present invention;

[0024] Figure 1b A schematic diagram of an object matching network provided by an embodiment of the present invention when a first object to be matched is inconsistent with a second object to be matched;

[0025] Figure 1c A schematic diagram of an object matching network provided by an embodiment of the present invention when a first object to be matched and a second object to be matched are both job descriptions of a recruiter;

[0026] Figure 1d A schematic diagram of an object matching network provided by an embodiment of the present invention in which a first object to be matched and a second object to be matched are both resumes of job seekers;

[0027] Figure 2 A schematic diagram of the structure of an object matching device provided in Embodiment 2 of the present invention;

[0028] Figure 3 A schematic diagram of the structure of an electronic device of an object matching method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the 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 creative work should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] Embodiment 1

[0032] Figure 1a This is a flow chart of an object matching method provided in the first embodiment of the present invention. This embodiment is applicable to the case of matching text objects. The method can be executed by an object matching device. The object matching device can be implemented in the form of hardware and / or software. The object matching device can be configured in a server. Figure 1a As shown, the method includes:

[0033] S110: Acquire a first object to be matched and a second object to be matched, wherein the first object to be matched has a first valid object feature, and the second object to be matched has a second valid object feature.

[0034] The first object to be matched and the second object to be matched may be text objects, such as a resume or a job description. That is, the first object to be matched may be a resume or a job description, and the second object to be matched may be a resume or a job description. Accordingly, the first valid object feature may refer to a text feature of a resume or a valid text feature in a job description.

[0035] In this embodiment, a first to-be-matched object having a first valid text feature and a second to-be-matched object having a second valid text feature may be obtained.

[0036] S120: Input the first object to be matched and the second object to be matched into a pre-trained object matching network, where the pre-trained object matching network includes a first feature processing structure of the first object to be matched and a second feature processing structure of the second object to be matched.

[0037] The pre-trained object matching network can be obtained by training based on the twin network method of the BERT model. The pre-trained object matching network can be trained in the following way: obtain the first category of objects and the second category of objects; the first category of objects corresponds to the first object to be matched, and the second category of objects corresponds to the second object to be matched; construct training samples according to user behaviors between the first category of objects and the second category of objects; select sample features for the training samples according to the description text of the first category of objects and the description text of the second category of objects; perform at least one round of training on the initial object matching network according to the training samples and sample features to obtain the pre-trained object matching network.

[0038] Among them, the first type of object can be a sample object corresponding to the first object to be matched. For example, if the first object to be matched is a personal resume, the first type of object can be an object of the type of personal resume; if the first object to be matched is a job description, the second type of object can be an object of the type of job description. Similarly, the second type of object can be a sample object corresponding to the second object to be matched. For example, if the second object to be matched is a personal resume, the second type of object can be an object of the type of personal resume; if the second object to be matched is a job description, the second type of object can be an object of the type of job description.

[0039] Optionally, constructing training samples based on user behavior between first-category objects and second-category objects may include: if the user behavior between first-category objects and second-category objects is positive behavior, constructing the first-category objects and the second-category objects as a group of positive samples; if the user behavior between second-category objects and second-category objects is negative behavior, constructing the first-category objects and the second-category objects as a group of negative samples.

[0040] Exemplarily, positive samples can be constructed based on positive behaviors between users of the first and second categories of objects on the recruitment platform, and negative samples can be constructed based on negative behaviors. Positive behaviors can refer to behaviors that are relatively recognized between the first and second categories of objects, such as behaviors such as long-term browsing, initiating inquiries, and positive replies between users of the first and second categories of objects. Negative behaviors can refer to behaviors that are relatively unrecognizable between at least one party of the first and second categories of objects, such as not browsing or browsing quickly, not initiating inquiries after browsing, or negative replies.

[0041] Optionally, according to the positive degree of the positive samples, the positive samples are divided into the first category of positive samples and the second category of positive samples, and the positive degree of the first category of positive samples is higher than that of the second category of positive samples; according to the negative degree of the negative samples, the negative samples are divided into the first category of negative samples and the second category of negative samples, and the negative degree of the first category of negative samples is higher than that of the second category of negative samples. Specifically, the training samples can be subjected to noise removal and hard negative sample screening by LLM (Large Language Model). Specifically, for paired first category objects and second category objects, LLM can be used to determine whether the two are related, and remove noise data caused by misoperation or low-quality behavior. On the other hand, hard negative samples can improve the distinguishing ability of the model, so LLM is used to extract hard negative samples from negative samples, such as "robot programming teacher" and "art teacher". In addition, a preset prompt word project can be used to guide LLM to generate small batches of hard negative samples. Easy negative samples can also be constructed by negative sampling of the entire library (for example, the talent database of the recruitment platform) to achieve the purpose of exposing the model to the world.

[0042] Exemplarily, if the first type of objects are personal resumes and the second type of objects are job descriptions, or if the first type of objects are job descriptions and the second type of objects are personal resumes, the training samples may be as shown in Table 1.

[0043] Table 1

[0044]

[0045]

[0046] For example, if both the first and second objects are resumes, the method for removing noise data in the training sample can be: according to the resume sequence that generates positive behavior on the recruiter, after forming resume pairs by sliding window, use LLM to determine whether the two are related, and remove noise data caused by misoperation or low-quality behavior. The training sample can be shown in Table 2.

[0047] Table 2

[0048]

[0049]

[0050] For example, if both the first and second objects are job descriptions, the method for removing noise data in the training sample can be: according to the job description sequence that generates positive behavior of the job seeker, after forming a job description pair by means of a sliding window, use LLM to determine whether the two are related, and remove noise data caused by misoperation or low-quality behavior. The training sample can be shown in Table 3.

[0051] Table 3

[0052]

[0053] Referring to Tables 1 to 3, strong positive samples are equivalent to the first type of positive samples mentioned above, weak positive samples are equivalent to the second type of positive samples mentioned above, easy negative samples are equivalent to the second type of negative samples mentioned above, and hard negative samples are equivalent to the first type of negative samples mentioned above.

[0054] In this embodiment, LLM is used to clean the data to remove noise, which can improve the accuracy and reliability of the training samples. In addition, hard negative samples are crucial to improving the model effect. LLM can also be used to screen negative behavior samples to construct a hard sample set, further improving the model's ability to distinguish.

[0055] In this embodiment, effective sample features can be selected for training samples based on the description text of the first category object and the description text of the second category object, that is, the text content of the first category object and the second category object. Since the text content of the first category object and the second category object itself can help describe the demands of the corresponding users, but not all text content can improve the expression ability of the object matching network, effective feature selection can be performed through ablation experiments. For example, the job description of the recruiter can select features such as those shown in Table 4, and the personal resume of the job seeker can select features such as those shown in Table 5.

[0056] Table 4

[0057]

[0058] Table 5

[0059]

[0060]

[0061] S130 , vectorizing the first valid object feature by using the first feature processing structure to obtain a first vector, and vectorizing the second valid object feature by using the second feature processing structure to obtain a second vector.

[0062] In the pre-training process of the object matching network, for example, if the first object to be matched is a job seeker's resume and the second object to be matched is a job description of a recruiter, or if the first object to be matched is a job description of a recruiter and the second object to be matched is a job seeker's resume, the object matching network can be as follows: Figure 1b As shown, it includes the job description processing structure JD tower on the left and the personal resume processing structure CV tower on the right.

[0063] Optionally, the resume features corresponding to the job seeker's resume can be multi-vectorized according to the resume module through the resume processing structure, and the multi-vectorized representation results can be processed through the attention mechanism to obtain the first vector; the job description features corresponding to the job description of the recruiter can be vectorized through the job description processing structure to obtain the second vector. That is, the job description and the features of the job description can be input into the JD tower for vectorization, and the resume and the features of the resume can be input into the CV tower to be multi-vectorized according to the resume modules such as job search intention, work experience and professional skills, and then the multi-vectorized results can be processed by the self-attention mechanism. During the pre-training process of the object matching network, a loss function can also be set to optimize the parameters of the JD tower and the CV tower. Weights can be shared between the JD tower and the CV tower.

[0064] For example, if the first object to be matched is a job description of a first employer, and the second object to be matched is a job description of a second employer, the pre-trained object matching network can be as follows: Figure 1c As shown, it includes a first job description processing structure (JD tower on the left) and a second job description processing structure (JD tower on the right), and the first job description processing structure and the second job description processing structure can share weights.

[0065] For example, if the first object to be matched is the resume of the first job seeker, and the second object to be matched is the resume of the second job seeker, the pre-trained object matching network can be as follows: Figure 1d As shown, it includes a first personal resume processing structure (CV tower on the left) and a second personal resume processing structure (CV tower on the right); the first personal resume processing structure and the second personal resume processing structure can share weights.

[0066] In this embodiment, positive behavior data is used to construct training samples, and negative sampling and contrastive learning within a batch are used to improve the similarity of matching JD and CV pairs, JD and CV pairs, while reducing the similarity of unmatched pairs.

[0067] In this embodiment, the input of the object matching network can also be optimized: for example, by extending the token_type_id inside the BERT model to distinguish different text domains in JD and CV, values ​​other than 0 and 1 are used to identify different features. This transformation can effectively improve the model performance while keeping the model complexity unchanged.

[0068] In this embodiment, a multi-vector representation method is used for the personal resume CV: since the personal resume contains multiple modules such as job application intention, work experience, educational background, etc., each module independently learns a vector and jointly participates in model training and downstream tasks, which can comprehensively mine and characterize the information of the personal resume.

[0069] S140: Determine a matching result between the first to-be-matched object and the second to-be-matched object according to a matching degree between the first vector and the second vector.

[0070] In the actual application process of the pre-trained object matching network, after obtaining the first vector corresponding to the first object to be matched and the second vector of the second object to be matched, the matching result (person-job matching result, similar resumes, similar positions) can be determined based on the matching degree between the first vector and the second vector.

[0071] The technical solution of the embodiment of the present invention is as follows: by obtaining a first object to be matched and a second object to be matched, the first object to be matched has a first valid object feature, and the second object to be matched has a second valid object feature; the first object to be matched and the second object to be matched are input into a pre-trained object matching network, the pre-trained object matching network includes a first feature processing structure of the first object to be matched and a second feature processing structure of the second object to be matched; the first valid object feature is vectorized by the first feature processing structure to obtain a first vector, and the second valid object feature is vectorized by the second feature processing structure to obtain a second vector; according to the degree of matching between the first vector and the second vector, the matching result between the first object to be matched and the second object to be matched is determined, and by using the object to be matched containing the valid object feature as the input of the BERT model, a method for applying a BERT model to the fields of resume and job matching, similar resume matching or similar job matching is provided, which improves the understanding of the BERT-based semantic expression model in the object matching scenario in the recruitment field with a large amount of data, and improves the accuracy of object matching.

[0072] Embodiment 2

[0073] Figure 2 FIG. 1 is a schematic diagram of the structure of an object matching device provided in Embodiment 2 of the present invention. Figure 2 As shown, the device includes: an object acquisition module 210, an object input module 220, a vector acquisition module 230 and an object matching result determination module 240. Among them:

[0074] An object acquisition module 210 is used to acquire a first to-be-matched object and a second to-be-matched object, wherein the first to-be-matched object has a first valid object feature, and the second to-be-matched object has a second valid object feature;

[0075] An object input module 220, configured to input the first to-be-matched object and the second to-be-matched object into a pre-trained object matching network, wherein the pre-trained object matching network includes a first feature processing structure of the first to-be-matched object and a second feature processing structure of the second to-be-matched object;

[0076] A vector acquisition module 230, configured to vectorize the first valid object feature through the first feature processing structure to obtain a first vector, and vectorize the second valid object feature through the second feature processing structure to obtain a second vector;

[0077] The object matching result determination module 240 is used to determine the matching result between the first to-be-matched object and the second to-be-matched object according to the matching degree between the first vector and the second vector.

[0078] The technical solution of the embodiment of the present invention is as follows: by obtaining a first object to be matched and a second object to be matched, the first object to be matched has a first valid object feature, and the second object to be matched has a second valid object feature; the first object to be matched and the second object to be matched are input into a pre-trained object matching network, the pre-trained object matching network includes a first feature processing structure of the first object to be matched and a second feature processing structure of the second object to be matched; the first valid object feature is vectorized by the first feature processing structure to obtain a first vector, and the second valid object feature is vectorized by the second feature processing structure to obtain a second vector; according to the degree of matching between the first vector and the second vector, the matching result between the first object to be matched and the second object to be matched is determined, and by using the object to be matched containing the valid object feature as the input of the BERT model, a method for applying a BERT model to the fields of resume and job matching, similar resume matching or similar job matching is provided, which improves the understanding of the BERT-based semantic expression model in the object matching scenario in the recruitment field with a large amount of data, and improves the accuracy of object matching.

[0079] Optionally, the pre-trained object matching network is trained in the following manner:

[0080] Acquire a first category of objects and a second category of objects; the first category of objects corresponds to the first object to be matched, and the second category of objects corresponds to the second object to be matched;

[0081] constructing training samples according to user behaviors between the first category of objects and the second category of objects;

[0082] selecting sample features for the training samples according to the description texts of the first category of objects and the description texts of the second category of objects;

[0083] The initial object matching network is trained for at least one round according to the training samples and the sample features to obtain the pre-trained object matching network.

[0084] Optionally, the object matching device further includes a training sample construction module, which is used to:

[0085] If the user behavior between the first category of objects and the second category of objects is a positive behavior, constructing the first category of objects and the second category of objects as a group of positive samples;

[0086] If the user behavior between the second-category objects and the second-category objects is a negative behavior, the first-category objects and the second-category objects are constructed as a group of negative samples.

[0087] Optionally, the object matching device further includes a sample distinguishing module, which is used to:

[0088] According to the positivity degree of the positive samples, the positive samples are divided into a first category of positive samples and a second category of positive samples, wherein the positivity degree of the first category of positive samples is higher than that of the second category of positive samples;

[0089] According to the negativity degree of the negative samples, the negative samples are divided into a first category of negative samples and a second category of negative samples, and the negativity degree of the first category of negative samples is higher than that of the second category of negative samples.

[0090] Optionally, the first object to be matched is a resume of a job seeker, and the second object to be matched is a job description of a recruiter; the pre-trained object matching network includes a resume processing structure and a job description processing structure; the resume processing structure and the job description processing structure share weights;

[0091] Correspondingly, the vector acquisition module can be used for:

[0092] The resume features corresponding to the resume of the job seeker are multi-vectorized according to the resume module by the resume processing structure, and the multi-vectorized representation result is processed by the attention mechanism to obtain the first vector;

[0093] The job description feature vector corresponding to the job description of the recruiter is vectorized by the job description processing structure to obtain the second vector.

[0094] Optionally, the first object to be matched is a job description of a first recruiter, and the second object to be matched is a job description of a second recruiter; the pre-trained object matching network includes a first job description processing structure and a second job description processing structure; the first job description processing structure and the second job description processing structure share weights;

[0095] Correspondingly, the vector acquisition module can be used for:

[0096] Processing the job description features corresponding to the job description of the first recruiter by the first job description processing structure to obtain the first vector;

[0097] The second vector is obtained by processing the job description features corresponding to the second recruiter's job description through the second job description processing structure.

[0098] Optionally, the first object to be matched is a resume of a first job seeker, and the second object to be matched is a resume of a second job seeker; the pre-trained object matching network includes a first resume processing structure and a second resume processing structure; the first resume processing structure and the second resume processing structure share weights;

[0099] Correspondingly, the vector acquisition module can be used for:

[0100] The first resume feature corresponding to the first job seeker's resume is multi-vectorized according to the resume module by the first resume processing structure, and the multi-vectorized representation result is processed by the attention mechanism to obtain the first vector;

[0101] The second resume feature corresponding to the second job seeker's resume is multi-vectorized according to the resume module through the second resume processing structure, and the multi-vectorized representation result is processed through the attention mechanism to obtain the second vector.

[0102] The object matching device provided in the embodiment of the present invention can execute the object matching method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0103] Embodiment 3

[0104] Figure 3 A schematic diagram of an electronic device 300 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers or various forms of mobile devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0105] like Figure 3As shown, the electronic device 300 includes at least one processor 301, and a memory connected to the at least one processor 301, such as a read-only memory (ROM) 302, a random access memory (RAM) 303, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 301 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 302 or the computer program loaded from the storage unit 308 to the random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 can also be stored. The processor 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0106] A number of components in the electronic device 300 are connected to the I / O interface 305, including: an input unit 306, such as a keyboard, a mouse, etc.; an output unit 307, such as various types of displays, speakers, etc.; a storage unit 308, such as a disk, an optical disk, etc.; and a communication unit 309, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 309 allows the electronic device 300 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0107] The processor 301 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 301 performs the various methods and processes described above, such as the object matching method.

[0108] In some embodiments, the object matching method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 300 via the ROM 302 and / or the communication unit 309. When the computer program is loaded into the RAM 303 and executed by the processor 301, one or more steps of the object matching method described above may be performed. Alternatively, in other embodiments, the processor 301 may be configured to perform the object matching method in any other appropriate manner (e.g., by means of firmware).

[0109] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0110] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0111] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0112] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0113] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0114] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0115] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0116] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An object matching method, characterized in that: include: Acquire a first to-be-matched object and a second to-be-matched object, wherein the first to-be-matched object has a first valid object feature, and the second to-be-matched object has a second valid object feature; Inputting the first to-be-matched object and the second to-be-matched object into a pre-trained object matching network, wherein the pre-trained object matching network includes a first feature processing structure of the first to-be-matched object and a second feature processing structure of the second to-be-matched object; Vectorizing the first valid object feature by the first feature processing structure to obtain a first vector, and vectorizing the second valid object feature by the second feature processing structure to obtain a second vector; A matching result between the first to-be-matched object and the second to-be-matched object is determined according to a matching degree between the first vector and the second vector.

2. The method according to claim 1, characterized in that The pre-trained object matching network is trained in the following way: Acquire a first category of objects and a second category of objects; the first category of objects corresponds to the first object to be matched, and the second category of objects corresponds to the second object to be matched; constructing training samples according to user behaviors between the first category of objects and the second category of objects; selecting sample features for the training samples according to the description texts of the first category of objects and the description texts of the second category of objects; The initial object matching network is trained for at least one round according to the training samples and the sample features to obtain the pre-trained object matching network.

3. The method according to claim 2, characterized in that Constructing training samples according to user behaviors between the first category of objects and the second category of objects, including: If the user behavior between the first category of objects and the second category of objects is a positive behavior, constructing the first category of objects and the second category of objects as a group of positive samples; If the user behavior between the second-category objects and the second-category objects is a negative behavior, the first-category objects and the second-category objects are constructed as a group of negative samples.

4. The method according to claim 3, characterized in that Also includes: According to the positivity degree of the positive samples, the positive samples are divided into a first category of positive samples and a second category of positive samples, wherein the positivity degree of the first category of positive samples is higher than that of the second category of positive samples; According to the negativity degree of the negative samples, the negative samples are divided into a first category of negative samples and a second category of negative samples, and the negativity degree of the first category of negative samples is higher than that of the second category of negative samples.

5. The method according to claim 1, characterized in that The first object to be matched is a job seeker's resume, and the second object to be matched is a job description of a recruiter; the pre-trained object matching network includes a resume processing structure and a job description processing structure; the resume processing structure and the job description processing structure share weights; Vectorizing the first valid object feature to obtain a first vector through the pre-trained object matching network, and vectorizing the second valid object feature to obtain a second vector, including: The resume features corresponding to the resume of the job seeker are multi-vectorized according to the resume module by the resume processing structure, and the multi-vectorized representation result is processed by the attention mechanism to obtain the first vector; The job description feature vector corresponding to the job description of the recruiter is vectorized by the job description processing structure to obtain the second vector.

6. The method according to claim 1, characterized in that The first object to be matched is a job description of a first recruiter, and the second object to be matched is a job description of a second recruiter; the pre-trained object matching network includes a first job description processing structure and a second job description processing structure; the first job description processing structure and the second job description processing structure share weights; Vectorizing the first valid object feature to obtain a first vector through the pre-trained object matching network, and vectorizing the second valid object feature to obtain a second vector, including: Processing the job description features corresponding to the job description of the first recruiter by the first job description processing structure to obtain the first vector; The second vector is obtained by processing the job description features corresponding to the second recruiter's job description through the second job description processing structure.

7. The method according to claim 1, characterized in that The first object to be matched is a resume of a first job seeker, and the second object to be matched is a resume of a second job seeker; the pre-trained object matching network includes a first resume processing structure and a second resume processing structure; the first resume processing structure and the second resume processing structure share weights; Vectorizing the first valid object feature to obtain a first vector through the pre-trained object matching network, and vectorizing the second valid object feature to obtain a second vector, including: The first resume feature corresponding to the first job seeker's resume is multi-vectorized according to the resume module by the first resume processing structure, and the multi-vectorized representation result is processed by the attention mechanism to obtain the first vector; The second resume feature corresponding to the second job seeker's resume is multi-vectorized according to the resume module through the second resume processing structure, and the multi-vectorized representation result is processed through the attention mechanism to obtain the second vector.

8. An object matching device, characterized in that: include: An object acquisition module, used to acquire a first to-be-matched object and a second to-be-matched object, wherein the first to-be-matched object has a first valid object feature, and the second to-be-matched object has a second valid object feature; An object input module, used for inputting the first to-be-matched object and the second to-be-matched object into a pre-trained object matching network, wherein the pre-trained object matching network includes a first feature processing structure of the first to-be-matched object and a second feature processing structure of the second to-be-matched object; A vector acquisition module, configured to vectorize the first valid object feature through the first feature processing structure to obtain a first vector, and vectorize the second valid object feature through the second feature processing structure to obtain a second vector; The object matching result determination module is used to determine the matching result between the first to-be-matched object and the second to-be-matched object according to the matching degree between the first vector and the second vector.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the object matching method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the object matching method according to any one of claims 1 to 7 when executed.

Images