Intelligent communication method and system based on recruitment scene of colleges and universities

Abstract

The invention provides an intelligent communication method and system based on a college recruitment scene, and relates to the technical field of intelligent recruitment, and the method comprises the steps: obtaining an online resume submitted by an applicant, and analyzing the online resume through a document analysis library to generate an original disordered text set; based on the original disordered text set, utilizing a pre-training language model to extract academic entity information, and generating a research direction vector; constructing an applicant academic portrait in combination with the academic entity and the research direction vector; calculating the matching degree between the applicant and the post through a quantitative matching model, judging whether the matching degree exceeds a preset threshold value, and if so, ending the process; otherwise, fusing the applicant portrait with college static information to construct a knowledge graph; generating personalized greeting words based on the map and the language model, and receiving natural language questions of candidates; and the system analyzes the question to generate a query statement, executes and obtains a structured answer in the knowledge graph, and finally converts the structured answer into a natural language reply through a language model and feeds back the natural language reply to an applicant.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent recruitment, in particular to an intelligent communication method and system based on a university recruitment scenario. BACKGROUND

[0002] University teacher recruitment is a core link to maintain academic vitality and scientific research competitiveness. With the acceleration of global academic exchanges and the frequent flow of talents, universities not only need to quickly and accurately screen the academic level and research direction of job seekers in the talent introduction process, but also need to have higher efficiency and flexibility in cross-border communication and talent attraction. How to use intelligent means to solve the pain points in the traditional recruitment mode has become an important direction that needs to be broken through in the university talent strategy.

[0003] The current university teacher recruitment mainly relies on manual review and communication. Job seekers need to submit complete academic resumes including academic achievements, research experience and teaching experience, which are reviewed and compared by recruitment experts to determine whether their research direction meets the development plan of the university. In cross-border recruitment, the employer usually communicates with the job seeker through email or video conference, and the talent attraction is mostly done by publishing job information on the official website or academic recruitment website, and passively waiting for the job seeker to submit an application.

[0004] However, the existing university teacher recruitment method is difficult to achieve efficient and fair evaluation due to the complex content of the recruitment materials, the time-consuming review process and the influence of subjective factors. At the same time, cross-border recruitment is limited by time zone differences and delayed responses, and the communication efficiency is low. In the absence of initiative and personalized outreach, it is difficult to lock in and compete for top talent in a timely manner, and it is easy to lose high-quality job seekers. SUMMARY

[0005] In view of the deficiencies of the prior art, the purpose of the embodiments of the present application is to provide an intelligent communication method based on a university recruitment scenario, which can solve the technical problems of the existing university teacher recruitment method, which is difficult to achieve efficient and fair evaluation due to the complex content of the recruitment materials, the time-consuming review process and the influence of subjective factors. At the same time, cross-border recruitment is limited by time zone differences and delayed responses, and the communication efficiency is low. In the absence of initiative and personalized outreach, it is difficult to lock in and compete for top talent in a timely manner, and it is easy to lose high-quality job seekers.

[0006] The first aspect of the embodiments of the present application proposes an intelligent communication method based on a university recruitment scenario, comprising: S1: obtaining an online resume submitted by a job seeker; S2: parsing the online resume through a document parsing library to extract a set of original unordered texts; S3: extracting academic entity information with academic meaning based on the original unordered text set through a pre-trained language model, and determining an academic research direction vector of the candidate based on the academic entity information; S4: constructing an academic portrait of the candidate based on the academic entity information and the academic research direction vector; S5: calculating a post matching degree of the candidate and the job position based on the academic portrait of the candidate through a quantitative matching model; S6: determining whether the post matching degree is greater than a preset post matching degree; if yes, proceeding to S7; otherwise, terminating the process of the candidate; S7: fusing the academic portrait of the candidate and pre-extracted college static information to construct a knowledge graph; S8: generating a greeting for the candidate based on the knowledge graph and the language model, and receiving a natural language question of the candidate; S9: analyzing the natural language question of the candidate to generate a knowledge graph query statement; S10: executing the knowledge graph query statement to obtain corresponding structured answer data, and generating a natural language reply based on the structured answer data through the language model and feeding back to the candidate.

[0007] The second aspect of the embodiment of the application provides an intelligent communication system based on a university recruitment scenario, comprising a processor and a memory. The memory stores programs or instructions executable on the processor, and the programs or instructions are executed by the processor to implement the steps of the intelligent communication method based on the university recruitment scenario.

[0008] The third aspect of the embodiment of the application provides a readable storage medium, and the readable storage medium stores programs or instructions, and the programs or instructions are executed by the processor to implement the steps of the intelligent communication method based on the university recruitment scenario.

[0009] The technical scheme provided by the embodiment of the application has at least the following beneficial effects: In the embodiment of the application, the candidate's resume is automatically analyzed, academic entity information is extracted, a research direction vector is determined, an academic portrait is constructed, and a post matching degree is calculated, so that the complicated manual review is converted into a standardized and quantitative calculation process, thereby significantly improving the screening efficiency and reducing subjectivity, realizing efficient and fair evaluation; at the same time, the system can also fuse the academic portrait and college information to construct a knowledge graph, automatically generate personalized greetings, and realize instant interaction and natural language reply across time zones based on knowledge graph question and answer, thereby improving the initiative of recruitment and effectively avoiding the loss of high-quality talents. Attached Figure Description

[0010] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0011] Figure 1 This is a flowchart illustrating an intelligent communication method based on a university recruitment scenario provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an intelligent communication system based on a university recruitment scenario provided by an embodiment of the present invention. Detailed Implementation

[0012] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0013] The intelligent communication method based on university recruitment scenarios provided by the present invention will be described in detail below with reference to the accompanying drawings and through specific embodiments and application scenarios.

[0014] Reference manual attached Figure 1 The diagram illustrates a flowchart of an intelligent communication method based on a university recruitment scenario provided by an embodiment of the present invention.

[0015] This invention provides an intelligent communication method based on a university recruitment scenario, which may include the following steps: S1: Obtain online resumes submitted by job applicants.

[0016] S2: Use a document parsing library to parse online resumes and extract the original unordered text set.

[0017] The document parsing library is a tool component that can parse the content of common electronic documents (such as PDF, Word, PPT, etc.). Its function is to automatically extract complete text information from files with complex formats, remove irrelevant elements such as layout, fonts, and images, and retain only the semantic plain text.

[0018] Optionally, the text parsing library uses PyMuPDF or PDFBox.

[0019] In a possible implementation, S2 specifically includes: S201: receiving an online resume submitted by a candidate and related resume attachment files.

[0020] S202: calling a document parsing library to parse the online resume and the attachment files word by word and section by section.

[0021] S203: based on the parsing result, extracting all the text content of the online resume to form a raw unordered text set.

[0022] In the embodiment of the application, by calling the document parsing library, the online resume submitted by the candidate is parsed word by word and section by section, and a raw unordered text set is formed by extraction. This process realizes compatible processing of multiple document formats. This process can effectively eliminate irrelevant elements such as layout, font, and pictures while preserving all semantic information of the resume, avoiding information omission and deviation caused by manual reading and inputting, significantly improving parsing efficiency and accuracy, and providing high-quality standardized input data for subsequent academic entity recognition and academic portrait construction.

[0023] S3: based on the raw unordered text set, extracting academic entity information with academic meaning through a pre-trained language model, and determining an academic research direction vector of the candidate based on the academic entity information.

[0024] It should be noted that the pre-trained language model is preferably a SciBERT model based on the BERT architecture. The model has different functions at different stages: on the one hand, it uses its named entity recognition capability to identify and extract academic entities from the raw unordered text set. On the other hand, it uses its semantic encoding capability to represent the features of the academic text fragments and generate fixed-length text representation vectors. The above two functions can be realized by the same pre-trained language model without introducing different models.

[0025] In a possible implementation, S3 specifically includes: S301: identifying and extracting academic entity information including a paper list, research keywords, research funds, cooperating candidates, educational background, and academic awards from the raw unordered text set through the language model.

[0026] S302: concatenating the research keywords and the paper titles and abstracts in the paper list into a coherent string, and inputting the coherent string into a word segmentation processing module of the language model to obtain a token sequence.

[0027] Need to explain, token sequence, simply speaking, is the minimum processing unit sequence after the segmentation of natural language text, is the input form when the NLP model (such as BERT, GPT, SciBERT, etc.) understands the text.

[0028] S303: the thesis list is regarded as a vector set, wherein each thesis in the thesis list is spliced by title and abstract: wherein, represents the text vector representation unit of the i th thesis, represents the title text of the i th thesis, represents the abstract text of the i th thesis, represents the thesis list set; S304: the token sequence is pooled to generate a query vector representing the core research direction; S305: using the query vector as the query, the thesis list vector set as the key and value, the attention weight of the query vector to each thesis is calculated through the attention mechanism: wherein, represents the attention weight of the query vector to the i th thesis, represents the activation function, represents the query vector, represents the vector representation of the i th thesis, represents the vector dimension used to scale the inner product result; S306: based on the attention weight, all thesis vectors are aggregated into a context-aware text vector through weighted summation: wherein, represents the text vector, n represents the number of theses, represents the attention weight of the i th thesis, represents the i th thesis vector; S307: the text vector and the query vector are fused and averaged to obtain a fusion vector, and the fusion vector is output as the academic research direction vector of the applicant.

[0029] Note that we discard naive mean pooling and use the keyword vectors as “queries” to dynamically assign weights to each paper. This makes papers that are highly relevant to the core direction (high-impact work) contribute more to the final vector.

[0030] In the embodiments of the present application, by introducing the joint vectorization processing flow of the language model and the attention mechanism, the structured and semantic extraction of the candidate's academic text information is realized. The method can automatically identify academic entities such as papers, keywords, and scientific research funds from unordered original texts, and generate a high-dimensional feature vector that can reflect the core research direction on this basis. Through the attention weighting and fusion operation of the query vector and the paper vector set, not only the representation accuracy of the academic features is improved, but also the perception ability of the model to the semantic association between different research topics is enhanced, thereby realizing the accurate characterization and intelligent matching of the candidate's academic research direction.

[0031] S4: Constructing the academic portrait of the candidate based on the academic entity information and the academic research direction vector.

[0032] Among them, the academic portrait of the candidate refers to the digital and structured information set formed after the text analysis, academic entity identification, and research direction vectorization processing based on the resume and related materials submitted by the candidate. The portrait usually includes structured academic entity information such as paper achievements, research keywords, scientific research funds, educational background, academic awards, and cooperative scholars, as well as a high-dimensional vector representing its research direction. Through this composite data structure, the academic portrait of the candidate can not only fully reflect the candidate's background and achievements in the academic field, but also describe the characteristics of the research direction in a computable way, thereby providing a reliable data foundation for job matching degree calculation, knowledge graph construction, and intelligent communication.

[0033] In one possible implementation, S4 specifically includes: S401: Assigning a unique identifier to the candidate for associating the academic entity information with the academic research direction vector.

[0034] S402: Combining the academic entity information and the academic research direction vector to form a composite data structure.

[0035] S403: Based on the unique identifier, establishing a data storage model including a main model and an associated model.

[0036] For example, the system will assign a unique ID, such as candidate-id: 1101, to the candidate being processed. This ID is the key hub that connects all the information, ensuring that all the data can be accurately associated with the same person.

[0037] Specifically, define the Candidate Profile Model: create a core data structure to represent a candidate. This structure will contain their basic information and that all-important research direction vector.

[0038] Define Relational Models: for information that contains multiple entries, such as the list of papers, educational background, create separate, associated data models for them. For example, for the list of papers, we will have a "Paper" model, each record contains fields such as title, author, journal, publication year, etc., and contains a candidate-id field to indicate which candidate this paper belongs to. Similarly, there will be "Education", "Research Fund", "Awards" models.

[0039] S404: Map the composite data structure to the data storage model to generate the candidate academic portrait.

[0040] It should be noted that building a complete candidate record: combine the above models to form a logical whole. This whole is the specific form of the candidate academic portrait.

[0041] In the embodiments of the present application, through the constructed candidate academic portrait, not only the academic entity information in the resume and the research direction vector can be uniformly stored and managed, ensuring the accurate association and integrity of the data, but also the academic background and research direction characteristics of the candidate can be comprehensively reflected in a structured and modularized manner. The academic portrait provides high-quality basic data for subsequent job matching degree calculation, knowledge graph construction, and intelligent communication, thereby significantly improving the automation degree of the recruitment process and the scientificity of the decision-making.

[0042] S5: Based on the candidate academic portrait, calculate the job matching degree of the candidate and the job by using the quantitative matching model.

[0043] Among them, the quantitative matching model refers to a mathematical model used to calculate the degree of fit between the candidate academic portrait and the job portrait in the job recruitment process. The model considers the similarity (such as cosine similarity) between the candidate's research direction vector and the job research direction vector, the scoring results of the candidate's academic indicators (such as h-index, number of top papers, fund level, etc.), and the matching situation of the job-specific keywords, and outputs a matching score by setting weight coefficients to weight and fuse the results of each part. The score can objectively quantify the degree of fit between the candidate and the job requirements, avoiding the subjectivity and randomness of manual screening, and providing comparable and measurable basis for recruitment decisions.

[0044] In one possible implementation, S5 specifically includes: S501: Define a job profile based on the job requirements of the recruiting unit, wherein the job profile includes core research direction keywords, achievement requirements, and soft requirements of the job.

[0045] S502: Convert the core research direction keywords in the job profile into a job research direction vector through a vectorization method.

[0046] S503: Extract research direction vectors and academic index information from the academic profile of the job applicant.

[0047] The research direction vector is a fixed dimension numerical vector generated by pre-training a language model on the research keywords, paper titles, and abstracts extracted from the applicant's resume and other academic text fragments.

[0048] The academic index information is a set of quantitative data extracted from the academic profile of the job applicant, which can objectively reflect their academic level and research achievements, including the number of top journal papers, fund project level, paper citation frequency, etc. After standardization and weighting, this information can be used as an important reference index to measure the academic achievements and research ability of the candidate, providing quantitative support for the matching model in job screening and evaluation.

[0049] S504: Input the job research direction vector, research direction vector, and academic index information into the quantitative matching model to calculate the job matching degree of the job applicant and the job: wherein, M represents the job matching degree, represents the weight represented by the cosine similarity, represents the cosine similarity function, represents the research direction vector of the job applicant, represents the job research direction vector, represents the weight represented by the score of the academic hard index of the job applicant, represents the scoring function of the academic hard index of the job applicant, represents the academic index of the job applicant, represents the weight represented by the matching bonus of the specific keywords of the job, represents the matching bonus function of the specific keywords of the job, represents the soft requirement keywords.

[0050] Specifically, according to the job requirements of the recruitment unit, a job portrait is defined in the system. The portrait contains the core research direction keywords required by the job, the minimum achievement requirements (such as h-index, the number of top journal papers), and other information. The system also uses the vectorization method in the previous step to convert the research direction keywords required by the job into an n-dimensional job research direction vector. In order to solve the subjectivity problem of matching degree evaluation, the following quantitative matching degree calculation model is designed to quantify the matching degree.

[0051] In the embodiment of the application, by using the quantitative matching model to calculate the matching degree of the candidate's academic portrait and the job portrait, not only can the research direction similarity, academic indicators and job keyword matching and other multi-dimensional factors be integrated to generate an objective and comparable job matching degree score, thereby effectively avoiding the subjectivity and uncertainty in the manual screening process, but also the weights can be flexibly adjusted according to the different needs of the recruitment unit, and the individualized configuration of the matching strategy is realized. This method significantly improves the automation level and scientificity of job screening and decision-making, and ensures the accuracy and efficiency of the recruitment results.

[0052] S6: Determine whether the job matching degree is greater than the preset job matching degree. If yes, go to S7. Otherwise, terminate the process of the candidate.

[0053] Optionally, if the job matching degree is greater than the preset job matching degree, it indicates that the candidate is highly matched with the job and is the key object of investigation. The process will automatically proceed to the next step and start personalized communication. This judgment step is the key link between calculation and communication, which ensures that the system only initiates communication to the most suitable candidates, and realizes the effective use of resources.

[0054] It should be noted that the size of the preset job matching degree can be set by the person skilled in the art according to actual needs, and the present application does not limit it.

[0055] S7: Fuse the candidate's academic portrait and the pre-extracted static information of the university to build a knowledge graph.

[0056] Among them, the static information of the university refers to the set of basic information related to colleges and universities or scientific research institutions in the recruitment scene, which is relatively fixed and long-term effective. Such information usually does not change with the change of individual candidates, and mainly includes the research direction of the department of the university, the composition of the teaching staff, the standard of scientific research starting fund, the policy of housing allowance, the process of children's enrollment, housing and welfare policy, etc.

[0057] In one possible implementation, S7 specifically includes: S701: Identify entity information related to the school from the static information of the school using information extraction technology. The entity information includes the research direction of the department, the teaching staff, the research starting fund standard, the relocation allowance policy, and the children's enrollment process.

[0058] S702: Identify the relationship between entities from the static information of the school based on information extraction technology. The relationship includes the relationship between the candidate and the organization, the relationship between the candidate and the paper, the relationship between the candidate and the fund project, the relationship between the candidate and the award, the relationship between the candidate and the keyword, the relationship between the candidate and the collaborator, the relationship between the paper and the organization, and the relationship between the keyword.

[0059] It should be noted that information extraction technology is a core technology in natural language processing, which is used to automatically identify and extract structured information from unstructured or semi-structured text. Its main goal is to convert raw natural language text into "entities" and "relationships" that can be understood and processed by computers. Information extraction usually includes two categories: one is named entity recognition (NER), which is used to identify entities such as people, organizations, papers, funds, awards, and keywords in the text; the other is relationship extraction (RE), which is used to identify the semantic relationship between these entities, such as "scholar - works for - organization" and "scholar - publishes - paper". Through information extraction technology, complex natural language descriptions can be converted into standardized knowledge units, providing structured data support for subsequent knowledge graph construction, intelligent question answering, and semantic retrieval.

[0060] S703: Map the structured data in the candidate's academic portrait to graph entities and relationship nodes.

[0061] S704: Based on the unified graph pattern, fuse the graph entities and relationship nodes with the entities and relationships between entities in the static information of the school, establish a graph structure containing school entities and candidate entities, and form a knowledge graph.

[0062] Specifically, the system uses information extraction technology (NER and relationship extraction RE) to process the public information and internal recruitment materials of the school. Entity and relationship definition: extract key "entities" (such as the research direction of each department, the teaching staff, the research starting fund standard, the relocation allowance policy, and the children's enrollment process) and their "relationships" from the materials.

[0063] The relationship types include: candidate-[graduated from]->institution, candidate-[employed at]->institution, candidate-[published]->paper, candidate-[hosted|participated]->fund project, candidate-[obtained]->award, candidate-[research field is]->keyword, paper-[belongs to]->institution (the institution published the paper), paper-[owns keyword]->keyword, candidate-[cooperated with]->candidate (the relationship is established through co-published papers). The graph is generated: the data of the structured portrait previously stored in the database is mapped to the pattern designed in the first step.

[0064] In the embodiment of the application, through the constructed knowledge graph, not only the academic portrait of the candidate and the static information of the institution can be fused to realize unified management and standardized expression of information, but also the entity and its relationship can be clearly presented in the form of a graph structure, thereby supporting semantic retrieval, path query and logical reasoning. The knowledge graph provides a high-quality data basis for subsequent post communication, candidate question and answer and automatic decision-making, and significantly improves the intelligent level and response efficiency of the recruitment process.

[0065] S8: generating a greeting for the candidate based on the knowledge graph and the language model, and receiving a natural language question of the candidate.

[0066] In a possible implementation, S8 specifically includes: S801: extracting post information, institution policy and candidate academic portrait information from the knowledge graph as input data for generating the greeting.

[0067] S802: inputting the input data into the language model to generate a personalized greeting and sending the greeting to the candidate, and receiving a natural language question of the candidate.

[0068] Specifically, the pre-defined template matching: the system pre-designs a query template for each intent. The identified entity: candidate=Zhang San, institution=Tsinghua University If the result is queried, it means that the relationship exists The system fills in the entity into the template to generate the final query.

[0069] In the embodiment of the application, a personalized greeting conforming to the background of the candidate is generated through the knowledge graph, automatic and efficient recruitment communication is realized, and the candidate can interact with the system through a natural language question. This method not only improves the experience and response rate of the candidate, but also establishes a complete closed loop from personalized communication to automatic question and answer, and significantly improves the intelligent level and actual effect of the recruitment process.

[0070] S9: analyzing the natural language question of the candidate to generate a knowledge graph query statement.

[0071] In a possible implementation, S9 specifically includes: S901: performing named entity recognition on the natural language question to extract key entities including candidates, institutions, papers, fund projects, awards and research keywords.

[0072] S902: performing intent classification on the natural language question based on the key entities to determine the question type.

[0073] It should be noted that the question type includes education background query, cooperation relationship query, research achievement query or post policy query, etc.

[0074] In the embodiment of the application, the natural language question of the candidate is converted into a knowledge graph query statement, realizing a no-threshold and automatic semantic analysis and question processing. This method not only improves the convenience of the interaction between the candidate and the system, but also ensures the accuracy and efficiency of the subsequent query and answer, thereby significantly enhancing the intelligent level of the recruitment communication link.

[0075] S10: executing the knowledge graph query statement to obtain corresponding structured answer data, and based on the structured answer data, generating a natural language reply through a language model and feeding back to the candidate.

[0076] In a possible implementation, S10 specifically includes: S1001: executing the knowledge graph query statement in the knowledge graph to obtain structured answer data matched with the knowledge graph query statement.

[0077] S1002: determining a query result based on the structured answer data, wherein the query result includes node, path and subgraph information.

[0078] S1003: inputting the structured answer data into the language model to generate a corresponding natural language reply, and feeding back the natural language reply to the candidate.

[0079] In the embodiment of the application, the structured data returned by the knowledge graph query is automatically converted into a natural language answer and fed back to the candidate, realizing a real-time and automatic closed loop of the candidate's question and the system's answer. This method not only improves the intuitiveness of information expression and the naturalness of interaction, but also ensures the accuracy and reliability of the answer content, thereby significantly improving the intelligent degree of the recruitment communication link and the candidate experience.

[0080] The technical scheme provided by the embodiment of the application has at least the following beneficial effects: In the embodiment of the present application, by automatically analyzing the resume of the job seeker, extracting academic entity information and determining the research direction vector, and then constructing the academic portrait and calculating the job matching degree, the complicated manual review is converted into a standardized and quantitative calculation process, thereby significantly improving the screening efficiency and reducing subjectivity, realizing efficient and fair evaluation; at the same time, the system can also fuse the academic portrait and the college information to construct the knowledge graph, automatically generate personalized greetings, and realize instant interaction and natural language reply based on the knowledge graph question and answer, thereby improving the initiative of recruitment and effectively avoiding the loss of high-quality talents.

[0081] With reference to the accompanying drawings Figure 2 The accompanying drawings show a structural schematic diagram of an intelligent communication system based on a university recruitment scene provided by the embodiment of the present application.

[0082] The embodiment of the present application provides an intelligent communication system 20 based on a university recruitment scene, which comprises a processor 201 and a memory 202. The memory 202 stores programs or instructions that can run on the processor 201, and the programs or instructions are executed by the processor 201 to realize the steps of the intelligent communication method based on the university recruitment scene described above, and can achieve the same technical effect. To avoid repetition, the present application will not be described again.

[0083] It should be understood that the processor 201 in the embodiment of the present application can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), ready-to-program gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0084] It is also to be understood that the memory 202 in embodiments of the present application can be volatile or nonvolatile memory, or can include both volatile and nonvolatile memory. Nonvolatile memory can be read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically EPROM (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), used as external cache. By way of example, and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0085] The above-described embodiments can be implemented in whole or in part by software, hardware (e.g., circuitry), firmware, or any combination of the three. When implemented in software, the above-described embodiments can be implemented in the form of one or more computer programs that are stored in a computer-readable storage medium. The computer-readable storage medium can be loaded into a computer, and the computer can execute the computer program to wholly or partially produce the processes or functions described in the embodiments of the present application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. The computer program can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer program can be transferred from one website, computer, server, or data center to another website, computer, server, or data center through a wired (e.g., infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that includes one or a collection of available media. The available media can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0086] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0087] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those of ordinary skill in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0088] Those of ordinary skill in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the devices, apparatuses and units described above can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0089] In several embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other manners. For example, the embodiments of the apparatus described above are merely schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation. For example, a plurality of units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0090] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0091] In addition, each functional unit in the various embodiments of the present application can be integrated into a processing unit, or each unit can be a physically separate unit, or two or more units can be integrated into a unit.

[0092] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0093] The embodiment of the present application provides a readable storage medium, which includes: a program or instruction stored on the readable storage medium, the program or instruction is executed by a processor to realize the steps of the intelligent communication method based on the university recruitment scene, and the same technical effect can be achieved. To avoid repetition, the present application will not be described again.

[0094] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present application, but not to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. Any changes or replacements that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application.

Claims

1. An intelligent communication method based on a university recruitment scene, characterized in that, Comprise: S1: obtaining an online resume submitted by a job applicant; S2: parsing the online resume through a document parsing library to extract a set of original unordered texts; S3: based on the set of original unordered texts, extracting academic entity information with academic meaning through a pre-trained language model, and determining an academic research direction vector of the job applicant based on the academic entity information; S4: based on the academic entity information and the academic research direction vector, constructing an academic portrait of the job applicant; S5: based on the academic portrait of the job applicant, calculating the job matching degree of the job applicant and the job position through a quantitative matching model; S6: determining whether the job matching degree is greater than a preset job matching degree; if yes, go to S7; Otherwise, terminate the process of the job applicant; S7: fusing the academic portrait of the job applicant and the pre-extracted static information of the university to construct a knowledge graph; S8: based on the knowledge graph and the language model, generating a greeting for the job applicant, and receiving a natural language question from the job applicant; S9: parsing the natural language question of the job applicant to generate a knowledge graph query statement; S10: executing the knowledge graph query statement to obtain corresponding structured answer data, and based on the structured answer data, generating a natural language reply through the language model and feeding back to the job applicant. 2.The intelligent communication method based on the university recruitment scene according to claim 1, characterized in that, The S2 specifically comprises: S201: receiving the online resume submitted by the job applicant and the related resume attachment file; S202: calling the document parsing library to parse the online resume and the attachment file by section and word by word; S203: based on the parsing result, extracting all the text content of the online resume to form the set of original unordered texts. 3.The intelligent communication method based on the university recruitment scene according to claim 1, characterized in that, The S3 specifically comprises: S301: through the language model, identifying and extracting academic entity information containing paper list, research keywords, scientific research fund, cooperative job applicant, education background and academic awards from the set of original unordered texts; S302: concatenating the research keywords and the paper titles and abstracts in the paper list into a coherent string, and inputting the coherent string into the word segmentation processing module of the language model to obtain a token sequence; S303: regarding the paper list as a vector set, wherein each paper in the paper list is concatenated by the title and the abstract: wherein, represents the first i text vector representation unit of the paper, represents the title text of the first i paper, represents the abstract text of the first i paper, represents a paper list set; S304: performing a pooling operation on the token sequence to generate a query vector representing the core research direction; S305: taking the query vector as a query, the paper list vector set as keys and values, and calculating the attention weight of the query vector to each paper through an attention mechanism: wherein, denotes the i attention weight of the query vector, denotes an activation function, denotes the query vector, denotes the i vector representation of the paper, denotes the vector dimension for scaling the inner product result; S306: based on the attention weight, aggregating all the paper vectors into a context-aware text vector through weighted summation: wherein, denotes the text vector, n denotes the number of papers, denotes the attention weight of the i th paper, denotes the vector of the i th paper; S307: fusing the text vector and the query vector to obtain a fusion vector, and outputting the fusion vector as the academic research direction vector of the job applicant. 4.The intelligent communication method based on the high school recruitment scene according to claim 3, characterized in that, The S4 specifically comprises: S401: assigning a unique identifier for the candidate for associating the academic entity information with the academic research direction vector; S402: combining the academic entity information and the academic research direction vector to form a composite data structure; S403: based on the unique identifier, establishing a data storage model including a main model and an association model; S404: mapping the composite data structure to the data storage model to generate the candidate academic portrait. 5.The intelligent communication method based on the university recruitment scene according to claim 1, characterized in that, The S5 specifically includes: S501: defining a post portrait based on the post demand of the recruitment unit, wherein the post portrait includes core research direction keywords, achievement requirements and post soft demand; S502: converting the core research direction keywords in the post portrait into a post research direction vector through vectorization method; S503: extracting research direction vector and academic index information from the candidate academic portrait; S504: inputting the post research direction vector, the research direction vector and the academic index information into the quantitative matching model to calculate the post matching degree of the candidate and the recruitment post: wherein, M represents a post matching degree, represents a weight represented by a cosine similarity, represents a cosine similarity function, represents a research direction vector of a candidate, represents a post research direction vector, represents a weight represented by a score of an academic hard index of a candidate, represents a score function of an academic hard index of a candidate, represents an academic index of a candidate, represents a weight represented by a matching bonus of a post specific keyword, represents a matching bonus function of a post specific keyword, represents a soft requirement keyword. 6.The intelligent communication method based on the high school recruitment scene according to claim 5, characterized in that, The S7 specifically includes: S701: using information extraction technology to identify entity information related to colleges and universities from the college static information, the entity information including department research direction, faculty team, research starting fund standard, relocation allowance policy and children enrollment process; S702: based on the information extraction technology, identifying the relationship between entities from the college static information, the relationship including candidate and organization, candidate and paper, candidate and fund project, candidate and award, candidate and keyword, candidate and collaborator, paper and organization paper and keyword; S703: mapping the structured data in the candidate academic portrait to graph entity and relationship node; S704: based on a unified graph pattern, fusing the graph entity and the relationship node with the entities and the relationships between entities in the college static information to establish a graph structure containing college entities and candidate entities, forming the knowledge graph. 7.The intelligent communication method based on the university recruitment scene according to claim 1, characterized in that, The S8 specifically includes: S801: extracting post information, college policy and candidate academic portrait information from the knowledge graph as input data for generating a greeting; S802: inputting the input data into the language model to generate personalized greetings and sending them to the candidate, and receiving natural language questions from the candidate. 8.The intelligent communication method based on the university recruitment scene according to claim 1, characterized in that, The S9 specifically includes: S901: performing named entity recognition on the natural language question to extract key entities including candidate, organization, paper, fund project, award and research keyword; S902: based on the key entities, classifying the natural language question to determine the question type; S903: based on the question type, generating the knowledge graph query statement according to the predefined query template. 9.The intelligent communication method based on the university recruitment scene according to claim 1, characterized in that, The S10 specifically includes: S1001: executing the knowledge graph query statement in the knowledge graph to obtain structured answer data matching the knowledge graph query statement; S1002: determining a query result based on the structured answer data, wherein the query result comprises node, path and subgraph information; S1003: inputting the structured answer data into the language model to generate a corresponding natural language reply, and feeding back the natural language reply to the job seeker.

10. An intelligent communication system based on a university recruitment scenario, characterized in that, Comprise: a processor and a memory; the memory stores programs or instructions executable on the processor, and the programs or instructions are executed by the processor to realize the steps of the intelligent communication method based on the university recruitment scene as claimed in any one of claims 1 to 9.

Images