A non-standard resume person-job matching method, device, computer equipment and readable storage medium based on artificial intelligence and knowledge graph

Abstract

The present invention discloses a method, device, computer equipment and readable storage medium for matching non-standard resumes and positions based on artificial intelligence and knowledge graphs, including: first obtaining a non-standard resume to be processed, extracting keywords and key phrases through word segmentation, obtaining target position description information, performing semantic association based on the industry knowledge graph to obtain resume and position association data, inputting the data into a pre-trained model to obtain multi-dimensional matching results, and finally generating a person-job matching report, thereby achieving efficient and accurate non-standard resume-person-job matching.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method, device, computer equipment and readable storage medium for matching non-standard resumes to positions based on artificial intelligence and knowledge graphs. Background Art

[0002] In today's recruitment field, processing a large number of non-standard resumes and achieving accurate job matching is a challenging task. Traditional resume screening methods are often inefficient and have limited accuracy, and are unable to cope with the diversity of resume formats and content. Summary of the invention

[0003] The purpose of the present invention is to provide a method, device, computer equipment and readable storage medium for matching non-standard resumes to positions based on artificial intelligence and knowledge graphs.

[0004] In a first aspect, an embodiment of the present invention provides a non-standard resume-job matching method based on artificial intelligence and knowledge graph, comprising:

[0005] Get non-standard resumes to be processed;

[0006] Segmenting the non-standard resume to be processed and extracting keywords and key phrases;

[0007] Obtain job description information of target positions;

[0008] Based on the pre-built industry knowledge graph, the keywords, the key phrases and the job description information are semantically associated to obtain resume-related data corresponding to the non-standard resume to be processed and job-related data corresponding to the target job, respectively;

[0009] Inputting the resume-related data and the position-related data into a pre-trained person-position matching model to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position;

[0010] A person-job matching report for the non-standard resume to be processed is generated based on the multi-dimensional matching results.

[0011] In a possible implementation, obtaining the non-standard resume to be processed includes:

[0012] Get original resumes;

[0013] The original resume is formatted and text cleaned to obtain the non-standard resume to be processed.

[0014] In a possible implementation, the word segmentation of the non-standard resume to be processed and the extraction of keywords and key phrases include:

[0015] Calling a word segmentation tool to segment the non-standard resume to be processed, and obtaining the word frequency and inverse document frequency of each word segmentation;

[0016] Calling the word vector model to convert each of the word segments into a word vector, and filtering out the key vectors in the recruitment scenario based on the separator built based on the support vector machine;

[0017] The keywords and the key phrases are determined from the multiple word segments according to the word frequency, the inverse document frequency and the key vector.

[0018] In a possible implementation, the industry knowledge graph is constructed in the following manner, including:

[0019] Obtain open source data in the field of human resources, industry research reports, and internal job description data;

[0020] Based on the open source data in the human resources field, the industry research report and the internal job description data of the enterprise, the industry knowledge graph including multiple entities and multiple relationships is constructed through automatic extraction and annotation operations; the industry knowledge graph is updated at a preset period.

[0021] In a possible implementation, the inputting of the resume-related data and the position-related data into a pre-trained person-position matching model to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position includes:

[0022] The resume-related data and the position-related data are standardized, and the standardized resume-related data and position-related data are input into a pre-trained person-job matching model based on a multi-layer perceptron and a convolutional neural network to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position; the convolutional layer and the pooling layer of the convolutional neural network are used to extract local features of the text, and the multi-layer perceptron is used to perform high-level feature fusion and prediction.

[0023] In a possible implementation, generating a person-job matching report for the non-standard resume to be processed according to the multi-dimensional matching result includes:

[0024] Generating a skill matching report, an experience relevance report, and a potential job performance prediction report for the non-standard resume to be processed according to the multi-dimensional matching results;

[0025] The skill matching report, the experience relevance report and the potential work performance prediction report are used as the person-job matching report of the non-standard resume to be processed.

[0026] In a possible implementation, the method further includes:

[0027] Collect case data on successful and failed person-job matching, including resume data and corresponding job information of successfully hired employees, as well as feedback data on resumes that were not successfully matched;

[0028] The person-job matching model is iteratively trained using the above case data to adjust the model parameters of the person-job matching model.

[0029] In a second aspect, an embodiment of the present invention provides a non-standard resume person-job matching device based on artificial intelligence and knowledge graph, comprising:

[0030] An acquisition module is used to acquire the non-standard resume to be processed; segment the non-standard resume to be processed and extract keywords and key phrases; acquire the job description information of the target job; semantically associate the keywords, key phrases and job description information based on a pre-built industry knowledge graph to obtain resume-related data corresponding to the non-standard resume to be processed and job-related data corresponding to the target job;

[0031] The matching module is used to input the resume-related data and the position-related data into a pre-trained person-position matching model to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position; and generate a person-position matching report for the non-standard resume to be processed based on the multi-dimensional matching result.

[0032] In a third aspect, an embodiment of the present invention provides a computer device, comprising a processor and a non-volatile memory storing computer instructions, wherein when the computer instructions are executed by the processor, the computer device executes the method described in at least one possible implementation of the first aspect.

[0033] In a fourth aspect, an embodiment of the present invention provides a readable storage medium, wherein the readable storage medium includes a computer program, and when the computer program is run, the computer device where the readable storage medium is located is controlled to execute the method described in at least one possible implementation manner of the first aspect.

[0034] Compared with the prior art, the beneficial effects provided by the present invention include: adopting a non-standard resume-job matching method, device, computer equipment and readable storage medium based on artificial intelligence and knowledge graph disclosed in the present invention, by obtaining the non-standard resume to be processed, extracting keywords and key phrases through word segmentation, obtaining the target job description information, performing semantic association based on the industry knowledge graph to obtain resume and job association data, inputting it into a pre-trained model to obtain multi-dimensional matching results, and finally generating a job matching report, thereby achieving efficient and accurate non-standard resume-job matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can also be obtained based on these drawings without creative work.

[0036] Figure 1 A schematic diagram of the steps of a non-standard resume-person-job matching method based on artificial intelligence and knowledge graph provided in an embodiment of the present invention;

[0037] Figure 2 A schematic block diagram of the structure of a non-standard resume-person-job matching device based on artificial intelligence and knowledge graph provided in an embodiment of the present invention;

[0038] Figure 3 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions 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 part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0040] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0041] In order to solve the technical problems in the aforementioned background technology, Figure 1 The flowchart of the non-standard resume-job matching method based on artificial intelligence and knowledge graph provided in the embodiment of the present disclosure is as follows. The non-standard resume-job matching method based on artificial intelligence and knowledge graph is introduced in detail.

[0042] Step S201, obtaining a non-standard resume to be processed;

[0043] Step S202, segmenting the non-standard resume to be processed and extracting keywords and key phrases;

[0044] Step S203, obtaining job description information of the target job;

[0045] Step S204, semantically associating the keywords, the key phrases, and the job description information based on the pre-constructed industry knowledge graph, and obtaining resume-associated data corresponding to the non-standard resume to be processed and job-associated data corresponding to the target job, respectively;

[0046] Step S205, inputting the resume-related data and the position-related data into a pre-trained person-position matching model to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position;

[0047] Step S206, generating a person-job matching report for the non-standard resume to be processed based on the multi-dimensional matching results.

[0048] In the embodiment of the present invention, for example, in a recruitment scenario, the server receives a large number of resumes, which have different formats and contents, some of which are free text forms, and some of which contain special graphics or multimedia elements, and are non-standard resumes.

[0049] For example, the server receives a resume submitted in PDF format. In this resume, the job seeker uses a unique layout and paragraph division to describe his work experience and skills. First, the server will parse and convert this PDF resume to extract the text content.

[0050] For example, another resume is uploaded in image format. The server uses image recognition technology to identify the text information in the image and convert it into processable text.

[0051] After extracting the text content, the server will also perform text cleaning operations. It will remove some meaningless characters, such as extra spaces, special symbols, etc. At the same time, for some spelling errors or non-standard expressions, the server will use natural language processing technology to correct and standardize them to ensure the accuracy of subsequent processing.

[0052] Suppose the server receives a non-standard resume text that has been cleaned and converted: "In the past five years, I have been working in the field of big data analysis. I am proficient in Python programming and Hadoop distributed computing framework. I have participated in many large-scale data mining projects and effectively improved the efficiency and accuracy of data processing."

[0053] The server first calls the word segmentation tool to segment the text. The word segmentation tool will segment the text into individual words, for example: "I" have been" committed to" the" field of" big data analysis" in" the" past" five" years" and" have" been" proficient in" Python" programming and" Hadoop" distributed" computing" framework" and" participated in" multiple" large" data mining" projects" and" effectively" improved" the" efficiency" and" accuracy" of" data processing".

[0054] Then, the server will count the term frequency (TF, TermFrequency) and inverse document frequency (IDF, InverseDocumentFrequency) of each word segment. Term frequency refers to the number of times a word appears in this resume. For example, "big data analysis" appears once, and "Python" appears once. Inverse document frequency is an indicator to measure the general importance of a word. If a word appears in many documents, then its inverse document frequency is low; conversely, if a word rarely appears in other documents but appears in the current document, then its inverse document frequency is high.

[0055] Next, the server calls the word vector model to convert each word segment into a word vector. Word vectors are a way to represent words as numerical vectors so that computers can process and calculate them. Suppose "big data analysis" is converted into a vector [0.1, 0.2, 0.3] and "Python" is converted into a vector [0.4, 0.5, 0.6].

[0056] The separator built based on the support vector machine will filter these word vectors to find out the key vectors that are important in the recruitment scenario. For example, in this example, word vectors related to technology and fields such as "big data analysis", "Python", and "Hadoop" may be filtered as key vectors.

[0057] Finally, the server considers the word frequency, inverse document frequency and key vectors to determine the keywords and key phrases. In this example, the keywords may include "big data analysis", "Python" and "Hadoop", and the key phrases may be "data mining project" and "efficiency and accuracy of data processing".

[0058] Assume that the target position received by the server is "Big Data Development Engineer", and its job description information is as follows:

[0059] "job description:

[0060] 1. Responsible for designing, developing and maintaining big data processing systems, using technologies in the Hadoop ecosystem, such as Hive, Spark, etc., to perform data extraction, transformation and loading (ETL) work.

[0061] 2. Use Python for data analysis and modeling to support business decision making.

[0062] 3. Participate in the performance optimization of the big data platform to solve the problems of high concurrency and low latency in data processing.

[0063] 4. Collaborate with team members to complete project delivery and ensure system stability and reliability.

[0064] Job requirements:

[0065] 1.Bachelor degree or above, major in computer science, statistics or related fields.

[0066] 2. Have more than three years of experience in big data development and be familiar with the Hadoop ecosystem.

[0067] 3. Proficient in Python programming language and have good data analysis and modeling capabilities.

[0068] 4. Have good teamwork spirit and communication skills, and be able to withstand work pressure. "

[0069] The server will receive and store this job description information in full and prepare for subsequent processing.

[0070] An industry knowledge graph covering the field of big data is pre-built in the server. This knowledge graph contains various entities, such as technologies, tools, projects, positions, etc., as well as the relationships between them.

[0071] Take the previously extracted resume keywords "big data analysis", "Python", "Hadoop" and the key phrase "data mining project" as an example, as well as the description information of the target position. The server will search for entities and relationships related to these keywords and key phrases in the knowledge graph.

[0072] For example, "big data analysis" is associated with entities such as "data analysis methods" and "data processing procedures" in the knowledge graph; "Python" is related to entities such as "programming language" and "data analysis tools"; "Hadoop" is connected to entities such as "big data processing framework" and "distributed computing".

[0073] For "Hadoop ecosystem" and "Python programming" in job description information, the corresponding related entities and relationships can also be found in the knowledge graph.

[0074] In this way, the server can generate corresponding related data for resumes and job descriptions. For resumes, the related data may include information such as technical depth and application scenarios related to keywords and key phrases; for jobs, the related data may cover the complexity and application frequency of the technology required for the job.

[0075] The server has pre-trained a person-job matching model based on a multi-layer perceptron and a convolutional neural network.

[0076] First, the server will standardize the resume-related data and job-related data and convert them into an input format that the model can accept.

[0077] Assume that the resume-related data after standardized processing is represented as a vector [0.2, 0.3, 0.5, 0.1, 0.4], and the position-related data is represented as a vector [0.3, 0.4, 0.2, 0.2, 0.5].

[0078] These standardized vectors are input into the person-job matching model. The convolutional layer and pooling layer of the convolutional neural network extract local features of the text, such as the matching features between the proficiency of a certain technology in the resume and the importance of the technology in the job requirements. The multi-layer perceptron is responsible for high-level feature fusion and prediction.

[0079] After calculation and processing by the model, a multi-dimensional matching result is finally obtained, which may include scores of multiple dimensions such as skill matching, experience matching, potential adaptability, etc.

[0080] For example, the skill match score is 0.8 (out of 1), the experience match score is 0.7, and the potential adaptability score is 0.6.

[0081] Based on the above multi-dimensional matching results, the server generates a detailed person-job matching report.

[0082] In terms of skill matching, the report will indicate the degree to which the job seeker's skills in "big data analysis", "Python programming", "Hadoop application" and other skills match the job requirements. For example, it means that the job seeker's skills in Python programming are highly matched with the job requirements, and he or she can skillfully use relevant libraries and frameworks for data analysis and processing.

[0083] In terms of experience relevance, the report may mention the relevance of the "data mining projects" that the job seeker has participated in and the project experience required for the target position. For example, the experience that the job seeker has accumulated in processing large-scale data and solving complex problems in past projects is more consistent with the project experience requirements for big data development engineers in the target position.

[0084] For the prediction of potential job performance, the report may predict the job applicant's potential job performance in the new position based on the assessment of the applicant's learning ability, adaptability, etc. For example, by comprehensively considering the applicant's past performance in learning and applying new technologies, it is inferred that the applicant has good learning and adaptability potential when facing new challenges and new technologies in the position.

[0085] Ultimately, this job matching report will provide recruiters with a comprehensive, detailed and objective reference, helping them to more accurately assess the suitability of job seekers for target positions.

[0086] In the embodiment of the present invention, the acquisition of the non-standard resume to be processed can be implemented through the following example.

[0087] Get original resumes;

[0088] The original resume is formatted and text cleaned to obtain the non-standard resume to be processed.

[0089] In the embodiment of the present invention, for example, in a busy recruitment season, the server has to process a large number of original resumes from different channels every day. These original resumes are in various formats, some are common Word documents, some are HTML web page formats, and some may be scanned images or handwritten PDF documents.

[0090] For example, the server first receives an original resume submitted in Word format. This resume contains a lot of formatting, such as different fonts, font sizes, colors, and various charts and bullet points. The server will use a special document parsing tool to extract the plain text content and remove the format-related codes and tags, leaving only the text information.

[0091] Next, the server receives a resume in HTML format. This format may contain a large number of web page tags and style codes. The server uses the HTML parser to accurately identify and remove these irrelevant codes and extract the main text, such as the applicant's name, work experience, educational background and other key information.

[0092] Another original resume was submitted as a scanned image. The server uses advanced optical character recognition (OCR) technology to convert the text in the image into editable text. In this process, some recognition errors may be encountered due to poor scanning quality and blurred handwriting. The server will correct and optimize the recognized text through context analysis and comparison with the common vocabulary library.

[0093] In addition, a handwritten PDF resume was also sent to the server, which also used OCR technology to recognize handwritten text and optimized it based on handwriting characteristics and common writing habits to improve recognition accuracy.

[0094] After completing the above preliminary processing of various raw resume formats, the server enters the text cleaning phase. For example, a text resume extracted from a web page may contain a large number of meaningless separators such as spaces, line breaks, and tabs. The server will uniformly process these redundant separators to make them more standardized and concise.

[0095] For example, some resumes may contain some spelling errors or irregular abbreviations. The server corrects these errors through built-in dictionaries and language models. For example, "professional" is corrected to "professional" and "devloper" is corrected to "developer".

[0096] At the same time, the server will also optimize some unclear or overly colloquial content. For example, the vague expression "did some work related to data analysis" was optimized to "participated in multiple data analysis projects" to make it more accurate and professional.

[0097] After a series of complex and sophisticated operations such as format conversion and text cleaning, the server finally obtained a clear, standardized, and easy-to-process non-standard resume, which is fully prepared for subsequent steps such as word segmentation and keyword extraction.

[0098] In the embodiment of the present invention, the word segmentation of the non-standard resume to be processed and the extraction of keywords and key phrases can be implemented through the following examples.

[0099] Calling a word segmentation tool to segment the non-standard resume to be processed, and obtaining the word frequency and inverse document frequency of each word segmentation;

[0100] Calling the word vector model to convert each of the word segments into a word vector, and filtering out the key vectors in the recruitment scenario based on the separator built based on the support vector machine;

[0101] The keywords and the key phrases are determined from the multiple word segments according to the word frequency, the inverse document frequency and the key vector.

[0102] In the embodiment of the present invention, the server receives, for example, a non-standard resume to be processed after format conversion and text cleaning, the content of which is as follows: "I have been engaged in software development for many years in a well-known Internet company, and I am proficient in programming languages ​​such as Java and Python. I have participated in the development of multiple large projects and have good teamwork and problem-solving skills."

[0103] First, the server calls the word segmentation tool to segment the resume. The word segmentation tool divides this text into individual words, and obtains "I" have been "engaged" in "software" development for "a" well-known "Internet" company for "many years" and "proficient" in "Java" and "Python" and "etc." programming "languages" and "participated in" "multiple" large-scale" projects and "possess" "good" "team" "collaboration" and "problem" "solving" abilities.

[0104] Then, the server starts to count the word frequency and inverse document frequency of each word segment. In this example, the word "development" appears 3 times, and its word frequency is relatively high. The word "well-known" appears less frequently and has a lower word frequency. The calculation of inverse document frequency requires reference to a large number of documents stored in the server. Assuming that the word "development" appears frequently in many resumes and related texts, its inverse document frequency is relatively low; while words like "Python" that are relatively professional and unique in a specific field will have a higher inverse document frequency.

[0105] Next, the server calls the word vector model to convert each word segment into a word vector. For example, "development" may be converted to a vector like [0.2, 0.3, 0.5], and "Python" may be converted to [0.6, 0.7, 0.8].

[0106] The separator built based on the support vector machine begins to work and screens out the key vectors in the recruitment scenario. In this process, the separator will determine which word vectors are more important for evaluating the match between job seekers and positions based on the characteristics and needs of the recruitment field. For example, since this is a resume related to software development, word vectors such as "Java", "Python", and "software development" may be identified as key vectors because they are directly related to technical capabilities.

[0107] Finally, the server comprehensively considers word frequency, inverse document frequency, and key vectors to determine keywords and key phrases. Although "development" has a high word frequency, it may not be determined as a keyword because of its low inverse document frequency and the fact that it is not a key vector. On the other hand, "Python" may not have a high word frequency, but it has a high inverse document frequency and is a key vector, so it will be determined as a keyword. "Software development" may be determined as a key phrase because it performs well in the comprehensive evaluation of word frequency, inverse document frequency, and key vectors.

[0108] Let's take another example. Suppose the server receives a non-standard resume to be processed: "I have accumulated extensive experience in the field of marketing, and am good at brand promotion and market research. I have successfully planned and executed multiple marketing activities, effectively increasing brand awareness and market share."

[0109] After word segmentation, we get “I”, “in”, “in”, “marketing”, “field”, “accumulated”, “rich”, “experience”, “good at”, “brand”, “promotion”, “and”, “market”, “research”, “once”, “successfully”, “planning”, “and”, “execute”, “multiple”, “marketing”, “activities”, “effectively”, “enhanced”, “brand”, “popularity”, “and”, “market”, “share”.

[0110] The word frequency statistics show that the word "market" appears many times and has a high word frequency. The inverse document frequency calculation shows that the more professional word "marketing" has a high inverse document frequency.

[0111] After word embedding conversion, “brand promotion” may be converted to [0.4, 0.5, 0.6], and “market research” may be converted to [0.7, 0.8, 0.9].

[0112] Through the screening of the separator constructed by the support vector machine, word vectors such as “brand promotion” and “market research” are considered as key vectors.

[0113] After comprehensive consideration, the keywords are determined to be "marketing", "brand promotion" and "market research", and the key phrase may be "effectively increase brand awareness and market share."

[0114] Through the above series of rigorous and meticulous processing, the server can accurately extract valuable keywords and key phrases from complex non-standard resumes, providing key basic data for subsequent job matching work.

[0115] In an embodiment of the present invention, the industry knowledge graph is constructed in the following manner and can be implemented through the following examples.

[0116] Obtain open source data in the field of human resources, industry research reports, and internal job description data;

[0117] Based on the open source data in the human resources field, the industry research report and the internal job description data of the enterprise, the industry knowledge graph including multiple entities and multiple relationships is constructed through automatic extraction and annotation operations; the industry knowledge graph is updated at a preset period.

[0118] In an embodiment of the present invention, exemplarily, in the process of constructing an industry knowledge graph, the server first obtains the required data from multiple channels.

[0119] The server obtains open source data in the field of human resources through web crawlers and data interfaces. These open source data may come from some professional human resources websites, open source databases, or data sets publicly shared by research institutions. For example, data such as the average salary range, common skill requirements, and career development paths for different industries and positions are obtained from a well-known human resources open source database.

[0120] At the same time, the server will also collect various industry research reports. These reports may come from professional market research companies, industry associations or consulting agencies. For example, a research report on the latest development trends of the Internet industry was obtained, which detailed the impact of emerging technologies on Internet job demand, including new positions and new skills required driven by technologies such as artificial intelligence and big data.

[0121] In addition, the server will also obtain job description data from the internal system of the partner company. This data describes in detail the specific responsibilities, qualifications, work processes and collaborative relationships of each position within the company. For example, detailed descriptions of software development, product managers, test engineers and other positions were obtained from a large technology company.

[0122] With these rich data sources, the server begins to build industry knowledge graphs through automatic extraction and labeling operations.

[0123] When processing open source data in the field of human resources, the server uses natural language processing technology and machine learning algorithms to automatically extract key information. For example, if information such as "software engineers usually need to master programming languages ​​such as Java and Python" is extracted from open source data, the server will mark "software engineer" as an entity, "Java" and "Python" as related skill entities, and "master" as the relationship between them.

[0124] For industry research reports, the server analyzes the text content in the report and extracts key statements such as "With the development of artificial intelligence technology, the demand for data scientists continues to grow, and they need to have knowledge of mathematical modeling and machine learning algorithms." Then, "data scientist" is annotated as an entity, "mathematical modeling" and "machine learning algorithms" are annotated as skill entities, and "demand growth" and "need to have" are annotated as relationships.

[0125] The server will process the internal job description data of the enterprise more carefully. For example, for a job description of a product manager, which mentions "the product manager is responsible for coordinating the R&D, marketing and sales teams, promoting the iteration and upgrade of products, and needs to have good communication and project management capabilities", the server will mark "product manager" as an entity, "R&D team", "marketing team" and "sales team" as related collaborative entities, "coordinate", "promote" and "possess" as relationships, and "communication ability" and "project management ability" as required capability entities.

[0126] Through such automatic extraction and labeling operations, the server builds an initial version of the industry knowledge graph that includes multiple entities (such as various positions, skills, capabilities, etc.) and multiple relationships (such as needs, mastery, possession, collaboration, etc.).

[0127] However, industry knowledge is constantly developing and changing. In order to ensure the timeliness and accuracy of the knowledge graph, the server will update it at a preset period.

[0128] Every month, the server will re-acquire the latest open source data and industry research reports in the field of human resources. For example, it is found in the newly acquired open source data that due to the widespread application of a certain emerging technology, a position that was not very popular has become in high demand, and the required skills have also been updated. The server will add new entities and relationships to the knowledge graph accordingly, or modify the attributes and relationships of the original entities.

[0129] For the internal job description data of an enterprise, the server may synchronize and update the data with the partner enterprise every six months. If the responsibilities and requirements of certain positions are modified due to business adjustments or strategic changes within the enterprise, the server will promptly reflect these changes in the knowledge graph.

[0130] Suppose that during an update, the server finds that due to changes in the market environment, a certain industry has greatly increased its emphasis on "user experience design", and the demand and skill requirements for related positions have new standards. The server will add the entity "user experience design" to the knowledge graph and update the relationships and attributes of related positions and skills.

[0131] By continuously acquiring new data, automatically extracting and labeling, and periodically updating, the industry knowledge graph built and maintained by the server can always accurately reflect the latest developments and needs in the human resources field, providing a solid and reliable knowledge foundation for matching people with non-standard resumes.

[0132] In an embodiment of the present invention, the resume-related data and the position-related data are input into a pre-trained person-position matching model to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position, which can be implemented through the following example.

[0133] The resume-related data and the position-related data are standardized, and the standardized resume-related data and position-related data are input into a pre-trained person-job matching model based on a multi-layer perceptron and a convolutional neural network to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position; the convolutional layer and the pooling layer of the convolutional neural network are used to extract local features of the text, and the multi-layer perceptron is used to perform high-level feature fusion and prediction.

[0134] In the embodiment of the present invention, illustratively, the server receives resume-related data corresponding to the non-standard resume to be processed and position-related data corresponding to the target position, and prepares to perform person-position matching calculations.

[0135] First, the server normalizes the data. Assume that the resume-related data contains the job seeker's skill level, such as Python skill proficiency of 8 points (out of 10 points) and project management experience of 3 years; the job-related data requires Python skill proficiency of at least 7 points and project management experience of at least 2 years. The server will standardize these data with different dimensions and ranges, for example, converting skill proficiency into values ​​between 0 and 1, and converting years of experience into relative proportions.

[0136] After the standardization process is completed, the server inputs this data into a pre-trained person-job matching model based on a multi-layer perceptron and convolutional neural network.

[0137] In this model, the convolutional layer of the convolutional neural network starts working. It is like a keen observer, focusing on extracting subtle features of the local text. For example, for the local text described in the resume "using Python to solve complex data processing problems in the project", the convolutional layer will capture key information such as "Python", "data processing", "complex problems", and extract the potential patterns and relationships between them.

[0138] The pooling layer then compresses and filters the features extracted by the convolutional layer, removing some redundant and unimportant information and retaining only the most critical and representative local features.

[0139] At the same time, the multi-layer perceptron is also busy working. It receives the features extracted and filtered by the convolutional neural network, as well as other numerical data after standardization, such as years of experience, skill scores, etc.

[0140] The multi-layer perceptron begins to perform high-level feature fusion. It integrates features and numerical data from different parts to form a comprehensive feature representation. For example, it integrates Python skill features, project management experience features, and other related features into a comprehensive feature vector that can fully reflect the match between job seekers and positions.

[0141] Next, the multi-layer perceptron makes predictions based on the fused features. Through complex calculations and weight adjustments, it gives the matching results between the job seeker and the target position in multiple dimensions.

[0142] Assume that in this example, the multi-dimensional matching results include skill matching, experience matching, personality matching, etc. In terms of skill matching, the server concludes that the match between the applicant's Python skills and the job requirements is 0.9 (full score 1), indicating a high match; in terms of experience matching, since the applicant's project management experience is slightly higher than the job requirements, the match is 0.8; in terms of personality matching, by analyzing the work style and teamwork experience described in the resume and comparing them with the personality traits required for the position, the match is 0.7.

[0143] In another scenario, the server receives a new resume-related data and job-related data. The resume mentions that the job seeker has strong innovation capabilities and has led multiple innovative projects; the job requirements emphasize the need for innovative thinking and pioneering spirit.

[0144] The convolutional and pooling layers of the convolutional neural network can accurately extract key local features such as "innovation ability", "leading", and "innovative projects". The multi-layer perceptron combines these features with other standardized data for prediction, and the matching degree of innovation ability is 0.85.

[0145] Through such precise and comprehensive calculation and analysis, the server can generate accurate, detailed and multi-dimensional matching results for each non-standard resume and target position to be processed, providing strong support and reference for recruitment decisions.

[0146] In the embodiment of the present invention, generating a person-job matching report for the non-standard resume to be processed according to the multi-dimensional matching result can be implemented through the following example.

[0147] Generating a skill matching report, an experience relevance report, and a potential job performance prediction report for the non-standard resume to be processed according to the multi-dimensional matching results;

[0148] The skill matching report, the experience relevance report and the potential work performance prediction report are used as the person-job matching report of the non-standard resume to be processed.

[0149] In the embodiment of the present invention, the server successfully obtains the multi-dimensional matching result of a non-standard resume to be processed and the target position. Assume that the multi-dimensional matching result shows that the skill matching degree is 80%, the experience correlation is 70%, and the potential job performance is predicted to be good.

[0150] First, the server generates a skill matching report based on the skill matching. It will analyze in detail the degree to which the job seeker's key skills match the requirements of the target position. For example, the target position requires proficiency in data analysis tools Excel and Python, and the job seeker's skill proficiency in Excel reaches 90% of the job requirement, while in Python it is 70%. The server will clearly point out these specific proportions and differences in the report, and explain which skills the job seeker performs well in and which skills need further improvement or training.

[0151] Next, the server generates an experience relevance report. It compares the applicant's past work experience with the target position. Suppose the applicant has worked in a similar industry, but the project size and complexity are slightly lower than the target position. The server will elaborate on the similarities and differences of this experience in the report, for example, pointing out that the applicant has some experience in dealing with similar business problems, but the projects involved in the target position may require higher coordination and management capabilities. At the same time, it will also mention in which aspects of the applicant's experience can bring value to the new position, and which aspects may need to be adapted and expanded in the new work environment.

[0152] The server then generates a report based on the potential job performance prediction. If the prediction results show that the job seeker has good learning and adaptability, and can quickly grow and cope with challenges in a new work environment, the server will provide relevant analysis in the report. For example, the job seeker has demonstrated the ability to quickly master new technologies and solve complex problems in past jobs, or can maintain a positive attitude and high work efficiency when facing work pressure.

[0153] Finally, the server integrates the skill matching report, experience relevance report, and potential job performance prediction report as a person-job matching report for the non-standard resume to be processed.

[0154] For example, for a job applicant applying for a marketing manager position, the skill match report may show that his skills in market research methods and marketing channel management are highly matched with the job requirements, but there is still room for improvement in the use of digital marketing tools; the experience relevance report may point out that the job applicant has rich experience in traditional marketing, but relatively little experience in emerging social media marketing; the potential job performance prediction report may predict that he will be able to effectively lead the team and promote marketing innovation in the new position based on his past team leadership experience and innovative thinking.

[0155] Such a comprehensive and detailed person-job matching report can provide recruiters with clear and accurate references, helping them to have a more comprehensive understanding of the compatibility between job seekers and target positions, thereby making more informed recruitment decisions.

[0156] In the embodiments of the present invention, the following implementation modes are also provided.

[0157] Collect case data on successful and failed person-job matching, including resume data and corresponding job information of successfully hired employees, as well as feedback data on resumes that were not successfully matched;

[0158] The person-job matching model is iteratively trained using the above case data to adjust the model parameters of the person-job matching model.

[0159] In the embodiment of the present invention, illustratively, after a period of recruitment work, the server starts to execute the task of collecting data on successful and failed cases of person-job matching.

[0160] For successful job-person matching cases, the server obtains the resume data of the successfully hired employees from the company's human resources system. For example, an employee who successfully joined the software development position has a resume that details his education background as a computer-related major, many years of Java development experience, participation in multiple large-scale projects, and familiarity with agile development processes. At the same time, the server also obtains the corresponding job information of the employee, including the specific responsibilities of the position, such as being responsible for the development and maintenance of a core system of the company, and being proficient in specific development frameworks and tools.

[0161] For cases where the person-job match fails, the server collects feedback data on the resumes that were not successfully matched. For example, although some job seekers' resumes seem to meet the job requirements in some aspects, they are found to have insufficient actual skills during the interview, or their communication and teamwork abilities do not meet the job expectations. These feedback information are recorded in detail as a valuable data resource.

[0162] The server organizes and categorizes the large amount of case data collected. It stores successful cases and failed cases in different database tables, and extracts and annotates key information in each case, such as the job seeker's key skills, work experience, personality traits, and specific requirements and characteristics of the position.

[0163] Next, the server uses these case data to iteratively train the person-job matching model. During the training process, the model will learn how to more accurately judge the degree of match based on the resume features in the successful cases and the matching patterns of the job requirements. At the same time, it will also learn lessons from the failed cases, analyze the factors that lead to mismatch, and adjust its own parameters and weights.

[0164] Suppose during a training session, the server found that the model’s weight distribution for certain specific skills was not reasonable, resulting in some job seekers who had key skills but were weak in other aspects being misjudged as mismatched. By analyzing successful cases, the server adjusted the weights of these skills in the model, making the model pay more attention to the evaluation of key skills.

[0165] For example, the server found that for some positions that require strong communication and teamwork skills, the model was not accurate enough in assessing the relevant abilities of job seekers. Therefore, through in-depth research on failed cases, the server introduced more effective feature extraction methods and evaluation indicators to improve the model's assessment accuracy of these abilities.

[0166] After continuous iterative training, the performance of the person-job matching model has gradually improved, and it can more accurately predict the degree of matching between resumes and positions, providing more reliable support for the company's recruitment work.

[0167] In order to more clearly describe the solution provided by the embodiment of the present invention, a more complete implementation method is provided below.

[0168] 1. Data preparation stage:

[0169] Collection: Collect a large number of non-standard resume samples, and widely collect non-standard resumes and corresponding job description information from various recruitment-related channels, such as corporate recruitment emails, online recruitment platform backends, etc. These non-standard resumes may have different formats (such as PDF, Word, HTML, etc.) and complex content structures, while job description information covers key content such as job responsibilities, skill requirements, education requirements, and work experience requirements.

[0170] Preprocessing: Format conversion: Convert the collected non-standard resumes from multiple formats into plain text format for subsequent processing. For example, use a specialized document parsing tool to extract text from a PDF document. Text cleaning: Remove special symbols (such as punctuation marks, tabs, and other characters that may interfere with information extraction), extra spaces, garbled characters, etc. in the text to make the text content more standardized and easier to process. This step can improve the accuracy of subsequent keyword extraction and analysis.

[0171] 2. Keyword and phrase extraction implementation:

[0172] Word segmentation: Use mature word segmentation tools (such as Jieba word segmentation, etc.) to perform word segmentation on the preprocessed resume text. Word segmentation is to divide the continuous text into individual words according to certain rules. For example, "having rich experience in data analysis" will be divided into words such as "having", "rich", "of", "data analysis", "experience", etc.

[0173] Statistical analysis and vector representation: Calculate the statistical information of each word, including word frequency (the number of times a word appears in a text) and inverse document frequency (used to measure the importance of a word. Words with high frequency but appearing in most documents are less important). At the same time, convert words into vector representations through word vector models (such as Word2Vec). Word vectors are a technology that projects words into low-dimensional vector space, making words with similar semantics closer in the vector space.

[0174] Keyword screening: Combined with machine learning classifiers (such as support vector machines), representative keywords and phrases are selected based on the vector representation of words, statistical information, and semantic features in the field of human resources. These keywords and phrases are a condensed representation of candidate information, covering key dimensions such as skills (such as "python programming" and "data analysis"), work experience (such as "3 years of project management experience"), and educational background (such as "undergraduate degree in computer science"). In addition, in order to cope with emerging vocabulary and professional terms, a dynamic vocabulary update mechanism is adopted to ensure that important information in the resume can be accurately captured.

[0175] 3. Knowledge graph application implementation:

[0176] Knowledge collection: The knowledge required to build the knowledge graph is obtained from multiple data sources, including professional human resources books, industry research reports, internal job descriptions, etc. These materials contain rich knowledge in the field of human resources, such as the relationship between different positions and skills, the correspondence between educational background and ability, etc.

[0177] Knowledge graph construction: The nodes (entities) and edges (relationships) of the knowledge graph are determined by combining manual annotation and automatic extraction. Manual annotation is mainly used to process some knowledge with clear semantic relationships but difficult to accurately obtain through automatic extraction, such as unique job skill requirements in certain specific industries. Automatic extraction uses natural language processing technology to automatically identify entities and relationships from large amounts of text data. For example, named entity recognition technology is used to identify that "software engineer" is a job entity, and "Java programming" is a skill entity, and relationship extraction technology is used to determine that there is a "job-skill requirement" relationship between them.

[0178] Knowledge graph update: In order to ensure the timeliness and accuracy of the knowledge graph, it needs to be updated regularly. With the development of the industry and changes in corporate needs, new knowledge continues to emerge. By introducing new industry knowledge and corporate feedback information, the entities and relationships in the knowledge graph are adjusted and supplemented. For example, when a new programming technology emerges and attracts attention in the recruitment market, it is added to the knowledge graph and linked to relevant positions.

[0179] Semantic association: Map the keywords and phrases extracted from the resume to the knowledge graph, and use the query and reasoning functions of the knowledge graph to obtain the semantic association and contextual understanding of this information. For example, when the keyword "data analysis" is mentioned in the resume, the knowledge graph can be associated with related tools (such as Excel, Python data analysis library, etc.), related industries (such as finance, e-commerce, etc.) and required capabilities (such as statistical knowledge, data mining capabilities, etc.), so as to have a more comprehensive understanding of the candidate's capabilities.

[0180] 4. AI model prediction implementation:

[0181] Model input preparation: Convert the experience information and job description information processed by the knowledge graph into vector form suitable for model input. These vectors contain rich semantic information and are a numerical representation of the original text after keyword extraction and knowledge graph association.

[0182] Model architecture and prediction: The AI ​​model is built based on an architecture that combines a multi-layer perceptron (MLP) and a convolutional neural network (CNN). The CNN part uses the advantages of its convolutional layer and pooling layer to effectively extract local features of text, and extracts features from the text information in the input vector, such as capturing the similarity between specific skill descriptions in a resume and job skill requirements. The MLP part is used to fuse and predict high-level features, further integrating the features extracted from CNN to calculate the match scores between candidates and jobs in various dimensions (such as skill matching, experience relevance, potential work performance, etc.).

[0183] Model training and optimization (during the training phase): Model data division: Use a large amount of historical recruitment data as the training set, and divide it into a training set and a validation set according to a certain ratio (such as 8:2). The training set is used to update the model parameters, and the validation set is used to evaluate the performance of the model during the training process to prevent overfitting. Loss function selection and parameter update: Select a suitable loss function (such as the mean square error loss function) to measure the difference between the model prediction results and the actual annotation results. During the training process, the prediction results are obtained by passing the input data into the model for forward propagation, and the loss value is calculated. Then, through the back-propagation algorithm, according to the gradient of the loss function to the model parameters, the optimization algorithm (such as the Adam optimizer) is used to update the model parameters. After multiple training rounds (epochs), the performance of the model on the validation set no longer improves or the preset number of training times is reached.

[0184] 5. Matching report generation implementation:

[0185] Report production basis: According to the matching score output by the AI ​​model, a detailed matching report is generated for each candidate. The matching score reflects the degree of matching between the candidate and the position in different dimensions and is an important basis for recruiters to evaluate candidates.

[0186] Report content: Skill matching: The report lists the matching of various skills in detail, such as "Python programming skill matching: 0.8", "Data analysis skill matching: 0.7", etc. In this way, recruiters can intuitively understand the degree of fit between the candidate's key skills and the job requirements. Experience relevance: Describes the match between the candidate's years of work, industry experience, project experience and job requirements. For example, "The candidate has 3 years of experience in data analysis in the e-commerce industry, which is highly matched with the industry experience required for the position", which helps recruiters determine whether the candidate has enough practical experience to be competent for the job. Potential job performance prediction: Based on the model's assessment of the candidate's comprehensive ability, a qualitative evaluation is given, such as "The candidate has high potential and is expected to be able to respond quickly to the job." This part of the content provides recruiters with a forward-looking reference for the candidate's future work performance.

[0187] Please refer to Figure 2 , Figure 2 A non-standard resume person-job matching device 110 based on artificial intelligence and knowledge graph provided in an embodiment of the present invention includes:

[0188] The acquisition module 1101 is used to acquire the non-standard resume to be processed; segment the non-standard resume to be processed and extract keywords and key phrases; acquire the job description information of the target job; semantically associate the keywords, the key phrases and the job description information based on the pre-built industry knowledge graph to obtain the resume association data corresponding to the non-standard resume to be processed and the job association data corresponding to the target job;

[0189] The matching module 1102 is used to input the resume-related data and the position-related data into a pre-trained person-position matching model to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position; and generate a person-position matching report for the non-standard resume to be processed based on the multi-dimensional matching result.

[0190] It should be noted that the implementation principle of the aforementioned non-standard resume person-job matching device 110 based on artificial intelligence and knowledge graph can refer to the implementation principle of the aforementioned non-standard resume person-job matching method based on artificial intelligence and knowledge graph, and will not be repeated here. It should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into a physical entity, or physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be fully implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the non-standard resume person-job matching device 110 based on artificial intelligence and knowledge graph can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the non-standard resume person-job matching device 110 based on artificial intelligence and knowledge graph. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in a processor element or an instruction in software form.

[0191] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more microprocessors (digital signal processors, DSP), or one or more field programmable gate arrays (FPGA), etc. For another example, when a module above is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0192] The embodiment of the present invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned non-standard resume person-job matching device 110 based on artificial intelligence and knowledge graph. Figure 3 As shown, Figure 3The computer device 100 is a block diagram of a structure of a computer device 100 provided in an embodiment of the present invention. The computer device 100 includes a non-standard resume person-job matching device 110 based on artificial intelligence and knowledge graph, a memory 111, a processor 112 and a communication unit 113.

[0193] In order to realize data transmission or interaction, the memory 111, the processor 112 and the communication unit 113 are electrically connected to each other directly or indirectly. For example, the electrical connection between these components can be realized through one or more communication buses or signal lines. The non-standard resume person-job matching device 110 based on artificial intelligence and knowledge graph includes at least one software function module that can be stored in the memory 111 in the form of software or firmware or solidified in the operating system (OS) of the computer device 100. The processor 112 is used to execute the non-standard resume person-job matching device 110 based on artificial intelligence and knowledge graph stored in the memory 111, such as the software function modules and computer programs included in the non-standard resume person-job matching device 110 based on artificial intelligence and knowledge graph.

[0194] An embodiment of the present invention provides a readable storage medium, which includes a computer program. When the computer program is running, it controls the computer device where the readable storage medium is located to execute the aforementioned non-standard resume-person-job matching device 110 based on artificial intelligence and knowledge graph.

[0195] For illustrative purposes, the foregoing description is made with reference to specific embodiments. However, the above illustrative discussion is not intended to be exhaustive or to limit the present disclosure to the precise form disclosed. Numerous modifications and variations are possible in accordance with the above teachings. These embodiments are selected and described in order to best illustrate the principles of the present disclosure and its practical application, so that those skilled in the art can best utilize the present disclosure and utilize various embodiments with different modifications to suit the intended specific application.

Claims

1. A non-standard resume-job matching method based on artificial intelligence and knowledge graph, characterized in that: include: Get non-standard resumes to be processed; Segmenting the non-standard resume to be processed and extracting keywords and key phrases; Obtain job description information of target positions; Based on the pre-built industry knowledge graph, the keywords, the key phrases and the job description information are semantically associated to obtain resume-related data corresponding to the non-standard resume to be processed and job-related data corresponding to the target job, respectively; Inputting the resume-related data and the position-related data into a pre-trained person-position matching model to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position; A person-job matching report for the non-standard resume to be processed is generated based on the multi-dimensional matching results.

2. The method according to claim 1, characterized in that: The obtaining of the non-standard resume to be processed includes: Get original resumes; The original resume is formatted and text cleaned to obtain the non-standard resume to be processed.

3. The method according to claim 1, characterized in that The word segmentation of the non-standard resume to be processed and the extraction of keywords and key phrases include: Calling a word segmentation tool to segment the non-standard resume to be processed, and obtaining the word frequency and inverse document frequency of each word segmentation; Calling the word vector model to convert each of the word segments into a word vector, and filtering out the key vectors in the recruitment scenario based on the separator built based on the support vector machine; The keywords and the key phrases are determined from the multiple word segments according to the word frequency, the inverse document frequency and the key vector.

4. The method according to claim 1, characterized in that: The industry knowledge graph is constructed in the following ways, including: Obtain open source data in the field of human resources, industry research reports, and internal job description data; Based on the open source data in the human resources field, the industry research report and the internal job description data of the enterprise, the industry knowledge graph including multiple entities and multiple relationships is constructed through automatic extraction and annotation operations; the industry knowledge graph is updated at a preset period.

5. The method according to claim 1, characterized in that The step of inputting the resume-related data and the position-related data into a pre-trained person-position matching model to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position includes: The resume-related data and the position-related data are standardized, and the standardized resume-related data and position-related data are input into a pre-trained person-job matching model based on a multi-layer perceptron and a convolutional neural network to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position; the convolutional layer and the pooling layer of the convolutional neural network are used to extract local features of the text, and the multi-layer perceptron is used to perform high-level feature fusion and prediction.

6. The method according to claim 1, characterized in that Generating a person-job matching report for the non-standard resume to be processed according to the multi-dimensional matching result includes: Generating a skill matching report, an experience relevance report, and a potential job performance prediction report for the non-standard resume to be processed according to the multi-dimensional matching results; The skill matching report, the experience relevance report and the potential work performance prediction report are used as the person-job matching report of the non-standard resume to be processed.

7. The method according to claim 1, characterized in that The method further comprises: Collect case data on successful and failed person-job matching, including resume data and corresponding job information of successfully hired employees, as well as feedback data on resumes that were not successfully matched; The person-job matching model is iteratively trained using the above case data to adjust the model parameters of the person-job matching model.

8. A non-standard resume-job matching device based on artificial intelligence and knowledge graph, characterized in that: include: The acquisition module is used to obtain the non-standard resumes to be processed; Segmenting the non-standard resume to be processed and extracting keywords and key phrases; Obtain job description information of target positions; Based on the pre-built industry knowledge graph, the keywords, the key phrases and the job description information are semantically associated to obtain resume-related data corresponding to the non-standard resume to be processed and job-related data corresponding to the target job, respectively; A matching module, used for inputting the resume-related data and the position-related data into a pre-trained person-position matching model to obtain a multi-dimensional matching result between the non-standard resume to be processed and the target position; A person-job matching report for the non-standard resume to be processed is generated based on the multi-dimensional matching results.

9. A computer device, characterized in that: The computer device comprises a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device executes the method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium includes a computer program, and when the computer program is executed, the computer device where the readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

Images