Intelligent evaluation method for resume matching degree

By obtaining and processing job seekers' resume information, using the BERT model and BiLSTM+CRF model to generate knowledge graph tags, it solves the problem of building and verifying the recruitment resume knowledge graph, realizes efficient recruitment business support and data management, and improves the enterprise's Competitiveness and recruitment efficiency.

CN118964620AInactive Publication Date: 2024-11-15HUBEI ZHENDAO DIGITAL INTELLIGENT TECH CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411172556.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-26
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

It is difficult for existing technology to quickly build, verify and apply the knowledge graph of recruitment resumes, affecting the competitiveness and recruitment efficiency of enterprises in the capital market.

Method used

By obtaining the resume information of the job seeker, after preprocessing, inputting the classification model to filter valid text information, using entity recognition model and word segmentation processing to generate tags, and performing similarity matching, building knowledge graph tags, using BERT model and BiLSTM+CRF model for Deep learning, combined with ClickHouse database for data processing and filtering.

Benefits of technology

It improves the accuracy and efficiency of text information processing, builds a high-quality knowledge graph software system, improves the matching recommendation and data usage efficiency of recruitment services, reduces data coupling, and provides diversified business guarantees.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964620A_ABST
    Figure CN118964620A_ABST
Patent Text Reader

Abstract

The invention discloses an intelligent resume matching degree evaluation method, which belongs to the technical field of natural language processing, and specifically comprises the following steps of: acquiring data to obtain resume information; text processing: pre-processing the resume information and inputting the pre-processed resume information into a classification model to screen effective text information; entity identification: inputting the text information into an entity identification model to obtain a first label; text word segmentation: performing word segmentation processing on the text information to obtain a second label; performing label optimization: performing similarity matching on the first labels and the second labels, screening the second labels of which the similarity with any first label is greater than a preset threshold value, marking the second labels as third labels, outputting the first labels and the third labels together, and marking the first labels and the third labels as knowledge graph labels; according to the method, the texts are identified based on the resume information, the labels are output, and the knowledge graph label output accuracy is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular to a resume matching intelligent evaluation method. Background Art

[0002] In the context of digital reform, the strength of data application capabilities increasingly reflects the competitiveness of enterprises in the capital market. The application of knowledge graphs is an indispensable part. The application of knowledge graphs is reflected in many aspects, such as: helping enterprises to accurately find the talents they need; helping enterprises to locate the positions they need; helping job seekers find their ideal job opportunities; and providing job seekers with career development roadmaps, etc. The application of knowledge graphs is of obvious benefit to enterprises whose main business is online recruitment.

[0003] In the increasingly competitive recruitment industry, how to quickly build, verify, and apply knowledge graphs will help companies improve their core competitiveness, while also providing the necessary conditions for the diversified development of companies. Summary of the invention

[0004] The purpose of the present invention is to provide a resume matching intelligent evaluation method to solve the following technical problems: How to quickly build, verify, and apply recruitment resume knowledge graphs.

[0005] The purpose of the present invention can be achieved by the following technical solutions: A resume matching intelligent evaluation method includes the following steps: Obtain resume information of job seekers; The resume information is pre-processed and then input into a classification model to filter out valid text information; Inputting the text information into an entity recognition model to obtain a first label; Performing word segmentation processing on the text information to obtain a second label; The first tag is matched with the second tag by similarity, and the second tag whose similarity with any first tag is greater than a preset threshold is selected and marked as a third tag. The first tag and the third tag are output together and marked as a knowledge graph tag.

[0006] As a further solution of the present invention: the text processing module preprocesses as follows: The text information is encoded and sorted by position embedding and then sent to the classification model.

[0007] As a further solution of the present invention: the classification model of the text processing module adopts the BERT model, and the BERT model is based on a multi-layer Transformer architecture.

[0008] As a further solution of the present invention: the entity recognition model of the entity recognition module includes the following steps: Input the text information into the BERT model and map it into a word vector, input the word vector into the BiLSTM layer, perform deep learning on the word vector context feature information through the BiLSTM layer, output the probability of each text information corresponding to a label, generate a label sequence, input the label sequence into the CRF layer for sorting, obtain the label sequence that optimizes the objective function, and mark it as the first label sequence; The text content is segmented based on a conventional dictionary, the text content is split into several words, the words are labeled, the labeled words are converted into vectors and then input into a text classification model. The text classification model is a hybrid model based on BERT, BiLSTM and CRF. The context feature information of any word vector is deeply learned, the probability of each text content corresponding to an entity is output, and an entity sequence is generated. The entity sequence is input into the CRF layer for sorting, and the entity sequence that optimizes the objective function is obtained. The entity sequence is screened based on a preset project dictionary, and valid entities related to the project dictionary are screened, and the text content including the valid entities is marked as valid text.

[0009] As a further solution of the present invention: the text segmentation module comprises the following steps: The text information is processed by NLP word segmentation to obtain a word segmentation result, and the word segmentation result is matched with a dictionary tool and filtered by position to obtain a second label.

[0010] As a further solution of the present invention: the server receives the request and splices it into an SQL statement that can be directly executed in Click House according to the conditions selected by the user, and executes the SQL in ClickHouse and generates a new Bitmap.

[0011] As a further solution of the present invention: the resume information collected by the data collection module includes the following steps: job seeker resume text and recruiter position text.

[0012] As a further solution of the present invention: the classification model screens the resume information into normal, simple and meaningless, and outputs the normal information and simple information as valid text information.

[0013] Beneficial effects of the present invention: The present invention performs entity recognition and word segmentation on text information respectively, matches the labels after the two processes, and thus screens out erroneous features existing in entity recognition, thereby improving the accuracy of model output. Based on the present invention, label screening is performed on text information in the recruitment field, and a knowledge graph software system is constructed and applied to various aspects of recruitment business to provide high-quality services such as matching recommendations in recruitment business and guiding users to write resumes / position information. At the same time, based on the construction of the knowledge graph software system, the link between data verification and development will be clearer, data coupling will be reduced, data utilization efficiency will be improved, and guarantee services will be provided for business diversification. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below in conjunction with the accompanying drawings.

[0015] Figure 1 It is a schematic diagram of the process of the present invention; DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] See also Figure 1 As shown, the present invention is a resume matching intelligent evaluation method, comprising the following steps: Obtain resume information of job seekers; The resume information is pre-processed and then input into a classification model to filter out valid text information; Inputting the text information into an entity recognition model to obtain a first label; Performing word segmentation processing on the text information to obtain a second label; Perform similarity matching between the first tag and the second tag, select a second tag whose similarity with any first tag is greater than a preset threshold, mark it as a third tag, and output the first tag and the third tag together and mark them as a knowledge graph tag; Based on the present invention, text information in the recruitment field is labeled and screened, a knowledge graph software system is constructed and applied to various aspects of the recruitment business to provide high-quality services such as matching recommendations in the recruitment business and guiding users to write resumes / position information. At the same time, based on the construction of the knowledge graph software system, the link between data verification and development will be clearer, data coupling will be reduced, data utilization efficiency will be improved, and guarantee services will be provided for business diversification.

[0018] In the field of recruitment, the processing and use of text information has always been an important topic. With the development of technology, based on the text tag screening technology of the present invention, we can build an efficient knowledge graph software system to provide strong support for all aspects of the recruitment business.

[0019] First, through label screening technology, we can quickly and accurately classify and organize a large amount of recruitment text information. This text information may come from various channels, such as recruitment websites, social media, corporate websites, etc. By screening and classifying this information, we can form a comprehensive and systematic knowledge map to provide strong data support for subsequent recruitment business.

[0020] In the recruitment business, matching recommendation is a very important link. Based on the well-built knowledge graph software system, we can recommend the most suitable positions for candidates based on their skills, experience, educational background and other information. At the same time, for enterprises, the system can also screen out the most suitable candidates according to their recruitment needs, improving the efficiency and accuracy of recruitment.

[0021] In addition, the knowledge graph software system can also help users write resumes and job information. By analyzing a large amount of data, the system can provide users with some suggestions and templates to help them better showcase their strengths and characteristics. At the same time, the system can also parse and classify job information to help companies and candidates better understand and match job information.

[0022] In addition to the functions mentioned above, the construction of the knowledge graph software system can also make the link of data verification development clearer. In the traditional recruitment process, data verification is a complex and time-consuming link. By building a knowledge graph, we can organically integrate various data, reduce data coupling, and improve data usage efficiency. This can not only improve the efficiency of recruitment business, but also provide guarantee services for the diversification of business.

[0023] In short, the knowledge graph software system built based on the text tag screening technology of the present invention can provide comprehensive support for recruitment business. From matching recommendations to resume writing, from data verification to business diversification, this system can provide powerful assistance for all aspects of recruitment business. In the future, with the continuous advancement of technology and the continuous expansion of application scenarios, we have reason to believe that the knowledge graph software system will play a more important role in the recruitment field.

[0024] In a preferred embodiment of the present invention, the text processing module preprocesses as follows: The text information is encoded and sorted by position embedding and then sent to the classification model; The reason for setting positional embedding is that when the vectors in the text pass through the Encoder / Decoder stack of the Transformer at the same time, the model itself has no sense of position / order for each vector, but for text, there is an order between a paragraph of text, so a position embedding is added to each vector. Position embedding uses a trainable parameter and is directly superimposed on the vector. The size of the position embedding is the same as the text vector.

[0025] In another preferred embodiment of the present invention, the classification model of the text processing module adopts a BERT model, and the BERT model is based on a multi-layer Transformer architecture; The BERT model adopts a multi-layer Transformer architecture, a tutor classifier is set behind the Transformer stack, and a student classifier is connected behind each layer of the Transformer architecture of the BERT model.

[0026] In another preferred embodiment of the present invention, the entity recognition model of the entity recognition module comprises the following steps: Input the text information into the BERT model and map it into a word vector, input the word vector into the BiLSTM layer, perform deep learning on the word vector context feature information through the BiLSTM layer, output the probability of each text information corresponding to a label, generate a label sequence, input the label sequence into the CRF layer for sorting, obtain the label sequence that optimizes the objective function, and mark it as the first label sequence; Since the output of BiLSTM is the score of each label of the unit, we can select the one with the highest score as the label of the unit. But we cannot guarantee that the label is predicted correctly every time, which is the so-called label bias problem. Therefore, by setting the CRF layer to reorder, this is the function of CRF (conditional random field), adding some constraint rules to greatly reduce the probability of prediction errors.

[0027] The text content is segmented based on a conventional dictionary, the text content is split into several words, the words are labeled, the labeled words are converted into vectors and then input into a text classification model. The text classification model is a hybrid model based on BERT, BiLSTM and CRF. The context feature information of any word vector is deeply learned, the probability of each text content corresponding to an entity is output, and an entity sequence is generated. The entity sequence is input into the CRF layer for sorting, and the entity sequence that optimizes the objective function is obtained. The entity sequence is screened based on a preset project dictionary, and valid entities related to the project dictionary are screened, and the text content including the valid entities is marked as valid text.

[0028] In another preferred embodiment of the present invention, the text segmentation module comprises the following steps: Processing the text information through NLP word segmentation to obtain a word segmentation result, matching the word segmentation result with a dictionary tool and filtering by position to obtain a second label; The natural language processing NLP of the present invention is based on a bidirectional maximum matching method, which compares the word segmentation result obtained by the forward maximum matching method with the result obtained by the reverse maximum matching method, and then selects the result with the least word segmentation according to the maximum matching principle.

[0029] In another preferred embodiment of the present invention, the server receives the request and splices it into an SQL statement that can be directly executed in Click House according to the conditions selected by the user, and executes the SQL in ClickHouse and generates a new Bitmap.

[0030] In another preferred embodiment of the present invention, the resume information collected by the data collection module includes the following steps: job seeker resume text and recruiter position text.

[0031] In another preferred embodiment of the present invention, the classification model screens the resume information into normal, simple and meaningless, and outputs the normal information and simple information as valid text information; The main application is the combination of BERT pre-trained language model + RNN model, which divides the job content into modules such as job responsibilities, job requirements, and company benefits; and splits the resume work experience into personal responsibilities and project responsibilities.

[0032] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", etc. should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0033] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A resume matching intelligent evaluation method, characterized in that: The following steps are involved: Obtain resume information of job seekers; The resume information is pre-processed and then input into a classification model to filter out valid text information; Inputting the text information into an entity recognition model to obtain a first label; Performing word segmentation processing on the text information to obtain a second label; The first tag is matched with the second tag by similarity, and the second tag whose similarity with any first tag is greater than a preset threshold is selected and marked as a third tag. The first tag and the third tag are output together and marked as a knowledge graph tag.

2. According to claim 1, a resume matching intelligent evaluation method is characterized in that: The pre-processing is: The text information is encoded and sorted by position embedding and then sent to the classification model.

3. The resume matching intelligent evaluation method according to claim 1, characterized in that: The classification model adopts the BERT model, and the BERT model is based on a multi-layer Transformer architecture.

4. The resume matching intelligent evaluation method according to claim 1, characterized in that: The entity recognition model includes the following steps: Input the text information into the BERT model and map it into a word vector, input the word vector into the BiLSTM layer, perform deep learning on the word vector context feature information through the BiLSTM layer, output the probability of each text information corresponding to a label, generate a label sequence, input the label sequence into the CRF layer for sorting, obtain the label sequence that optimizes the objective function, and mark it as the first label sequence; The text content is segmented based on a conventional dictionary, the text content is split into several words, the words are labeled, the labeled words are converted into vectors and then input into a text classification model. The text classification model is a hybrid model based on BERT, BiLSTM and CRF. The context feature information of any word vector is deeply learned, the probability of each text content corresponding to an entity is output, and an entity sequence is generated. The entity sequence is input into the CRF layer for sorting, and the entity sequence that optimizes the objective function is obtained. The entity sequence is screened based on a preset project dictionary, and valid entities related to the project dictionary are screened, and the text content including the valid entities is marked as valid text.

5. The resume matching intelligent evaluation method according to claim 1, characterized in that: The text segmentation comprises the following steps: The text information is processed by NLP word segmentation to obtain a word segmentation result, and the word segmentation result is matched with a dictionary tool and filtered by position to obtain a second label.

6. The resume matching intelligent evaluation method according to claim 1, characterized in that: The resume information collected by the data includes the following steps: job seeker resume text and recruiter position text.

7. The sparse matrix-based label storage system according to claim 6, characterized in that: The server receives the request and splices it into an SQL statement that can be directly executed in Click House according to the conditions selected by the user, executes the SQL in ClickHouse and generates a new Bitmap.

8. The resume matching intelligent evaluation method according to claim 1, characterized in that: The classification model screens the resume information into normal, simple and meaningless, and outputs the normal information and simple information as valid text information.

Citation Information

Patent Citations

  • Automatic resume searching method, automatic recruitment system and computer storage medium

    CN113434687A

  • Sequence labeling method based on deep learning and application

    CN114692568A

  • Label storage system based on sparse matrix

    CN116401333A

  • Clue business opportunity insight system based on NLP text processing technology

    CN116523555A

  • Labeling method and system for recruitment field knowledge graph construction

    CN116541528A