Deep learning-based resume information extraction method and system

Abstract

The invention belongs to the technical field of computers, and aims to provide a resume information extraction method and system based on deep learning. The method comprises the steps of obtaining a to-be-processed resume text, and preprocessing the to-be-processed resume text to obtain a preprocessed word segmentation sequence; inputting the word segmentation sequence into a pre-trained resume information extraction model to obtain an optimal tag sequence corresponding to the resume text to be processed; performing post-processing on the resume text to be processed according to the optimal tag sequence to obtain an enhanced field sequence corresponding to the resume text to be processed; and according to the enhanced field sequence, constructing and obtaining structured resume information corresponding to the resume text to be processed. According to the invention, the comprehensiveness and accuracy of resume information extraction can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and specifically relates to a resume information extraction method and system based on deep learning. Background Art

[0002] In the modern job search and recruitment process, resume information extraction technology is one of the key links. Resume information extraction aims to automatically identify and extract key information from unstructured resume text, such as name, contact information, education background, work experience and skills. This process has an important impact on the recruitment efficiency of enterprises and also provides data support for human resource management.

[0003] At present, resume information extraction technology mainly relies on natural language processing (NLP) and traditional machine learning methods. Specifically, existing resume information extraction methods are generally divided into the following steps: (1) Text preprocessing: segment the resume text, remove stop words and punctuation marks to reduce complexity and improve processing efficiency; (2) Feature extraction: extract candidate information from the text using keyword matching, rule methods or specific domain dictionaries, such as locating educational experience and work experience by identifying keywords such as "university" or "company"; (3) Classification and labeling: input the extracted features into machine learning models, such as support vector machines (SVM), random forests, etc., for classification and labeling, and identify the specific content of each field in the resume; (4) Information extraction: extract specific information based on the classification results for subsequent use.

[0004] However, in the process of using the prior art, the inventors found that the prior art has at least the following problems: The format and expression of resumes vary significantly. Job seekers may use a variety of typesetting styles or expressions to describe their experience and skills, and traditional rule matching or keyword recognition methods often cannot cover all situations. For example, different candidates may use different words to describe positions or company information, which greatly reduces the accuracy of information extraction. In addition, most traditional methods rely on manually defined features to extract information from resume content. These features are usually effective for specific resume formats or fields, but often perform poorly when faced with ever-changing resume styles; especially when encountering emerging positions, skills or atypical resume formats, they lack flexible feature generalization capabilities, resulting in a decrease in the accuracy of information extraction. In addition, existing methods mainly rely on specific keywords or structured features, while resume texts usually contain rich semantic information and may involve complex syntactic structures. For example, a job description may contain multiple relevant skills, responsibilities, and project details. This information is difficult to accurately extract by relying solely on feature matching, resulting in some key information being omitted or mis-extracted. Summary of the invention

[0005] The present invention aims to solve the above technical problems at least to a certain extent, and provides a resume information extraction method and system based on deep learning.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a resume information extraction method based on deep learning, comprising: Obtaining a resume text to be processed, and preprocessing the resume text to be processed to obtain a preprocessed word segmentation sequence; Input the word segmentation sequence into a pre-trained resume information extraction model to obtain the optimal label sequence corresponding to the resume text to be processed; wherein the resume information extraction model includes a BERT encoding layer, a BiLSTM layer and a CRF layer connected in sequence; the BERT encoding layer is used to perform text feature extraction on the word segmentation sequence to obtain a context feature representation vector sequence of the word segmentation sequence, wherein the context feature representation vector sequence includes context feature representation vectors corresponding to each word in the word segmentation sequence; the BiLSTM layer is used to perform feature sequence encoding processing on the context feature representation vector sequence to obtain a feature vector sequence of the context feature representation vector sequence, wherein the feature vector sequence includes feature vectors corresponding to each word in the word segmentation sequence; the CRF layer is used to perform sequence labeling on the feature vector sequence to obtain the optimal label sequence corresponding to the resume text to be processed; Post-processing the resume text to be processed according to the optimal label sequence to obtain an enhanced processed field sequence corresponding to the resume text to be processed; The structured resume information corresponding to the resume text to be processed is constructed based on the enhanced field sequence.

[0007] In a possible design, the resume text to be processed is preprocessed to obtain a preprocessed word segmentation sequence, including: Cleaning the resume text to be processed to obtain a denoised resume text; Performing word segmentation and sentence segmentation processing on the denoised resume text to obtain a sentence segmented resume text; Part-of-speech tagging and named entity recognition are performed on the sentence-segmented resume text to obtain a preprocessed word segmentation sequence.

[0008] In a possible design, the resume information extraction model is trained by multi-task learning.

[0009] In a possible design, the resume text to be processed is post-processed according to the optimal label sequence to obtain an enhanced processed field sequence corresponding to the resume text to be processed, including: Perform field association and integration processing on the resume text to be processed according to the optimal label sequence to obtain an integrated field sequence corresponding to the resume text to be processed; The integrated field sequence is subjected to knowledge graph enhancement processing to obtain an enhanced field sequence.

[0010] In a possible design, field association and integration processing is performed on the resume text to be processed according to the optimal tag sequence to obtain an integrated field sequence corresponding to the resume text to be processed, including: The tags in the optimal tag sequence are confirmed in sequence. If there are consecutive tags that are the same, the word segments corresponding to the tags in the resume text to be processed are merged into one field to obtain a field sequence after the field is merged. According to resume domain knowledge, the fields in the field sequence after the fields are merged are processed by predefined rules to obtain a field sequence after rule processing; Redundant information in the field sequence after the rule processing is removed to obtain an integrated field sequence.

[0011] In a possible design, the integrated field sequence is subjected to knowledge graph enhancement processing to obtain an enhanced field sequence, including: Build a knowledge graph in the resume field based on a preset public database; The completeness of the integrated field sequence is checked. If there is missing information in a field of the integrated field sequence, the entity node corresponding to the field in the knowledge graph is found through the entity linking method, and then the integrated field sequence is completed based on the entity node to obtain an enhanced field sequence.

[0012] In a second aspect, the present invention provides a resume information extraction system based on deep learning, which is used to implement a resume information extraction method based on deep learning as described in any one of the above; the resume information extraction system based on deep learning includes: A preprocessing module is used to obtain the resume text to be processed, and preprocess the resume text to be processed to obtain a preprocessed word segmentation sequence; A label sequence acquisition module is communicatively connected with the preprocessing module, and is used to input the word segmentation sequence into a pre-trained resume information extraction model to obtain an optimal label sequence corresponding to the resume text to be processed; wherein the resume information extraction model includes a BERT encoding layer, a BiLSTM layer, and a CRF layer connected in sequence; the BERT encoding layer is used to perform text feature extraction on the word segmentation sequence to obtain a context feature representation vector sequence of the word segmentation sequence, wherein the context feature representation vector sequence includes context feature representation vectors corresponding to each word in the word segmentation sequence; the BiLSTM layer is used to perform feature sequence encoding processing on the context feature representation vector sequence to obtain a feature vector sequence of the context feature representation vector sequence, wherein the feature vector sequence includes feature vectors corresponding to each word in the word segmentation sequence; the CRF layer is used to perform sequence annotation on the feature vector sequence to obtain an optimal label sequence corresponding to the resume text to be processed; A field sequence enhancement module, which is in communication with the label sequence acquisition module and is used to post-process the resume text to be processed according to the optimal label sequence to obtain an enhanced processed field sequence corresponding to the resume text to be processed; The structured processing module is in communication with the field sequence enhancement module and is used to construct structured resume information corresponding to the resume text to be processed based on the enhanced field sequence.

[0013] In a third aspect, the present invention provides an electronic device, comprising: a memory for storing computer program instructions; and, A processor is used to execute the computer program instructions to complete the operation of a resume information extraction method based on deep learning as described in any one of the above.

[0014] In a fourth aspect, the present invention provides a computer program product, comprising a computer program or instructions, which, when executed by a computer, implements a resume information extraction method based on deep learning as described in any one of the above.

[0015] In a fifth aspect, the present invention provides a computer-readable storage product having instructions stored thereon. When the instructions are run on a computer, a resume information extraction method based on deep learning as described in any one of the above is executed.

[0016] The beneficial effects of the present invention are: The present invention discloses a resume information extraction method and system based on deep learning, which can improve the comprehensiveness and accuracy of resume information extraction. Specifically, in the implementation process of the present invention, first, the resume text to be processed is obtained, and the resume text to be processed is preprocessed to obtain the preprocessed word segmentation sequence; then, the word segmentation sequence is input into a pre-trained resume information extraction model to obtain the optimal label sequence corresponding to the resume text to be processed; then, the resume text to be processed is post-processed according to the optimal label sequence to obtain the enhanced processed field sequence corresponding to the resume text to be processed; finally, the structured resume information corresponding to the resume text to be processed is constructed according to the enhanced processed field sequence. The resume information extraction model in the present invention can automatically learn features, has strong generalization ability, and can adapt to different resume formats; wherein, the BERT encoding layer can understand the context of the word segmentation sequence, thereby more accurately capturing the information in the resume, so that the resume information extraction model can better understand the semantics and adapt to different expressions; the BiLSTM layer and the CRF layer can enhance the resume information extraction model's ability to understand sentence structure, thereby more accurately extracting entities (such as name, company, position, etc.). Based on this, the present invention is not only suitable for resumes in standard formats, but also has high robustness and adaptability to diverse texts and multimodal information, making the extraction results more comprehensive and accurate, and can provide enterprises with more efficient and flexible resume analysis technology.

[0017] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of a resume information extraction method based on deep learning in Example 1; Figure 2 It is a module block diagram of a resume information extraction system based on deep learning in Example 2; Figure 3 This is a module block diagram of an electronic device in Example 3. DETAILED DESCRIPTION

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0020] Embodiment 1: This embodiment discloses a resume information extraction method based on deep learning, which can be executed by, but is not limited to, a computer device or a virtual machine with certain computing resources, such as a personal computer, a smart phone, a personal digital assistant, or a wearable device, or by a virtual machine.

[0021] like Figure 1 As shown, a resume information extraction method based on deep learning may include but is not limited to the following steps: S1. Obtain the resume text to be processed, and preprocess the resume text to obtain the preprocessed word segmentation sequence; it should be noted that in actual applications, the resume text to be processed has various formats and may contain unstructured content such as images and tables. Therefore, it is necessary to preprocess it before extracting information from it. Specifically, in this embodiment, the process of preprocessing the resume text to be processed includes but is not limited to: S101. Cleaning the resume text to be processed to remove noise data therein and retain the text, thereby obtaining a denoised resume text; the noise data therein may include extra spaces, special characters and table symbols in the text; S102. Perform word segmentation and sentence segmentation on the denoised resume text, and use sentence segmentation symbols such as periods or semicolons to sentence the text after word segmentation to obtain a sentence-separated resume text; S103. Perform part-of-speech tagging and named entity recognition on the sentence-segmented resume text to preliminarily mark the parts of speech of the text and identify basic entities such as date and place, obtain a preprocessed word segmentation sequence, and provide auxiliary information for subsequent processes.

[0022] It should be noted that in the process of preprocessing the resume text to be processed, if the initial resume text to be processed contains image data, the OCR (Optical Character Recognition) method is used to recognize the text content in the image and merge it with the text data; for example, the company logo in the image can be matched with the company name through image recognition and used as auxiliary information to enhance the recognition accuracy of company word segmentation. Based on this improved solution, the present embodiment has the ability to fuse multimodal information, and can effectively extract image information in the resume while processing the text.

[0023] S2. Input the word segmentation sequence into a pre-trained resume information extraction model to obtain the optimal label sequence corresponding to the resume text to be processed; wherein the resume information extraction model includes a BERT (Bidirectional Encoder Representation of Transformer, a pre-trained language representation model) encoding layer, a BiLSTM (Bidirectional Long Short-Term Memory Network) layer and a CRF (Conditional Random Field) layer connected in sequence; the BERT encoding layer is used to perform text feature extraction on the word segmentation sequence to obtain a context feature representation vector sequence of the word segmentation sequence, wherein the context feature representation vector sequence includes context feature representation vectors corresponding to each word in the word segmentation sequence. In this embodiment, the context feature representation vector sequence can be expressed as { h 1, h 2,……, h n}, the context feature representation vector in the context feature representation vector sequence is a high-dimensional vector containing rich context information; the BiLSTM layer is used to perform feature sequence encoding processing on the context feature representation vector sequence to obtain a feature vector sequence of the context feature representation vector sequence, wherein the feature vector sequence includes feature vectors corresponding to each word in the word segmentation sequence, and the feature vector sequence can be represented as { h 1 BiLSTM , h 2 BiLSTM ,……, h n BiLSTM}, including the feature representation of each participle in the context h 1 BiLSTM , h 2 BiLSTM ,……, h n BiLSTM In this embodiment, the BiLSTM layer can further capture the contextual features of each word segment, and any feature vector is obtained by concatenating the bidirectional LSTM (Long Short-Term Memory) feature vectors of the corresponding word segment; the CRF layer is used to perform sequence labeling on the feature vector sequence to obtain the optimal label sequence corresponding to the resume text to be processed; the optimal label sequence can be expressed as { y 1, y 2,……, y n}, including the labels of each word in the word sequence y 1, y 2,……, y n; It should be understood that the optimal label sequence includes multiple fields corresponding to the resume text to be processed and carrying labels that match it.

[0024] In this embodiment, the resume information extraction model is trained by multi-task learning. Specifically, in this embodiment, during the training process of the initial resume information extraction model, various resume fields (such as education and work experience) in the training data set are used as different subtasks, and a multi-task learning framework is used to define subtasks for each resume field. The initial resume information extraction model is trained by sharing parameters, and then the trained resume information extraction model is obtained. Based on this, the extraction effect of the overall model can be improved.

[0025] In addition, when training the initial resume information extraction model, due to the diversity of resume formats, the training data set is expanded in advance through data augmentation methods, or the initial resume information extraction model is pre-trained on large-scale corpus in other fields and then fine-tuned to the resume field in order to improve the generalization ability of the model.

[0026] It should also be understood that in the process of training the initial resume information extraction model, the cross entropy loss function is used to train the model, and the extraction effect is evaluated by precision, recall rate and F1 value.

[0027] Through the above method, the resume information extraction model trained in this embodiment can greatly improve the accuracy and generalization ability of resume information extraction, while reducing the dependence on manual rules and improving the robustness under different resume formats.

[0028] It should be noted that during the implementation of this embodiment, the resume information extraction model can gradually process and infer the input word segmentation sequence, and finally output specific field information, that is, the optimal label sequence, which can be helpful in constructing structured resume information.

[0029] S3. Post-process the resume text to be processed according to the optimal label sequence to obtain an enhanced processed field sequence corresponding to the resume text to be processed.

[0030] It should be noted that the optimal label sequence output by the resume information extraction model may contain redundant information or lack contextual semantics. For this reason, this embodiment further sets a post-processing step. Specifically, the post-processing step mainly further analyzes and enhances the optimal label sequence output by the model so as to finally output structured resume information.

[0031] Specifically, in step S3, the resume text to be processed is post-processed according to the optimal label sequence to obtain an enhanced processed field sequence corresponding to the resume text to be processed, including: S301. Perform field association and integration processing on the to-be-processed resume text according to the optimal tag sequence to obtain an integrated field sequence corresponding to the to-be-processed resume text. For example, merge multiple fields with consecutive tags of "Name" into a complete name, and identify multiple date fields as a complete work experience time period. It should be noted that in a resume, some fields often consist of multiple consecutive words or phrases, such as a complete name, company name, position, and skills. Therefore, after obtaining the optimal tag sequence based on the resume information extraction model, through the field association and integration processing step, it is beneficial to merge related fields in the optimal tag sequence into complete structured information.

[0032] Specifically, in step S301, performing field association and integration processing on the to-be-processed resume text according to the optimal tag sequence to obtain an integrated field sequence corresponding to the to-be-processed resume text includes: S3011. Sequentially confirm the tags in the optimal tag sequence. If there are consecutive identical tags (for example, multiple consecutive "Company" tags appear), then merge the word segments corresponding to this tag in the to-be-processed resume text into one field to obtain a field sequence after field merging; S3012. According to resume domain knowledge, process the fields in the field sequence after field merging through predefined rules to obtain a field sequence after rule processing. For example, merge consecutive date fields (such as "2021 - 2023") into a complete time period; S3013. Remove redundant information in the field sequence after rule processing to obtain an integrated field sequence. For example, if the tag contains multiple meaningless phrases (such as "of", "and", etc.), then filter to retain the core information in the field.

[0033] S302. Perform knowledge graph enhancement processing on the integrated field sequence to obtain an enhanced field sequence after processing. It should be noted that the knowledge graph can associate and complement resume information, which helps to make the finally extracted resume content more comprehensive. Specifically, in this embodiment, add association information to the fields identified from the integrated field sequence through a pre-constructed knowledge graph. For example, associate and complement the position field with corresponding skills, and associate and complement the company name with information such as the industry it belongs to, etc. Thus, the content that may be missing in the resume can be filled, and further enhance the integrity of the resume content corresponding to the tag sequence.

[0034] Specifically, in step S302, performing knowledge graph enhancement processing on the integrated field sequence to obtain an enhanced field sequence after processing includes: S3021. Construct a knowledge graph in the resume field based on a preset public database; specifically, the knowledge graph contains information such as positions, companies, industries, skills, etc.; for example, a node in the knowledge graph represents a company name, and an edge represents the industry to which the company belongs.

[0035] S3022. Perform a completeness check on the integrated field sequence. If there is missing information in a field of the integrated field sequence, find the entity node corresponding to the field in the knowledge graph through the entity linking method, and then complete the integrated field sequence based on the entity node to obtain an enhanced field sequence. For example, if the identified position is "data scientist" and the relevant skills are not mentioned in the resume, the skills (such as "Python", "machine learning") of the "data scientist" node in the knowledge graph are automatically associated as completion information. In addition, in this embodiment, when the entity node corresponding to the field in the knowledge graph is found through the entity linking method, for example, for the "software engineer" position field, the "software engineer" node in the knowledge graph is searched and associated with the relevant skill nodes (such as "programming language").

[0036] In this embodiment, the knowledge graph can be expressed as G =( V , E ),in, V is a collection of entity nodes, E is an edge set, given a field in the integrated field sequence F When there is a corresponding node in the knowledge graph v ∈ V , then complete the relevant information according to the adjacent nodes, that is, the completed field is F'=F { v' | (v,v')∈E},in, v' For the field F The associated node.

[0037] It should be noted that, in this embodiment, based on the above steps S301 to S302, by merging continuous fields in the label sequence, the integrity of the field content can be ensured; then, the identified resume fields are further associated with the knowledge graph to complete the relevant skills, industry and other information, which can improve the semantic understanding and information comprehensiveness of the field sequence, effectively solve the problem of missing content in traditional information extraction, and make the structured results of the resume information more complete and reliable.

[0038] S4. Constructing structured resume information corresponding to the resume text to be processed according to the enhanced processed field sequence. Specifically, the structured resume information includes complete field information such as name, company, position, skills and education experience.

[0039] In addition, this embodiment should also be able to utilize user feedback to achieve self-update of the model through an incremental learning mechanism to improve adaptability in long-term use.

[0040] This embodiment can improve the comprehensiveness and accuracy of resume information extraction. Specifically, during the implementation of this embodiment, first, the resume text to be processed is obtained, and the resume text to be processed is preprocessed to obtain the preprocessed word segmentation sequence; then, the word segmentation sequence is input into the pre-trained resume information extraction model to obtain the optimal label sequence corresponding to the resume text to be processed; then, the resume text to be processed is post-processed according to the optimal label sequence to obtain the enhanced processed field sequence corresponding to the resume text to be processed; finally, the structured resume information corresponding to the resume text to be processed is constructed according to the enhanced processed field sequence. The resume information extraction model in this embodiment can automatically learn features, has strong generalization ability, and can adapt to different resume formats; wherein, the BERT encoding layer can understand the context of the word segmentation sequence, so as to more accurately capture the information in the resume, so that the resume information extraction model can better understand the semantics and adapt to different expressions; the BiLSTM layer and the CRF layer can enhance the resume information extraction model's ability to understand sentence structure, so as to more accurately extract entities (such as name, company, position, etc.). Based on this, this embodiment is not only applicable to resumes in standard formats, but also has high robustness and adaptability to diverse texts and multimodal information, making the extraction results more comprehensive and accurate, and providing companies with more efficient and flexible resume analysis technology.

[0041] Embodiment 2: This embodiment discloses a resume information extraction system based on deep learning, which is used to implement the resume information extraction method based on deep learning in Example 1; Figure 2 As shown, the resume information extraction system based on deep learning includes: A preprocessing module is used to obtain the resume text to be processed, and preprocess the resume text to be processed to obtain a preprocessed word segmentation sequence; A label sequence acquisition module is communicatively connected with the preprocessing module, and is used to input the word segmentation sequence into a pre-trained resume information extraction model to obtain an optimal label sequence corresponding to the resume text to be processed; wherein the resume information extraction model includes a BERT encoding layer, a BiLSTM layer, and a CRF layer connected in sequence; the BERT encoding layer is used to perform text feature extraction on the word segmentation sequence to obtain a context feature representation vector sequence of the word segmentation sequence, wherein the context feature representation vector sequence includes context feature representation vectors corresponding to each word in the word segmentation sequence; the BiLSTM layer is used to perform feature sequence encoding processing on the context feature representation vector sequence to obtain a feature vector sequence of the context feature representation vector sequence, wherein the feature vector sequence includes feature vectors corresponding to each word in the word segmentation sequence; the CRF layer is used to perform sequence annotation on the feature vector sequence to obtain an optimal label sequence corresponding to the resume text to be processed; A field sequence enhancement module, which is in communication with the label sequence acquisition module and is used to post-process the resume text to be processed according to the optimal label sequence to obtain an enhanced processed field sequence corresponding to the resume text to be processed; The structured processing module is in communication with the field sequence enhancement module and is used to construct structured resume information corresponding to the resume text to be processed based on the enhanced field sequence.

[0042] It should be noted that the working process, working details and technical effects of the resume information extraction system based on deep learning provided in this embodiment 2 can be found in embodiment 1 and will not be repeated here.

[0043] Embodiment 3: Based on Embodiment 1 or 2, this embodiment discloses an electronic device, which may be a smart phone, a tablet computer, a laptop computer, or a desktop computer. The electronic device may be referred to as a user terminal, a portable terminal, a desktop terminal, etc. Figure 3 As shown, the electronic equipment includes: a memory for storing computer program instructions; and, A processor is used to execute the computer program instructions to complete the operation of a resume information extraction method based on deep learning as described in any one of Example 1.

[0044] Specifically, the processor 301 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 301 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 301 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 301 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen.

[0045] The memory 302 may include one or more computer-readable storage media, which may be non-transitory. The memory 302 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 302 is used to store at least one instruction, which is used to be executed by the processor 301 to implement the resume information extraction method based on deep learning provided in Example 1 of the present application.

[0046] In some embodiments, the terminal may further optionally include: a communication interface 303 and at least one peripheral device. The processor 301, the memory 302 and the communication interface 303 may be connected via a bus or a signal line. Each peripheral device may be connected to the communication interface 303 via a bus, a signal line or a circuit board. Specifically, the peripheral device includes: at least one of a radio frequency circuit 304, a display screen 305 and a power supply 306.

[0047] The communication interface 303 may be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 301 and the memory 302. In some embodiments, the processor 301, the memory 302, and the communication interface 303 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 301, the memory 302, and the communication interface 303 may be implemented on a separate chip or circuit board, which is not limited in this embodiment.

[0048] The radio frequency circuit 304 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 304 communicates with a communication network and other communication devices through electromagnetic signals.

[0049] The display screen 305 is used to display a UI (User Interface). The UI may include any combination of graphics, text, icons, and videos.

[0050] The power supply 306 is used to supply power to various components in the electronic device.

[0051] Embodiment 4: Based on any one of Embodiments 1 to 3, this embodiment discloses a computer program product, including a computer program or instructions, which, when executed by a computer, implements a resume information extraction method based on deep learning as described in any one of Embodiments 1. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0052] Embodiment 5: Based on any one of Embodiments 1 to 3, this embodiment discloses a computer-readable storage product, on which instructions are stored. When the instructions are run on a computer, a resume information extraction method based on deep learning as described in any one of Embodiments 1 is executed. The computer-readable storage product refers to a carrier for storing data, which may include but is not limited to computer-readable storage media such as floppy disks, optical disks, hard disks, flash memories, USB flash drives, and / or memory sticks, and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0053] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A resume information extraction method based on deep learning, characterized in that: include: Obtaining a resume text to be processed, and preprocessing the resume text to be processed to obtain a preprocessed word segmentation sequence; Input the word segmentation sequence into a pre-trained resume information extraction model to obtain the optimal label sequence corresponding to the resume text to be processed; wherein the resume information extraction model includes a BERT encoding layer, a BiLSTM layer and a CRF layer connected in sequence; the BERT encoding layer is used to perform text feature extraction on the word segmentation sequence to obtain a context feature representation vector sequence of the word segmentation sequence, wherein the context feature representation vector sequence includes context feature representation vectors corresponding to each word in the word segmentation sequence; the BiLSTM layer is used to perform feature sequence encoding processing on the context feature representation vector sequence to obtain a feature vector sequence of the context feature representation vector sequence, wherein the feature vector sequence includes feature vectors corresponding to each word in the word segmentation sequence; the CRF layer is used to perform sequence labeling on the feature vector sequence to obtain the optimal label sequence corresponding to the resume text to be processed; Post-processing the resume text to be processed according to the optimal label sequence to obtain an enhanced processed field sequence corresponding to the resume text to be processed; The structured resume information corresponding to the resume text to be processed is constructed based on the enhanced field sequence.

2. A resume information extraction method based on deep learning according to claim 1, characterized in that: Preprocess the resume text to be processed to obtain a preprocessed word segmentation sequence, including: Cleaning the resume text to be processed to obtain a denoised resume text; Performing word segmentation and sentence segmentation processing on the denoised resume text to obtain a sentence segmented resume text; Part-of-speech tagging and named entity recognition are performed on the sentence-segmented resume text to obtain a preprocessed word segmentation sequence.

3. The resume information extraction method based on deep learning according to claim 1 is characterized in that: The resume information extraction model is trained by multi-task learning.

4. The resume information extraction method based on deep learning according to claim 1 is characterized in that: Post-processing the resume text to be processed according to the optimal label sequence to obtain an enhanced processed field sequence corresponding to the resume text to be processed, including: Perform field association and integration processing on the resume text to be processed according to the optimal label sequence to obtain an integrated field sequence corresponding to the resume text to be processed; The integrated field sequence is subjected to knowledge graph enhancement processing to obtain an enhanced field sequence.

5. A resume information extraction method based on deep learning according to claim 4, characterized in that: According to the optimal label sequence, the resume text to be processed is subjected to field association and integration processing to obtain an integrated field sequence corresponding to the resume text to be processed, including: The tags in the optimal tag sequence are confirmed in sequence. If there are consecutive tags that are the same, the word segments corresponding to the tags in the resume text to be processed are merged into one field to obtain a field sequence after the field is merged. According to resume domain knowledge, the fields in the field sequence after the fields are merged are processed by predefined rules to obtain a field sequence after rule processing; Redundant information in the field sequence after the rule processing is removed to obtain an integrated field sequence.

6. A resume information extraction method based on deep learning according to claim 4, characterized in that: Performing knowledge graph enhancement processing on the integrated field sequence to obtain an enhanced field sequence includes: Build a knowledge graph in the resume field based on a preset public database; The completeness of the integrated field sequence is checked. If there is missing information in a field of the integrated field sequence, the entity node corresponding to the field in the knowledge graph is found through the entity linking method, and then the integrated field sequence is completed based on the entity node to obtain an enhanced field sequence.

7. A resume information extraction system based on deep learning, characterized in that: Used to implement a resume information extraction method based on deep learning as described in any one of claims 1 to 6; the resume information extraction system based on deep learning comprises: A preprocessing module is used to obtain the resume text to be processed, and preprocess the resume text to be processed to obtain a preprocessed word segmentation sequence; A label sequence acquisition module is communicatively connected with the preprocessing module, and is used to input the word segmentation sequence into a pre-trained resume information extraction model to obtain an optimal label sequence corresponding to the resume text to be processed; wherein the resume information extraction model includes a BERT encoding layer, a BiLSTM layer, and a CRF layer connected in sequence; the BERT encoding layer is used to perform text feature extraction on the word segmentation sequence to obtain a context feature representation vector sequence of the word segmentation sequence, wherein the context feature representation vector sequence includes context feature representation vectors corresponding to each word in the word segmentation sequence; the BiLSTM layer is used to perform feature sequence encoding processing on the context feature representation vector sequence to obtain a feature vector sequence of the context feature representation vector sequence, wherein the feature vector sequence includes feature vectors corresponding to each word in the word segmentation sequence; the CRF layer is used to perform sequence annotation on the feature vector sequence to obtain an optimal label sequence corresponding to the resume text to be processed; A field sequence enhancement module, which is in communication with the label sequence acquisition module and is used to post-process the resume text to be processed according to the optimal label sequence to obtain an enhanced processed field sequence corresponding to the resume text to be processed; The structured processing module is in communication with the field sequence enhancement module and is used to construct structured resume information corresponding to the resume text to be processed based on the enhanced field sequence.

8. An electronic device, characterized in that: include: a memory for storing computer program instructions; as well as, A processor, used to execute the computer program instructions to complete the operation of a resume information extraction method based on deep learning as described in any one of claims 1 to 6.

9. A computer program product comprising a computer program or instructions, characterized in that When the computer program or the instruction is executed by a computer, it implements a resume information extraction method based on deep learning as described in any one of claims 1 to 6.

10. A computer readable storage product, characterized in that: The computer-readable storage product stores instructions, and when the instructions are executed on a computer, a resume information extraction method based on deep learning as described in any one of claims 1 to 6 is executed.

Images