Intelligent occupational coding method and device in social investigation

Abstract

The invention discloses an intelligent occupational coding method and device in social survey, and relates to the technical field of statistical survey, natural language processing and computers. According to the technical scheme, firstly, word segmentation, part-of-speech tagging and feature extraction are carried out on an occupational text to obtain a feature set; mapping into numeric word embedding through a pre-training word embedding model; analyzing by using an occupational coding model to obtain a potential code (set) and a potential matching probability (set); and determining a target code according to the coding decision scheme. The beneficial effects are that the scheme effectively improves the coding efficiency and accuracy through a series of automatic processing and flexible decision making, can intelligently select the coding mode according to the occupational text features and the matching probability, achieves efficient and accurate intelligent occupational coding, and cooperatively guarantees the implementation of the method through the device, the storage medium and the electronic equipment, thereby improving the efficiency and accuracy of the intelligent occupational coding. And occupational coding work in social investigation is helped to be more scientific and intelligent.

Description

Technical Field

[0001] The present invention relates to the fields of statistical survey, natural language processing and computer technology, and in particular to an intelligent occupation coding method and device in social survey. Background Art

[0002] Many social surveys, such as population censuses and labor force surveys under the national statistical survey system, collect occupational information from respondents. These surveys typically include open-ended questions like "What is your occupation?" or similarly worded questions. The occupational information collected is unstructured text data. To facilitate management and analysis, it must be converted into structured numerical data. Occupational coding is the process of converting occupational text into corresponding numerical codes according to designated occupational classification standards. Occupational coding is essential for organizing occupational survey data.

[0003] In existing technology, occupational coding is primarily done manually, with professional coders assigned to occupational texts based on their understanding and mastery of occupational classification standards. Manual coding has three main drawbacks. First, it requires a large number of pre-selected personnel, which is very costly. Second, coding is time-consuming, compromising the timeliness of collected data. Third, the coding process relies on the coder's experience and subjective judgment, making it difficult to ensure the reliability and effectiveness of the coding.

[0004] Existing computerized occupational coding methods have partially addressed the challenges of manual coding. This method abstracts the occupational coding process into a text classification problem within the field of natural language processing, establishes a computerized coding model, and applies it to social surveys. However, a significant drawback of current computerized coding methods is that they still require human input for final decision-making. This method's ability to save costs and improve efficiency in large-scale coding scenarios is limited. Summary of the Invention

[0005] In order to solve the technical problems existing in the above-mentioned prior art, the present invention provides an intelligent occupation coding method and device for social survey scenarios, which realizes computer-automatic or assisted occupation coding, and is particularly suitable for large-scale occupation coding work in social surveys, greatly saving labor costs and time costs, and improving the timeliness of occupation data.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The first aspect of the present invention is to provide an intelligent occupation coding method for social surveys, comprising the following steps:

[0008] Perform word segmentation, part-of-speech tagging, and feature extraction on the occupational text to be coded to determine a feature set corresponding to the occupational text;

[0009] Mapping the feature set using a pre-trained word embedding model to represent the occupation text as a numerical word embedding;

[0010] Using a preset occupation coding model, perform coding analysis on the word embedding to determine the potential coding (set) and potential matching probability (set) of the occupation text;

[0011] A coding decision plan is formulated, and a target code corresponding to the occupational text is determined from the potential codes (set) according to the coding decision plan.

[0012] Furthermore, the steps of segmenting and cleaning the occupational texts collected in the survey to determine the feature set corresponding to the occupational texts are as follows:

[0013] Performing word segmentation processing on the occupation text to process the occupation text into semantically independent word sequences;

[0014] Part-of-speech tagging is performed on each word in the word sequence, and verbs, nouns, adjectives, and adverbs therein are extracted to determine the feature set.

[0015] Furthermore, the steps of mapping the feature set using the pre-trained word embedding model to represent the occupation text as a numerical word embedding are as follows:

[0016] Using a pre-trained word embedding model, the features in the feature set are mapped one by one into word embeddings (vectors), so that the features of the occupation text can be recognized and calculated by a computer;

[0017] The mean of all the word embeddings is calculated to determine the word embedding of the occupation text.

[0018] Furthermore, the use of a preset occupation coding model to perform coding analysis on the word embedding to determine the potential coding (set) and potential matching probability (set) of the occupation text includes:

[0019] Using a preset occupation coding model, perform coding analysis on the word embedding to generate a matching probability between the occupation text and any occupation code in the coding table;

[0020] One or more occupation codes with the highest matching probability are selected to determine a potential code (set) and a potential matching probability (set) corresponding to the potential code.

[0021] Furthermore, the formulating of a coding decision scheme and determining the target coding corresponding to the occupation text from the potential coding (set) according to the coding decision scheme include:

[0022] Develop coding decision plans;

[0023] Determining the encoding method used for the occupational text according to the encoding decision scheme;

[0024] The encoding method is used to determine the target encoding corresponding to the occupation text.

[0025] Furthermore, the formulation of the coding decision plan includes:

[0026] Determine the cut-off score;

[0027] Determine the minimum score.

[0028] Furthermore, the encoding method includes:

[0029] The computer automatic coding method refers to the computer autonomously determining the target coding corresponding to the occupational text;

[0030] Computer-assisted manual coding means that the computer provides an auxiliary coding set, and the coder selects elements from the auxiliary coding set based on the occupational text to determine the target code corresponding to the occupational text;

[0031] Manual independent coding means that the coder determines the target code corresponding to the occupational text based on the occupational text and relies solely on his or her own knowledge and experience.

[0032] Furthermore, the determining of the encoding method to be used for the occupation text according to the encoding decision scheme includes:

[0033] Identifying the maximum of the potential matching probability (set);

[0034] Comparing the potential matching probability (set) with the cutoff score and the minimum score;

[0035] When the maximum value is not lower than the cutoff score, a computer automatic coding method is adopted.

[0036] When the maximum value is lower than the cutoff score but not lower than the minimum score, a computer-assisted manual coding method is used.

[0037] When all of the potential matching probabilities are lower than the minimum score, manual independent coding is adopted.

[0038] Furthermore, the use of the encoding method to determine the target code corresponding to the occupation text includes:

[0039] When a computer automatic coding method is adopted, the computer autonomously determines the potential code corresponding to the maximum value as the target code corresponding to the occupational text.

[0040] When computer-assisted manual coding is used, the potential matching probability (set) is compared with the minimum score, and the potential codes with a potential matching probability not less than the minimum score are selected and determined as the computer-assisted coding set. The computer-assisted coding set is then submitted to a human coder, who, based on their knowledge and experience, selects an element from the computer-assisted coding set and determines it as the target code corresponding to the occupational text.

[0041] When the manual independent coding method is adopted, the occupational text is submitted to a manual coder, and the coder determines the target code corresponding to the occupational text based on his or her own knowledge and experience.

[0042] The second aspect of the present invention is to provide an intelligent occupation coding device for social surveys. The device comprises:

[0043] A feature set determination module is used to perform word segmentation, part-of-speech tagging, and feature extraction on the occupation text, and determine a feature set corresponding to the occupation text;

[0044] a word embedding determination module, configured to map the feature set using a pre-trained word embedding model to determine word embeddings for the occupation text;

[0045] a potential coding set determination module, configured to perform coding analysis on the word embedding using a preset occupation coding model to determine the potential coding (set) and potential matching probability (set) of the occupation text;

[0046] The target coding determination module is used to determine the target coding corresponding to the occupation text from the potential coding (set) according to a preset coding decision scheme.

[0047] A third aspect of the present invention is to provide a computer-readable storage medium comprising:

[0048] The storage medium is used to store computer programs.

[0049] The computer program is used to execute the intelligent occupation coding method in social survey described in the present invention.

[0050] A fourth aspect of the present invention is to provide an electronic device, comprising:

[0051] processor;

[0052] a memory for storing instructions executable by the processor;

[0053] The processor is used to read the executable instructions from the memory and execute the instructions to implement the intelligent occupation coding method in social surveys described in the present invention.

[0054] Beneficial effects of the present invention:

[0055] Compared with the prior art, the intelligent occupation coding method and device for social surveys described in the present invention have the following technical features and beneficial effects:

[0056] (1) Improve coding efficiency and automation level: Through automated processes such as word segmentation, part-of-speech tagging, feature extraction, word embedding mapping, and occupational coding model analysis, manual intervention is greatly reduced and the occupational coding cycle is significantly shortened. This is especially suitable for large-scale data processing in social surveys, which improves the timeliness of data.

[0057] (2) Improve coding accuracy and consistency: Utilizing pre-trained word embedding models and hierarchical multi-label occupation coding models, combined with occupational classification standards, a potential coding set is generated through semantic analysis and probabilistic matching, which reduces the subjective differences in manual coding and enhances the reliability and standardization of the coding results.

[0058] (3) Flexible and intelligent coding decision-making mechanism: By setting the cutoff score and the minimum score, the system dynamically selects computer automatic coding, assisted manual coding, or manual independent coding. It ensures full automation in high-confidence scenarios and introduces manual review in low-confidence scenarios, thus balancing efficiency and accuracy and optimizing resource allocation.

[0059] (4) Significantly reduce labor and time costs: Through computer-assisted or automatic coding, the reliance on professional coders is reduced, thus reducing labor costs. At the same time, the automated process shortens coding time and improves the efficiency of processing social survey data.

[0060] (5) Adaptability to complex texts and large-scale scenarios: It supports efficient processing of unstructured professional texts, and the system design is scalable and can stably cope with large-scale professional coding needs.

[0061] (6) Enhanced interpretability and practicality: Part-of-speech tagging, feature extraction, and probability matching processes enhance the interpretability of the model and facilitate manual review. At the same time, the computer-assisted coding set provides a reference for manual coders, further improving the scientificity and practicality of coding decisions.

[0062] This invention achieves efficient, accurate and intelligent occupational coding by deeply integrating natural language processing technology and intelligent decision-making mechanism, providing strong support for the scientific management and analysis of social survey data. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0064] in:

[0065] Figure 1 A flowchart of an intelligent occupation coding method for social surveys provided by an embodiment of the present invention;

[0066] Figure 2 A schematic diagram of the structure of an intelligent occupation coding device for social surveys provided by an embodiment of the present invention.

[0067] Figure 3 This is a schematic diagram of the interface structure of the intelligent occupational encoding device of the present invention;

[0068] Figure 4 This is a schematic diagram of the structure and interface module relationship of the intelligent occupational coding device of the present invention. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Figures 1-4 Further explanation of the intelligent occupation coding method and device in social surveys.

[0070] Example 1: Method Example

[0071] This embodiment provides an intelligent occupation coding method for social surveys, such as Figure 1 As shown, the embodiment includes the following steps:

[0072] Step 1: perform word segmentation, part-of-speech tagging, and feature extraction on the occupational texts to be coded after the survey collection to determine the feature set corresponding to the occupational texts.

[0073] The occupational texts to be coded are textual data reported by respondents in social surveys, describing their occupations. Chinese occupational texts are typically composed of several or dozens of characters, lack natural word separators, and are interspersed with nonstandard terms and redundant morphemes. Therefore, before using computers to analyze occupational texts, they require text preprocessing, which involves the following three steps:

[0074] Step 1.1: Text segmentation. This involves dividing the sentences in the occupational text into semantically independent word sequences. Three examples were selected from occupational texts submitted by respondents in a social survey. Using an existing word segmentation database and algorithm, the original occupational texts and the word sequences after segmentation are shown in the following table:

[0075] Example Number Original Professional Text Word Sequence after Word Segmentation 1 Serving as a secretary in the village committee {'in', 'a certain village', 'village committee','serving as','secretary'} 2 3

[0076] Step 1.2, part-of-speech tagging, assigns an appropriate part-of-speech tag to each word or morpheme in the occupation text, such as noun, verb, adjective, adverb, etc. Part-of-speech tagging helps improve the accuracy of feature extraction and enhance the interpretability of the occupation coding model. The results of part-of-speech tagging for Examples 1-3 in this embodiment are shown in the following table:

[0077] Example Number Part-of-Speech Tagging Result 1 {'in'-preposition, 'a certain village'-place name, 'village committee'-noun,'serving as'-verb,'secretary'-noun} 2 3

[0078] Step 1.3, feature extraction, aims to extract the key features of the occupational text. Nouns, verbs, nouns, adjectives, and adverbs in the occupational text are of obvious significance to the occupational coding. Therefore, the words in the word sequence with the above four parts of speech are extracted to form the feature set of the occupational text. The feature set extracted in Examples 1-3 of this embodiment is as follows:

[0079] Example Number Part-of-Speech Tagging Result 1 {'village committee','serving as','secretary'} 2 3

[0080] Step 2: Use the pre-trained word embedding model to map the feature set to represent the occupation text as a numerical word embedding.

[0081] Since professional text cannot be directly recognized and operated by computer algorithms, it is necessary to convert the text into a numerical form that can be processed by computers. Text vectorization is the process of mapping natural language information from text space to vector space. It specifically includes the following two steps:

[0082] Step 2.1, feature word embedding, maps words into numerical vectors so that words can be recognized and calculated by computers. Specifically, word embedding maps words into a high-dimensional vector space so that semantically similar words are closer in the vector space. The pre-trained word embedding model used in this example is based on the Chinese Wikipedia corpus and is trained using the Word2Vec model method. The word embedding dimension is set to 100. The word embedding model is used to map the features in the feature sets of Examples 1-3 in this embodiment to word embeddings one by one. The word embeddings are as follows:

[0083] Example Number Part-of-Speech Tagging Result 1 {'village committee'-[n],'serving as'-[v],'secretary'-[n]} 2 3

[0084] Step 2.2, Text Sentence Embedding, maps the occupational text sentences into numerical vectors, making them computer-recognizable and computationally identifiable. For occupational texts with more than one feature element, the mean of the word embeddings for all features is calculated as the sentence embedding for the occupational text. The sentence embeddings for Examples 1-3 of this embodiment are shown in the following table:

[0085] Example Number Text Sentence Embedding Result 1 serving as: [-0.16527951 0.27950493... -0.09866355 -0.10846316] 2 village committee: [-0.12884776... -0.10463286 0.09527693] 3 secretary: [0.09244609 0.03663632... -0.28884773 0.35641267]

[0086] Step 3: Use the preset occupation coding model to perform coding analysis on the word embedding of the occupation text to determine the potential coding (set) and potential matching probability (set) of the occupation text.

[0087] The intelligent process of occupational coding is abstracted into a natural language text classification problem. Using text classification based on supervised learning as a technical means, a hierarchical multi-label coding model is constructed. The words of the occupational text to be coded are embedded into the input model. The model generates the matching probability between the occupational text and any occupational code in the coding table through analysis. The higher the matching probability, the more likely the corresponding occupational code is to be the target code of the occupational text. One or more occupational codes with the highest matching probability are selected to determine the potential code (set). The target code of the occupational text is very likely to be one of the elements in the potential code set. Specifically, it includes the following three steps:

[0088] Step 3.1, preset the occupation coding model. The occupation coding model can be a knowledge system based on the national standard "Occupational Classification and Code" (GB6565-2015) and the occupational knowledge graph, or it can be a machine learning text classification model trained based on occupational text and code annotation samples. In this example, in order to improve the accuracy of occupational coding, a deep learning occupation coding model with a hierarchical structure was established. Specifically, the "Occupational Classification and Code" (GB6565-2015) is used as the occupation coding table, and the occupational texts and coding results accumulated from a social survey are used as training samples. A hierarchical perception global model structure is adopted to make full use of the interaction between text features and occupational codes, and an occupational coding model that is simultaneously oriented towards major, medium and minor occupational categories is trained.

[0089] Step 3.2: Use the occupation coding model to generate the matching probability of the occupation text. The word embedding of the occupation text is input into the preset hierarchical multi-label coding model. The model returns the matching probability with any occupation code in the coding table. The matching probability ranges from [0, 1]. The higher the matching probability, the more likely the corresponding occupation code is the target code of the occupation text. This example takes the occupation subcategory coding result returned by the model as an example. The result returned by the occupation text in Example 1 is:

[0090] Occupation Code (Subcategory) Matching Probability 20212 0.84 20209 0.16

[0091] Example 2: The result of the occupation text is:

[0092] Occupation Code (Subcategory) Matching Probability 20609 0.57 20610 0.33 20611 0.10

[0093] Example 3: The occupation text returns the following result:

[0094] Occupation Code (Subcategory) Matching Probability 41202 0.31 60801 0.39 61109 0.30

[0095] Step 3.3: Determine the potential code (set). Based on the preset selection criteria, select one or more occupation codes with the highest matching probability to determine the potential code (set) and the potential matching probability (set) corresponding to the potential code. In this embodiment, the five occupation codes with the highest matching probability are selected to form the potential code set. The potential code set and the corresponding potential matching probability set for the occupation text in Example 1 in this embodiment are:

[0096] Rank Occupation Code (Subcategory) Matching Probability 1 20212 0.84 2 20209 0.16 3 4 5

[0097] In this embodiment, the potential coding set of the occupation text in Example 2 and the corresponding potential matching probability set are:

[0098] Rank Occupation Code (Subcategory) Matching Probability 1 20609 0.57 2 20610 0.33 3 20611 0.10 4 5

[0099] In this embodiment, the potential coding set of the occupation text in Example 3 and the corresponding potential matching probability set are:

[0100] Rank Occupation Code (Subcategory) Matching Probability 1 41202 0.31 2 60801 0.39 3 61109 0.30 4 5

[0101] Step 4: formulate a coding decision plan, and determine the target code corresponding to the occupational text from the potential codes (set) based on the coding decision plan.

[0102] The so-called coding decision plan determines whether the occupational text is suitable for coding by computer automatic coding, computer-assisted manual coding, or independent manual coding by setting two indicators: a cutoff score and a minimum score. In the context of large-scale coding in social surveys, the coding decision plan determines the extent to which occupational texts rely on manual coding, which directly affects the cost and efficiency of occupational coding work. Setting a lower cutoff score and minimum score means that there is more trust in the computer coding results, which may lead to certain systematic coding errors, relatively less use of manual labor, lower coding costs and higher efficiency. Conversely, setting a higher cutoff score and minimum score means that the coding work relies more on manual labor, which helps reduce systematic errors (and may introduce some variable errors), but the coding cost is higher and the efficiency is lower. Determining the target coding corresponding to the occupational text based on the coding decision plan specifically includes the following three steps:

[0103] Step 4.1: Develop a coding decision plan. Set a cutoff score and minimum score. These can be set empirically or based on analysis of the distribution of matching probabilities. The cutoff score and minimum score are in the range [0, 1], and the cutoff score must not be lower than the minimum score. In this example, the cutoff score is set to 0.8, and the minimum score is set to 0.4.

[0104] Step 4.2, determining the encoding method used for the occupation text: comparing the potential matching probability (set) of the occupation text with the minimum score and the numerical value of the minimum score to determine the encoding method used for the occupation text.

[0105] Identify the maximum value in the potential matching probability set of the occupational text. If the maximum value is not lower than the cutoff score, the occupational text will be automatically coded by the computer, and the computer will independently determine the target coding corresponding to the occupational text.

[0106] If the maximum value in the matching probability set is lower than the cutoff score but not lower than the minimum score, the occupational text will be coded manually using computer-assisted coding, that is, the computer provides an auxiliary coding set, and the coder selects elements from it to determine the target coding corresponding to the occupational text.

[0107] If the maximum value in the matching probability set is lower than the minimum score, the occupational text will be coded manually and independently, that is, the coder relies entirely on his or her own knowledge and experience to determine the target coding corresponding to the occupational text.

[0108] The encoding methods used in Examples 1-3 of this embodiment are as follows:

[0109]

[0110]

[0111] Step 4.3, determine the target coding corresponding to the occupational text.

[0112] When a computer automatic coding method is adopted, the potential code corresponding to the maximum value is determined as the target code corresponding to the occupational text.

[0113] When using computer-assisted manual coding, all occupation codes with a potential matching probability of at least a minimum score are first selected to form a computer-assisted code set. This set is then submitted to a human coder, who, based on their knowledge and experience, selects an element from the set as the target code corresponding to the occupation text.

[0114] When using the manual independent coding method, the occupational text is submitted to a manual coder, who uses his or her own knowledge and experience to determine the target code corresponding to the occupational text.

[0115] The process of determining the target code in Examples 1-3 of this embodiment is as follows:

[0116]

[0117] Example 2: Device Example

[0118] This embodiment provides an intelligent occupation coding device for social surveys, such as Figure 2 This embodiment is a physical device that can execute the above method embodiment, and its functions correspond one-to-one with the steps of the above method embodiment. The device of this embodiment includes:

[0119] A feature set determination module (module 1) is used to perform word segmentation, part-of-speech tagging, and feature extraction on the occupation text, and determine the feature set corresponding to the occupation text;

[0120] A word embedding determination module (module 2) is configured to map the feature set using a pre-trained word embedding model to determine the word embedding of the occupation text;

[0121] A potential coding set determination module (module 3) is used to perform coding analysis on the word embedding using a preset occupation coding model to determine the potential coding (set) and potential matching probability (set) of the occupation text;

[0122] The target coding determination module (module 4) is used to determine the target coding corresponding to the occupation text from the potential coding (set) according to a preset coding decision scheme.

[0123] Specifically, in Figure 2 The framework of the device embodiment shown also includes:

[0124] The feature set determination module includes:

[0125] The text segmentation unit (unit 1.1) is used to divide sentences in professional texts into semantically independent word sequences.

[0126] The part-of-speech tagging unit (unit 1.2) is used to assign an appropriate part-of-speech tag to each word or morpheme in the occupational text.

[0127] Feature extraction unit (unit 1.3) is used to extract key features of occupational texts.

[0128] The word embedding determination module includes:

[0129] The feature word embedding unit (unit 2.1) is used to map words into numerical vectors.

[0130] The text sentence embedding unit (unit 2.2) is used to project professional text sentences into numerical vectors.

[0131] The potential coding set determination module includes:

[0132] The preset occupation coding model unit (unit 3.1) is used to store the preset occupation coding model.

[0133] The matching probability generation unit (unit 3.2) is used to generate the matching probability between the occupation text and any occupation code in the coding table.

[0134] Determine the potential code(s) The determination unit (unit 3.3) is used to determine the potential code(s) of the occupational text.

[0135] The target encoding determination module includes:

[0136] The coding decision scheme formulation unit (unit 4.1) is used to set indicators in the coding decision scheme.

[0137] The coding method determination unit (Unit 4.2) is used to determine the coding method used for professional texts.

[0138] The target code determination unit (unit 4.3) is used to determine the target code corresponding to the occupational text.

[0139] The present invention first determines the feature set and word embeddings of the occupation text to be coded. It then uses a pre-set coding model to determine the potential codes (sets) and potential matching probabilities (sets). Finally, it flexibly adopts coding methods to determine the target code. This enables computer-automated or computer-assisted occupation coding, making it particularly suitable for large-scale occupation coding in social surveys, significantly saving labor and time costs and improving the timeliness of occupational data.

[0140] Example 3: Computer Program Product Example

[0141] This embodiment is a computer program product capable of executing the above-described method embodiment. The program instructions contained in the computer program product, when executed, are capable of performing the steps in the above-described method embodiment. The computer program product can be written in one or more programming languages, such as C, C++, and Java. The program code can be executed in whole or in part, and can be executed on a local computer device, a remote computing device, or a server.

[0142] Embodiment 4: Computer-readable storage medium embodiment

[0143] This embodiment is a computer-readable storage medium, and the program instructions stored therein are capable of executing the steps of the above-described method embodiment when executed. The computer-readable storage medium may be a single readable medium or a combination of multiple readable media. The readable medium may be a readable signal medium or a readable storage medium. Readable signal media include, but are not limited to, infrared, magnetic, semiconductor, etc.; readable storage media include, but are not limited to, hard disks, portable disks, card-type memories (such as SD cards), read-only memories (ROMs), random access memories (RAMs), etc.

[0144] The above implementation is only a preferred embodiment of the intelligent occupation coding method and device in social survey of the present invention, and does not limit the form and scope of the present invention. Any disassembly, reassembly, equivalent structure and equivalent process made using the contents of the description and drawings of the present invention are also included in the patent protection scope of the present invention.

[0145] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. An intelligent occupation coding method for social surveys, characterized in that: Here are the steps: S1. Perform word segmentation, part-of-speech tagging, and feature extraction on the occupational text to be encoded to determine a feature set corresponding to the occupational text; S2. Mapping the feature set using a pre-trained word embedding model to represent the occupation text as a numerical word embedding; S3. Using a preset occupation coding model, perform coding analysis on the word embedding to determine a potential code or code set, and a potential matching probability or probability set of the occupation text; S4. Formulate a coding decision plan, and determine the target code corresponding to the occupational text from the potential codes or code sets based on the coding decision plan.

2. The intelligent occupation coding method in social survey according to claim 1 is characterized in that: The specific steps of step S1 are as follows: S11, performing word segmentation processing on the occupation text to process the occupation text into semantically independent word sequences; S12. Tagging the words in the word sequence one by one, extracting the verbs, nouns, adjectives and adverbs therein, and determining the feature set.

3. The intelligent occupation coding method in social survey according to claim 1 is characterized in that: The specific steps of step S2 are as follows: S21, using a pre-trained word embedding model to map the features in the feature set into word embedding vectors one by one; S22. Calculate the mean of all the word embeddings to determine the word embedding of the occupation text.

4. The intelligent occupation coding method in social survey according to claim 1 is characterized in that: The specific steps of step S3 are as follows: S31, using a preset occupation coding model, analyzing the word embedding to generate a matching probability between the occupation text and any occupation code in the coding table; S32. Select one or more occupation codes with the highest matching probability to determine a potential code or code set, and a potential matching probability or probability set.

5. The intelligent occupation coding method in social survey according to claim 1 is characterized in that: In step S4, the steps for formulating a coding decision plan are as follows: S41. Determine the cut-off score and minimum score; S42, comparing the maximum value of the potential matching probability set with the cutoff score and the minimum score; S43. When the maximum value is not lower than the cutoff score, computer automatic coding is used; when the maximum value is lower than the cutoff score but not lower than the minimum score, computer-assisted manual coding is used; when the maximum value is lower than the minimum score, manual independent coding is used.

6. The intelligent occupation coding method in social survey according to claim 5 is characterized in that: Encoding methods include: (1) Computer automatic coding method: The computer independently determines the target code corresponding to the occupational text; (2) Computer-assisted manual coding: The computer provides an auxiliary code set, and the coder selects from the auxiliary code set; (3) Manual independent coding method: The coder determines the target code corresponding to the occupational text based solely on his or her own knowledge and experience.

7. An intelligent occupation coding device for social surveys, characterized in that: include: A feature set determination module, configured to perform word segmentation, part-of-speech tagging, and feature extraction on the occupation text to determine a feature set corresponding to the occupation text; a word embedding determination module, configured to map the feature set using a pre-trained word embedding model to determine a word embedding for the occupation text; a latent coding set determination module, configured to perform coding analysis on the word embedding using a preset occupation coding model to determine a latent code or code set, and a potential matching probability or probability set of the occupation text; The target coding determination module is used to determine the target coding corresponding to the occupation text from the potential coding or coding set according to a preset coding decision scheme.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, which, when executed, implements the intelligent occupation coding method in social surveys described in any one of claims 1 to 6.

9. An electronic device, characterized in that: include: processor; Memory; The memory stores instructions that can be executed by the processor, and when the instructions are executed, the intelligent occupation coding method in social survey described in any one of claims 1-6 is implemented.

Images