Generative language model assisted rooting theory coding optimization method
Through the generative language model assists rooted theoretical coding, automatically process text and conduct in-depth semantic analysis, combined with manual review and dynamic update, the problem of time-consuming and poor consistency of rooted theoretical coding is solved, and an efficient and accurate coding process is achieved.
Patent Information
- Application Number
- CN202510375621.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-27
- Publication Date
- 2025-07-11
AI Technical Summary
The existing grounded theory coding process is time-consuming, subjective and poor consistency. The existing automation tools lack dynamic optimization mechanisms and cannot adapt to complex and changeable text environments.
Generative language model is used to assist in grounded theoretical coding, and the original text is processed through transcription and sentence-making tools. The pre-trained language model is used for semantic analysis, and feature vectors are generated. The text is automatically classified based on cosine similarity and semantic clustering algorithms, and the encoding corpus and feature sets are updated dynamically.
Improve coding efficiency and accuracy, reduce manual workload, ensure coding consistency and reliability, and the model realizes automated and intelligent coding through continuous optimization and adaptation to data changes.
Smart Images
Figure CN120297231A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of artificial intelligence and qualitative research, and in particular to a method for optimizing grounded theory coding assisted by a generative language model. Background Art
[0002] Grounded theory is the core method of qualitative research, but its traditional coding process relies on manual operation, with defects such as time-consuming, strong subjectivity, and poor consistency. In the prior art, although individual studies have attempted to introduce natural language processing technology, most are limited to keyword matching or simple classification and have not been deeply integrated with the three-level coding framework of grounded theory. In addition, existing automated tools lack a dynamic optimization mechanism and cannot adapt to complex and changing text environments. Therefore, there is an urgent need for an intelligent coding scheme that combines deep semantic analysis and continuous learning. Summary of the Invention
[0003] To improve the coding efficiency, accuracy, and consistency in qualitative research, this application provides a method for optimizing grounded theory coding assisted by a generative language model.
[0004] The first above-mentioned invention object of this application is achieved through the following technical solutions:
[0005] A method for optimizing grounded theory coding assisted by a generative language model includes the steps of:
[0006] Receiving original text data, and converting the original text data into an initial corpus set after clause segmentation through a transcription and clause segmentation tool;
[0007] Using a pre-trained generative language model to perform semantic analysis on the initial corpus set, extract text features, and generate a set of feature vectors;
[0008] Based on the set of feature vectors, by calculating the similarity between the new text and the preset open coding, axial coding, and selective coding corpora, automatically classifying the new text into the corresponding three-level coding categories, and outputting the automatic classification result;
[0009] Manually reviewing the automatic classification result, correcting coding items with low matching degrees or ambiguities, and feeding back the correction result to the generative language model to optimize the model parameters;
[0010] According to the new data and the manual review result, dynamically update the coding corpus and the text feature set to achieve continuous optimization of the model.
[0011] By adopting the above technical solution, through automated transcription and sentence segmentation, a large amount of original text data is quickly converted into a structured corpus, improving the efficiency of data processing. Utilizing the deep semantic analysis ability of the generative language model, it can more accurately extract text features, generate a multi-dimensional set of feature vectors, which helps to deeply understand the text content. Automatically classifying new texts into the third-level coding categories significantly reduces the workload of manual coding and improves the consistency and accuracy of coding. The manual review step ensures the reliability of the coding results, and the feedback mechanism helps to continuously improve the performance of the model and the coding quality. By dynamically updating the coding corpus and the text feature set, the model can continuously adapt to new data and knowledge, maintain its optimized state, realizing the automation and intelligence of the grounded theory coding process. Compared with the traditional manual coding method, it can greatly reduce the time and effort required by researchers during the coding process and significantly improve the coding efficiency.
[0012] In a preferred example of the present application: The step of using the pre-trained generative language model to perform semantic analysis on the initial corpus, extract text features, and generate a set of feature vectors specifically includes the steps of:
[0013] Extract keywords and perform semantic association analysis on the segmented text;
[0014] Generate the context representation of the text based on the attention mechanism;
[0015] Combine the pre-trained word embedding technology to generate feature vectors with multi-dimensional semantic information.
[0016] By adopting the above technical solution, keyword extraction and semantic association analysis help to focus on the key information in the text and understand the deep connections between words, laying a solid foundation for the subsequent generation of feature vectors. Using the attention mechanism to capture the key information in the text, the generated context representation can better reflect the true meaning of the text. Combining the pre-trained word embedding technology, the generated feature vectors contain rich semantic information, which helps to improve the accuracy of coding and the classification effect.
[0017] In a preferred example of the present application: The step of automatically classifying the new text into the corresponding third-level coding category and outputting the automatic classification result based on the set of feature vectors by calculating the similarity between the new text and the preset open coding, axial coding, and selective coding corpora specifically includes the steps of:
[0018] Calculate the similarity between the new text and the open coding corpus through the cosine similarity algorithm to determine the initial coding category;
[0019] Based on the initial coding category, combine the semantic clustering algorithm to further classify the new text into the axial coding;
[0020] Through dependency parsing and topic modeling, map the spindle coding results to the selective coding categories and output the automatic classification results.
[0021] By adopting the above technical solutions, the cosine similarity algorithm can quickly determine the initial coding category of the new text, providing a basis for subsequent more refined classification. Conducting more detailed classification based on the initial coding category helps improve the accuracy of coding. By using advanced language processing techniques, map the text to the most appropriate coding category, improving the scientificity and rationality of the classification results.
[0022] In a preferred example of this application: The steps of dynamically updating the coding corpus and text feature set according to the new data and the manual review results to achieve continuous optimization of the model specifically include the steps of:
[0023] Based on the incremental learning algorithm, input the new data into the generative language model for incremental training;
[0024] Through the adaptive weight adjustment strategy, balance the feature weights of historical data and new data;
[0025] Regularly perform redundancy removal on the coding corpus and retain the feature entries with high confidence.
[0026] By adopting the above technical solutions, the incremental learning algorithm enables the model to continuously evolve using new data, maintaining the timeliness and accuracy of the model, balancing the influence of old and new data, so that the model does not bias towards any party and maintains the fairness of its prediction. Regularly cleaning the coding corpus to ensure its simplicity and efficiency helps improve the operation efficiency and classification effect of the model.
[0027] The second above-mentioned invention object of this application is achieved through the following technical solutions:
[0028] A grounded theory coding optimization system assisted by a generative language model, comprising:
[0029] A data preprocessing module, used to receive the original text data and convert the original text data into an initial corpus set after sentence segmentation through a transcription and sentence segmentation tool;
[0030] A feature extraction module, used to perform semantic analysis on the initial corpus set by using a pre-trained generative language model, extract text features, and generate a set of feature vectors;
[0031] An intelligent three-level coding classification module, used to automatically classify the new text into the corresponding three-level coding category based on the set of feature vectors by calculating the similarity between the new text and the preset open coding, spindle coding, and selective coding corpora, and output the automatic classification results;
[0032] Manual review and adjustment module, which is used to manually review the automatic classification results, correct the coding items with low matching degree or ambiguity, and feedback the correction results to the generative language model to optimize the model parameters;
[0033] Feature set iterative optimization module, which is used to dynamically update the coding corpus and text feature set according to the new data and manual review results, so as to realize the continuous optimization of the model.
[0034] By adopting the above technical solutions, the functions of each module of the system are clearly defined, which is easy to manage and maintain, improves the stability and scalability of the system. The system can continuously optimize according to data and feedback, and adapt to the changing research requirements and environment.
[0035] The above-mentioned third object of the present application is achieved by the following technical solutions:
[0036] A computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned generative language model-assisted grounded theory coding optimization method are realized.
[0037] The above-mentioned fourth object of the present application is achieved by the following technical solutions:
[0038] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned generative language model-assisted grounded theory coding optimization method are realized.
[0039] In summary, the present application includes at least one of the following beneficial technical effects:
[0040] 1. Through automated transcription and sentence segmentation, a large amount of raw text data can be quickly converted into a structured corpus set, improving the efficiency of data processing. Utilizing the deep semantic analysis ability of the generative language model, text features can be extracted more accurately, generating a multi-dimensional feature vector set, which helps to deeply understand the text content. Automatically classifying new texts into the third-level coding categories can greatly reduce the workload of manual coding and improve the consistency and accuracy of coding. The manual review step ensures the reliability of the coding results, and the feedback mechanism helps to continuously improve the performance of the model and the coding quality. By dynamically updating the coding corpus and text feature set, the model can continuously adapt to new data and knowledge, maintain its optimized state, realize the automation and intelligence of the grounded theory coding process, and can significantly reduce the time and effort required by researchers in the coding process compared with the traditional manual coding method, and significantly improve the coding efficiency;
[0041] 2. Keyword extraction and semantic association analysis help to focus on the key information in the text and understand the deep connections between words, laying a solid foundation for the subsequent generation of feature vectors. By using the attention mechanism to capture the key information in the text, the generated context representation can better reflect the true meaning of the text. Combining with the pre-trained word embedding technology, the generated feature vectors contain rich semantic information, which helps to improve the accuracy of encoding and classification effect;
[0042] 3. The cosine similarity algorithm can quickly determine the initial encoding category of the new text, providing a basis for subsequent more refined classification. Conducting more detailed classification based on the initial encoding category helps to improve the accuracy of encoding. By using advanced language processing technology to map the text to the most appropriate encoding category, the scientificity and rationality of the classification results are improved;
[0043] 4. The incremental learning algorithm enables the model to continuously evolve using new data, maintaining the timeliness and accuracy of the model. Balancing the influence of new and old data so that the model does not favor either side and maintaining the fairness of its predictions. Regularly cleaning the encoding corpus to ensure its simplicity and efficiency helps to improve the operation efficiency and classification effect of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 is a flowchart of an embodiment of a grounded theory coding optimization method assisted by a generative language model of the present application;
[0045] Figure 2 is a flowchart of the implementation of step S20 in an embodiment of a grounded theory coding optimization method assisted by a generative language model of the present application;
[0046] Figure 3 is a flowchart of the implementation of step S30 in an embodiment of a grounded theory coding optimization method assisted by a generative language model of the present application;
[0047] Figure 4 is a schematic diagram of a grounded theory coding optimization system assisted by a generative language model of the present application;
[0048] Figure 5 is a schematic block diagram of a computer device of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0049] The following further describes the present application in detail with reference to the accompanying drawings.
[0050] In one embodiment, as Figures 1 - 3 shown, the present application discloses a grounded theory coding optimization method assisted by a generative language model, which specifically includes the following steps:
[0051] S10: Receive the original text data, and convert the original text data into an initial corpus set after clause segmentation through a transcription and clause segmentation tool;
[0052] S20: Use a pre-trained generative language model to perform semantic analysis on the initial corpus set, extract text features, and generate a set of feature vectors;
[0053] S30: Based on the set of feature vectors, by calculating the similarity between the new text and the preset open coding, axial coding, and selective coding corpora, automatically classify the new text into the corresponding three-level coding categories, and output the automatic classification result;
[0054] S40: Manually review the automatic classification result, correct the coding items with low matching degree or ambiguity, and feedback the correction result to the generative language model to optimize the model parameters;
[0055] S50: Dynamically update the coding corpus and the text feature set according to the new data and the manual review result to achieve continuous optimization of the model.
[0056] In this embodiment, the original text data may be interview records, observation notes, etc. Use a transcription tool to convert speech or video data into text. The preset coding corpus includes open coding, axial coding, and selective coding.
[0057] Specifically, use a clause segmentation tool to split continuous text into independent sentences to form an initial corpus set. Use a pre-trained generative language model to deeply analyze the initial corpus set, understand the semantic content of the text, extract text features, and convert these features into a set of feature vectors. These feature vectors can represent the semantic information of the text. Based on the set of feature vectors, by calculating the similarity between the new text and the preset coding corpus, achieve automatic classification of the text, output the automatic classification result, and assign the text to the corresponding three-level coding category. Manually review the automatic classification result, correct the parts with low matching degree or ambiguity in the classification result, and feedback the correction result of the manual review to the generative language model to optimize the parameters of the model and improve the classification performance of the model. Over time, collect new data, and according to these new data and the results of the manual review, dynamically update the coding corpus and the text feature set. Through this iterative process, achieve continuous optimization of the model to better adapt to the changing data and coding requirements.
[0058] In one embodiment, step S20 specifically includes the steps of:
[0059] S21: Extract keywords and perform semantic association analysis on the text after clause segmentation;
[0060] S22: Generate a context representation of the text based on the attention mechanism;
[0061] S23: Combine the pre-trained word embedding technology to generate feature vectors with multi-dimensional semantic information.
[0062] In this embodiment, the keywords are usually nouns, verbs, and adjectives, which can represent the main content and core idea of the text. The attention mechanism is a commonly used technology in neural networks for emphasizing certain important parts in the input sequence. The word embedding technology can be Word2Vec, GloVe, or BERT, etc. These technologies can map words to vectors in a high-dimensional space, where the distance between vectors can reflect the semantic similarity between words.
[0063] Specifically, extract keywords from the text after clause splitting, identify the most important words in the text, conduct semantic association analysis by analyzing the grammatical structure, co-occurrence relationship, and semantic network between words, understand the relationship between keywords and their semantic roles in the sentence, use the attention mechanism to generate a context representation of the text. The attention mechanism can help the model focus on the part of the text most relevant to the current task, thereby generating a representation that can reflect the context of the entire sentence or paragraph. Combine the pre-trained word embedding technology to generate feature vectors with multi-dimensional semantic information. These feature vectors not only contain the semantic information of words but may also contain the grammatical information and context information of words, thus being able to represent the text content more comprehensively.
[0064] In one embodiment, step S30 specifically includes the steps:
[0065] S31: Calculate the similarity between the new text and the open coding corpus through the cosine similarity algorithm to determine the initial coding category;
[0066] S32: Based on the initial coding category, combine the semantic clustering algorithm to further classify the new text into the axial coding;
[0067] S33: Through dependency syntactic analysis and topic modeling, map the axial coding result to the selective coding category and output the automatic classification result.
[0068] In this embodiment, the cosine similarity algorithm is a commonly used method for measuring text similarity. It determines the similarity between two vectors by measuring the degree of similarity in their directions; the semantic clustering algorithm can group texts into different categories according to their semantic content; the axial coding is a further abstraction and induction of the results of open coding, aiming to find the main themes and patterns in the data; the dependency syntactic analysis is used to analyze the dependency relationships between words in a sentence, which helps to understand the structure and deep meaning of the text. Topic modeling is an unsupervised learning method that can identify the topic distribution in a large-scale text collection; selective coding is a process of further refining and selecting core categories on the basis of axial coding to form a theoretical framework.
[0069] Specifically, the cosine similarity between the feature vector of the new text and the center vectors of each category in the open coding corpus is calculated by the cosine similarity algorithm. According to the similarity score, the initial coding category of the new text is determined, and the category with the highest similarity will be regarded as the initial classification of the new text. Using the initial coding category determined in step S31, the semantic clustering algorithm is used to further analyze the new text and classify it into a more refined axial coding category. The results of the axial coding are mapped to the selective coding category through dependency syntactic analysis and topic modeling, and the final automatic classification result is output, which classifies the new text into the existing coding categories or creates new coding categories as needed.
[0070] In one embodiment, step S50 specifically includes the steps of:
[0071] S51: Based on the incremental learning algorithm, the new data is input into the generative language model for incremental training;
[0072] S52: Through the adaptive weight adjustment strategy, balance the feature weights of the historical data and the new data;
[0073] S53: Regularly perform redundancy removal processing on the coding corpus and retain the feature entries with high confidence.
[0074] In this embodiment, the incremental learning algorithm allows the model to be updated when new data is received, rather than retraining from scratch; the new data is input into the generative language model, and this data can be new text samples or the latest coding information.
[0075] Specifically, based on the incremental learning algorithm, new data is input into the generative language model, and the model will use this new data to perform incremental training on itself to adapt to the changes in the data distribution and improve the processing ability for new data. Since historical data may still contain valuable information, while new data may reflect the latest trends or changes, different emphases need to be given. The adaptive weight adjustment strategy is used to dynamically adjust the feature weights of historical data and new data during the model training process. Through this strategy, the model can better balance the importance of old and new data, avoiding overfitting to historical data or ignoring key information in new data. As time goes by, the encoded corpus may accumulate a large amount of data, which may contain duplicate or similar feature entries. During the redundancy removal process, high-confidence feature entries, that is, those considered reliable and important in the model, can be retained, which can be achieved by setting a confidence threshold, and only feature entries exceeding this threshold will be retained.
[0076] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution is prior or subsequent. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of this application.
[0077] In one embodiment, a grounded theory coding optimization system assisted by a generative language model is provided, and this system corresponds one-to-one with a grounded theory coding optimization method assisted by a generative language model in the above embodiment. As Figure 4 shown, this system includes:
[0078] A data preprocessing module, configured to receive the original text data and convert the original text data into an initial corpus set after sentence segmentation through a transcription and sentence segmentation tool;
[0079] A feature extraction module, configured to perform semantic analysis on the initial corpus set by using a pre-trained generative language model, extract text features, and generate a set of feature vectors;
[0080] An intelligent three-level coding classification module, configured to, based on the set of feature vectors, automatically classify the new text into the corresponding three-level coding category by calculating the similarity between the new text and the preset open coding, axial coding, and selective coding corpora, and output an automatic classification result;
[0081] An artificial review and adjustment module, configured to perform artificial review on the automatic classification result, correct low-matching or ambiguous coding items, and feedback the correction result to the generative language model to optimize the model parameters;
[0082] A feature set iterative optimization module, configured to dynamically update the coding corpus and the text feature set according to the new data and the artificial review result, and achieve continuous optimization of the model.
[0083] For the specific limitations of a generative language model-assisted grounded theory coding optimization system, reference can be made to the limitations of a generative language model-assisted grounded theory coding optimization method in the foregoing text, which will not be elaborated herein. Each module in the above-mentioned generative language model-assisted grounded theory coding optimization system can be implemented in whole or in part by software, hardware, and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory in the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to the above-mentioned modules.
[0084] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 5 shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it realizes a generative language model-assisted grounded theory coding optimization method.
[0085] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it realizes the steps of a generative language model-assisted grounded theory coding optimization method.
[0086] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, it realizes the steps of a generative language model-assisted grounded theory coding optimization method.
[0087] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.
[0088] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0089] The above embodiments are only used to illustrate the technical solutions of the present application, not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.
Claims
1. A grounded theory coding optimization method assisted by a generative language model, characterized in that Including the steps: Receiving the original text data, and converting the original text data into an initial corpus set after clause segmentation through a transcription and clause segmentation tool; Performing semantic analysis on the initial corpus set by using a pre-trained generative language model, extracting text features, and generating a set of feature vectors; Based on the set of feature vectors, by calculating the similarity between the new text and the preset open coding, axial coding, and selective coding corpora, automatically classifying the new text into the corresponding third-level coding category, and outputting an automatic classification result; Performing manual review on the automatic classification result, correcting the coding items with low matching degree or ambiguity, and feeding back the correction result to the generative language model to optimize the model parameters; According to the new data and the manual review result, dynamically updating the coding corpus and the text feature set to achieve continuous optimization of the model.
2. The grounded theory coding optimization method assisted by a generative language model according to claim 1, wherein The step of performing semantic analysis on the initial corpus set by using a pre-trained generative language model, extracting text features, and generating a set of feature vectors specifically includes the steps of: extracting keywords and performing semantic association analysis on the text after clause segmentation; Generating a context representation of the text based on the attention mechanism; Combining the pre-trained word embedding technology to generate feature vectors with multi-dimensional semantic information.
3. The grounded theory coding optimization method assisted by a generative language model according to claim 1, wherein The step of, based on the set of feature vectors, by calculating the similarity between the new text and the preset open coding, axial coding, and selective coding corpora, automatically classifying the new text into the corresponding third-level coding category, and outputting an automatic classification result specifically includes the steps of: Calculating the similarity between the new text and the open coding corpus through the cosine similarity algorithm to determine the initial coding category; Based on the initial coding category, combining with the semantic clustering algorithm, further classifying the new text into the axial coding; Mapping the axial coding result to the selective coding category through dependency syntax analysis and topic modeling, and outputting an automatic classification result.
4. A grounded theory coding optimization method assisted by a generative language model according to claim 1, characterized in that, The step of, according to the new data and the manual review result, dynamically updating the coding corpus and the text feature set to achieve continuous optimization of the model specifically includes the steps of: Based on the incremental learning algorithm, inputting the new data into the generative language model for incremental training; Balancing the feature weights of historical data and new data through an adaptive weight adjustment strategy; Regularly performing redundancy removal processing on the coding corpus, and retaining the feature entries with high confidence.
5. A grounded theory coding optimization system assisted by a generative language model, characterized in that Including: A data preprocessing module, configured to receive the original text data, and convert the original text data into an initial corpus set after clause segmentation through a transcription and clause segmentation tool; A feature extraction module, configured to perform semantic analysis on the initial corpus set by using a pre-trained generative language model, extract text features, and generate a set of feature vectors; An intelligent third-level coding classification module, configured to, based on the set of feature vectors, by calculating the similarity between the new text and the preset open coding, axial coding, and selective coding corpora, automatically classify the new text into the corresponding third-level coding category, and output an automatic classification result; A manual review and adjustment module, configured to perform manual review on the automatic classification result, correct the coding items with low matching degree or ambiguity, and feed back the correction result to the generative language model to optimize the model parameters; A feature set iterative optimization module, which is used to dynamically update the coding corpus and text feature set according to new data and manual review results, so as to realize the continuous optimization of the model.
6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of a grounded theory coding optimization method assisted by a generative language model according to any one of claims 1 to 4.
7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of a grounded theory coding optimization method assisted by a generative language model according to any one of claims 1 to 4.