Search

CN-121980385-A - Method, device, equipment, medium and product for measuring enterprise digital transformation degree

CN121980385ACN 121980385 ACN121980385 ACN 121980385ACN-121980385-A

Abstract

The application discloses a method, a device, equipment, a medium and a product for measuring the digital transformation degree of enterprises, which relate to the fields of enterprise information technology and data analysis and comprise the steps of dividing enterprise annual reports of each enterprise into a plurality of chapters; extracting a part of chapters, manually marking the grade value of the enterprise digital transformation index to be evaluated corresponding to each chapter to obtain a marked data set, determining an optimal task execution strategy and an optimal knowledge excitation mode according to the marked data set, generating an instruction prompt word according to the optimal task execution strategy and the optimal knowledge excitation mode, inputting the instruction prompt word and the rest chapters of the enterprise to be measured into a large language model to obtain the grade value of the enterprise digital transformation index to be evaluated corresponding to each chapter, and finishing the measurement of the enterprise digital transformation degree to be measured.

Inventors

  • ZHOU ZHENKUN
  • HE YIHAN
  • REN TAO
  • YU MENGLI

Assignees

  • 首都经济贸易大学

Dates

Publication Date
20260505
Application Date
20260123

Claims (10)

  1. 1. A method for measuring the degree of transformation of an enterprise digital transformation, characterized in that the method for measuring the degree of transformation of the enterprise digital transformation comprises the following steps: Acquiring enterprise annual reports of a plurality of enterprises, wherein the enterprises comprise enterprises to be measured; Dividing enterprise annual reports of each enterprise into a plurality of chapters to obtain a chapter set of each enterprise; extracting a part of chapters from the chapter set of each enterprise, and manually marking the grade value of the digitized transformation index of each enterprise to be evaluated corresponding to each extracted chapter to obtain a marked data set, wherein the marked data set comprises each extracted chapter and the grade value of each digitized transformation index of each enterprise to be evaluated corresponding to each extracted chapter which is manually marked; Determining an optimal task execution strategy and an optimal knowledge excitation mode according to the annotation data set, wherein the task execution strategy is an output mode of a large language model, and the knowledge excitation mode is a generation mode of instruction type prompt words; Generating instruction type prompt words according to an optimal knowledge excitation mode, wherein the instruction type prompt words comprise digitized transformation indexes of enterprises to be evaluated and an optimal task execution strategy; Inputting the instruction prompt words and the chapters remained after the extraction of the enterprise to be measured into a large language model to obtain the grade value of the digitized transformation index of each enterprise to be evaluated corresponding to each remaining chapter, and finishing the digitized transformation degree measurement of the enterprise to be measured.
  2. 2. The method for measuring the digitized transformation degree of an enterprise according to claim 1, wherein the determining an optimal task execution strategy and an optimal knowledge excitation mode according to the labeling data set specifically comprises: dividing the labeling data set into a training set, a verification set and a test set; executing task execution strategies of outputting the single indexes one by one to obtain prompt words of the digitized transformation indexes of each enterprise to be evaluated; inputting prompt words of the digitized transformation indexes of the enterprises to be evaluated and chapters in the test set into a large language model to obtain first prediction grade values of the digitized transformation indexes of the enterprises to be evaluated corresponding to the chapters in the test set; Determining an evaluation result corresponding to a task execution strategy which is output by the single index one by one according to a first prediction grade value of each enterprise digital transformation index to be evaluated corresponding to each chapter in the test set and a grade value of each enterprise digital transformation index to be evaluated corresponding to each chapter manually marked in the test set; executing a task execution strategy of multi-index synchronous output to obtain an overall prompt word, wherein the overall prompt word comprises all digitized transformation indexes of enterprises to be evaluated; Inputting the whole prompt words and each chapter in the test set into a large language model to obtain a second prediction grade value of the digitalized transformation index of each enterprise to be evaluated corresponding to each chapter in the test set; Determining an evaluation result corresponding to a task execution strategy synchronously output by multiple indexes according to a second prediction grade value of each enterprise digital transformation index to be evaluated corresponding to each chapter in the test set and a grade value of each enterprise digital transformation index to be evaluated corresponding to each chapter manually marked in the test set; Determining an optimal task execution strategy according to the evaluation results corresponding to the task execution strategies output one by the single index and the evaluation results corresponding to the task execution strategies synchronously output by the multiple indexes; executing a knowledge excitation mode of a direct prediction strategy on the basis of an optimal task execution strategy to obtain a direct prediction prompt word; Inputting the direct prediction prompt word into a large language model to obtain a third prediction grade value of the digitized transformation index of each enterprise to be evaluated corresponding to each chapter in the test set; determining an evaluation result corresponding to a knowledge excitation mode of a direct prediction strategy according to a third prediction grade value of each enterprise digital transformation index to be evaluated corresponding to each chapter in the test set and a grade value of each enterprise digital transformation index to be evaluated corresponding to each chapter manually marked in the test set; executing a knowledge excitation mode of a thinking chain excitation strategy on the basis of an optimal task execution strategy to obtain a thinking chain prompt word; Inputting the thinking chain prompt words into a large language model to obtain a fourth prediction grade value of the digitalized transformation index of each enterprise to be evaluated corresponding to each chapter in the test set; Determining an evaluation result corresponding to a knowledge excitation mode of the thinking chain excitation strategy according to a fourth prediction grade value of each enterprise digital transformation index to be evaluated corresponding to each chapter in the test set and a grade value of each enterprise digital transformation index to be evaluated corresponding to each chapter manually marked in the test set; And determining an optimal knowledge excitation mode according to the evaluation result corresponding to the knowledge excitation mode of the direct prediction strategy and the evaluation result corresponding to the knowledge excitation mode of the thinking chain excitation strategy.
  3. 3. The method for measuring the digitized transformation degree of the enterprise according to claim 2, wherein before inputting the instruction prompt words and the chapters remained after the enterprise to be measured is extracted into the large language model, the method further comprises the steps of fine tuning the large language model on a training set, evaluating the fine-tuned large language model on a verification set and testing the evaluated large language model on a test set by utilizing a low-rank self-adaption technology on the basis of an optimal task execution strategy and an optimal knowledge excitation mode.
  4. 4. The method of claim 1, further comprising preprocessing each chapter before extracting a portion of the chapters.
  5. 5. The method for measuring the degree of transformation of the enterprise digital transformation according to claim 1, wherein the class value of each enterprise digital transformation index to be evaluated corresponding to each remaining chapter outputted by the large language model is a JSON-format file, and the method for measuring the degree of transformation of the enterprise digital transformation further comprises the step of performing format conversion on the JSON-format file outputted by the large language model.
  6. 6. The method for measuring the digitized transformation degree of enterprises according to claim 1, wherein the annual newspaper of the enterprises is in PDF format, and each enterprise annual newspaper is divided into a plurality of chapters to obtain a chapter set of each enterprise, specifically comprising: for any enterprise, obtaining a logic starting page number of each chapter according to a directory page of an enterprise annual report of the enterprise; acquiring an actual initial page number of a first chapter; Obtaining a page offset according to the actual start page number of the first chapter and the logic start page number of the first chapter; obtaining the actual initial page numbers of other chapters according to the page number offset and the logic initial page numbers of other chapters, wherein the other chapters are all chapters except the first chapter; Obtaining the actual end page numbers of all chapters according to the actual start page numbers of all chapters; And dividing the enterprise annual report of the enterprise into a plurality of chapters according to the actual start page numbers and the actual end page numbers of all chapters to obtain a chapter set of the enterprise.
  7. 7. An enterprise digital transformation level measurement apparatus, the enterprise digital transformation level measurement apparatus comprising: The system comprises a data preprocessing module, a data processing module and a data processing module, wherein the data preprocessing module is used for acquiring enterprise annual messages of a plurality of enterprises, the enterprises comprise enterprises to be measured, and the enterprise annual messages of each enterprise are divided into a plurality of chapters to obtain a chapter set of each enterprise; The system comprises a large model analysis module, a labeling data set, an instruction type prompt word and a large language model, wherein the large model analysis module is used for extracting a part of chapters in a chapter set of each enterprise, manually labeling the grade values of the digitized transformation indexes of the enterprise to be evaluated corresponding to the extracted chapters to obtain the labeling data set, determining an optimal task execution strategy and an optimal knowledge excitation mode according to the labeling data set, the labeling data set comprises the grade values of the digitized transformation indexes of each enterprise to be evaluated corresponding to the extracted chapters and the manually labeled extracted chapters, the knowledge excitation mode is a generation mode of an instruction type prompt word, the task execution strategy is an output mode of a large language model, the instruction type prompt word is generated according to the optimal knowledge excitation mode, the instruction type prompt word comprises the digitized transformation indexes of each enterprise to be evaluated and the optimal task execution strategy, and the grade values of the digitized transformation indexes of each remaining enterprise to be measured after the extraction of each enterprise to be measured are input into the large language model to obtain the grade values of the digitized transformation indexes of each enterprise to be measured corresponding to be measured.
  8. 8. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the enterprise digital transformation level measurement method of any one of claims 1-6.
  9. 9. A computer readable storage medium having stored thereon a computer program, which when executed by a processor implements the enterprise digital transformation level measurement method of any one of claims 1-6.
  10. 10. A computer program product comprising a computer program which, when executed by a processor, implements the enterprise digital transformation level measurement method of any one of claims 1-6.

Description

Method, device, equipment, medium and product for measuring enterprise digital transformation degree Technical Field The application relates to the field of enterprise information technology and data analysis, in particular to an enterprise digital transformation degree measuring method, device, equipment, medium and product. Background With the digital technology becoming a core driving force for enterprise development, accurate and effective measurement of the enterprise digital transformation process has become an important topic in commercial practice. The enterprise digital transformation measurement result plays an indispensable important role in the aspects of enterprise strategic decision basis, investor investment decision reference, industry competition analysis and the like. The decision-making basis of the enterprise strategy is that an enterprise manager needs to clearly know the transformation stage, advantages and defects of the enterprise manager when making a digital transformation strategy. Accurate quantitative evaluation can provide detailed diagnosis report for enterprises, help the enterprises to definitely transform directions, reasonably allocate resources, formulate targeted transformation strategies, avoid blind investment and decision errors, and improve transformation success rate. The investment decision reference is that for investors, the digitized transformation degree of enterprises is a key index for evaluating future development potential and investment value of enterprises. Accurate quantitative evaluation results can help investors to screen enterprises with high growth and competitiveness, and make intelligent investment decisions. And the quantitative evaluation is beneficial to the transverse comparison of enterprises in industry and is used for knowing the status and the competitiveness of the enterprises in the industry. Enterprises can find gaps by comparing with digital transformation indexes of excellent enterprises in the same industry, learn to reference advanced experiences, and formulate differential competition strategies so as to improve the competition advantages in the industry. At present, the prior art scheme for measuring the digital transformation degree of enterprises mainly focuses on the following categories, and has certain limitations (1) a measurement method based on a single index or a comprehensive evaluation index. For example, the digitization level is indirectly reflected by calculating the proportion of the enterprise digital assets to the total assets, or a multi-dimensional index system is constructed and weights are manually set for comprehensive scoring. The method has the advantages that the data is easy to obtain, but the subjectivity of index selection and weight setting is strong. (2) Text analysis based on word frequency. The word frequency in public documents such as annual newspaper of enterprises is statistically analyzed to measure the transformation degree by constructing a digital technical keyword dictionary. This is the dominant method in current demonstration research. However, the method has obvious technical bottlenecks that firstly, a preset keyword dictionary is seriously relied on, semantic recognition deviation and missed judgment are caused by incapability of recognizing expressions which are not in the dictionary but are related to semantics, secondly, the method can only judge the existence of keywords, cannot understand the context, is easy to generate misjudgment due to fragmentation of texts (such as 'we do not greatly put into cloud computing'), and thirdly, the method is essentially binary judgment of 'existence or non-existence', and cannot carry out grading quantization on the application depth and effect of the technology. (3) Sentence classification based on a supervised machine learning model. To overcome the limitations of traditional text analysis, there have been recent studies attempting to use pre-trained models for supervised sentence classification. The method improves semantic understanding accuracy, but the model training and recognition process is not completely free from the path dependence on a dictionary label system. More importantly, the method is still essentially framed as a two-class task, whose output is limited to determining whether a technology is "mentioned" and not how "applied" and what effect "it produces" in the enterprise. In summary, the prior art scheme solves the core problem of enterprise digital transformation degree measurement, and faces the key technical bottleneck that 1, semantic understanding is shallow, the traditional method relies on surface keyword matching, and reasoning capability on text deep semantic and context logic is lacking, so that accuracy and robustness of recognition results are insufficient. 2. The measurement dimension singleization is that the existing method generally only realizes binary judgment of yes and no, does not carry out continuous an