Method and device for extracting report key information and generating abstract
By splitting multi-level headings in long text reports and analyzing them using a pre-trained large language model, the accuracy issues of key information extraction and summary generation in long text reports in existing technologies have been resolved, achieving efficient and low-cost information extraction and generation.
Patent Information
- Application Number
- CN202511706644.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-20
- Publication Date
- 2026-02-13
AI Technical Summary
Existing technologies struggle to effectively handle the complex semantic text in long reports, and deep learning-based methods suffer from content forgetting and model illusion, leading to inaccurate key information extraction and summary generation.
By employing a pre-trained large language model combined with multi-level heading segmentation technology, long text reports are formatted and segmented. Through regular expression matching and large language model analysis, key information is located and extracted, and a summary in a specified format is generated.
It enables accurate extraction of key information and generation of summaries from long text reports, reducing manual costs and improving the efficiency and accuracy of information extraction.
Smart Images
Figure CN121524342A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of file information extraction and abstract generation, in particular to a report key information extraction and abstract generation method and device, a storage medium and electronic equipment. BACKGROUND
[0002] At present, file information extraction and abstract generation mainly focuses on literature search, financial report analysis, policy interpretation, engineering design verification and the like, and information extraction and abstract generation technologies include rule matching, sequence labeling, deep learning and artificial processing methods. Among them, rule matching and sequence labeling and the like are based on fixed mode search, and it is difficult to deeply analyze files containing complex semantic texts; the artificial processing method requires personnel to have a high professional background, and the cost is high; the artificial intelligence technology based on deep learning is difficult to process long text report files, and there are problems such as content forgetting and model illusion.
[0003] Therefore, there is an urgent need to design a report key information extraction and abstract generation method, device, storage medium and electronic equipment, which fully utilizes the template framework of long text reports, accurately extracts key information content and generates abstracts according to specified formats. SUMMARY
[0004] The purpose of the present application is to overcome the shortcomings of the prior art, and provide a report key information extraction and abstract generation method, device, storage medium and electronic equipment for key information extraction and abstract generation of long text reports using fixed template formats.
[0005] The present application is realized by the following technical solutions: The application discloses a method for extracting and summarizing key information in a report, which comprises the following steps: S1. report format conversion, converting a report file with a fixed template into a source file; S2. initially setting N=1, M=I[i], i=1, wherein N is the current title level to be matched, M is the current key information content to be queried, I is a list of all key information, and i is the list index; S3. matching the N-level title of the text, splitting the source file into multiple N-level subfiles; S4. using a pre-trained large language model to analyze the content of the multiple N-level subfiles, and locating the N-level subfile containing the key information M; S5. for the N-level subfile containing the key information M, setting N=N+1, and continuing to repeat the steps S3-S4 until N exceeds a specified maximum value Nmax; S6. using the pre-trained large language model to extract the key information M from the final Nmax-level subfile containing the key information M, and outputting the extracted key information M in a specified format and saving in a draft file; S7. setting i=i+1 and N=1, and continuing to repeat the steps S3-S6 until all the key information in the list I is traversed; and S8. using the pre-trained large language model to analyze the content of the draft file, and generating a final summary in a specified format.
[0006] Preferably, the application further comprises that the source file format in step S1 is markdown, HTML, TXT, RST or other common text-based structured formats, which are used to clearly show the title hierarchy of the source file.
[0007] Preferably, the application further comprises that the text matching in step S3 comprises setting a regular matching rule for the title of a specified level, which is used for regular search on the full text of the file.
[0008] Preferably, the application further comprises that the file splitting in step S3 comprises performing regular matching on the title of the source file content, inserting a file splitting point before the matched text, and splitting the source file into multiple subfiles with non-overlapping contents according to the file splitting point.
[0009] Preferably, the application further comprises that the pre-trained large language model is used to analyze the content of the subfile in step S4, and the content comprises inputting the content of the corresponding subfile into the large language model, and judging whether the subfile contains the key information M through a proper prompt word.
[0010] Preferably, the application further comprises that the N-level subfile containing the key information M is located in step S4, and the implementation manner comprises but is not limited to recording the path information of the subfile containing the key information in the memory or the hard disk, writing the path information of the subfile containing the key information and the judgment result into a self-defined data table, and adding a specific marker to the name of the subfile containing the key information, so as to determine the subfile containing the key information from the multiple subfiles.
[0011] Preferably, the present application further comprises: the draft file described in step S6, the format of which includes but is not limited to: text file format, markdown format, HTML file format, etc., and the purpose is to save one or more key information after the format is regularized.
[0012] Preferably, the present application further comprises: the final abstract described in step S8, the content of which includes but is not limited to: summary of all key information in the draft file, computational analysis of all key information in the draft file, logical inference of all key information in the draft file, and the purpose is to provide an abstract that meets the application scenario.
[0013] The present application also provides a device for executing the report key information extraction and abstract generation method, which is used to execute the report key information extraction and abstract generation method, and includes the following modules: a format conversion module for converting a report file with a fixed template into a source file; a file splitting module for inserting a file splitting point before the searched matching text by performing N-level title regular matching on the content of the source file, and splitting the source file into multiple content non-overlapping sub-files according to the file splitting point; a file analysis module for inputting the content of the corresponding sub-file into a large language model, and judging whether the sub-file contains the key information through appropriate prompt words, if the sub-file contains the key information, positioning the current sub-file; the positioning method includes but is not limited to: recording the path and file name information of the sub-file in the memory or hard disk, writing the sub-file judgment result into a custom data table, adding a specific marker to the sub-file name, etc., and the purpose is to determine the sub-file containing the key information in multiple sub-files; an analysis loop module for repeatedly executing the file splitting and file analysis steps until N exceeds the specified maximum value Nmax; an information extraction module for extracting the key information existing in the given file by using the pre-trained large language model, and outputting to the draft file in the specified format; an intermediate storage module for maintaining the draft file, which is used to summarize various key information extracted by the pre-trained large language model; an extraction loop module for repeatedly executing the key information extraction step until all key information is extracted; and an abstract generation module for analyzing the content of the draft file by using the pre-trained large language model, generating a final abstract in a specified format, and the final abstract content includes but is not limited to: summary of all key information in the draft file, computational analysis of all key information in the draft file, logical inference of all key information in the draft file, etc.
[0014] The present application also provides a storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the report key information extraction and abstract generation method.
[0015] The application further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory, wherein the processor implements the method for reporting key information extraction and summary generation when executing the computer program.
[0016] Compared with the prior art, the application has the following advantages and beneficial effects: For long text reports with fixed templates, the report template architecture is fully utilized, the report file is split according to multiple levels of titles, a pre-trained large language model is used for key information positioning, the effective context containing key information can be shortened to a small scale, the pre-trained large language model is applied to the effective context to extract key information, and multiple key information can be summarized by repeating the above steps, and finally the pre-trained large language model is used to generate an accurate summary. BRIEF DESCRIPTION OF DRAWINGS
[0017] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the application, and therefore should not be considered as a limitation to the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor.
[0018] Figure 1 is a flowchart of a method for reporting key information extraction and summary generation based on a fixed template in an embodiment of the application.
[0019] Figure 2 is a structural diagram of a device for reporting key information extraction and summary generation based on a fixed template in an embodiment of the application.
[0020] Figure 3 is a structural diagram of an electronic device for reporting key information extraction and summary generation based on a fixed template in an embodiment of the application.
[0021] Figure 4 is a schematic diagram of a report content architecture in an embodiment of the application.
[0022] Figure 5 is a schematic diagram of a required output summary template in an embodiment of the application.
[0023] Figure 6 is a schematic diagram of a 1-level sub-file split in an embodiment of the application.
[0024] Figure 7 is a schematic diagram of a 2-level sub-file split of a specific content (protection) chapter in an embodiment of the application.
[0025] Figure 8 Figure 2 is a content schematic diagram of a draft file after the end of the first round of analysis in the embodiment of the present application.
[0026] Figure 9 Figure 3 is a 2-level sub-file splitting schematic diagram of a specific content (dispatching) chapter in the embodiment of the present application.
[0027] Figure 10 Figure 4 is a content schematic diagram of a draft file after the end of the second round of analysis in the embodiment of the present application.
[0028] Figure 11 Figure 5 is a 2-level sub-file splitting schematic diagram of a specific content (communication) chapter in the embodiment of the present application.
[0029] Figure 12 Figure 6 is a content schematic diagram of a draft file after the end of the third round of analysis in the embodiment of the present application.
[0030] Figure 13 Figure 7 is a content schematic diagram of a final abstract file in the embodiment of the present application. DETAILED DESCRIPTION
[0031] In order to make the purpose, technical scheme and advantages of the present application clearer, further detailed description of the present application will be given below in combination with embodiments and drawings. The illustrative embodiments of the present application and their descriptions are only used to explain the present application and do not limit the present application. EMBODIMENT
[0032] Figure 1 Figure 8 is a flow schematic diagram of a method for extracting and generating abstracts of report key information in the embodiment of the present application. As shown in Figure 8, the method includes the following contents. Figure 1
[0033] Step S1, report format conversion, converting a report file with a fixed template into a source file; Specifically, after a user submits a report with a fixed template format, the embodiment of the present application can use OCR, format conversion tools and other means to convert the report into a markdown format source file.
[0034] Figure 4 Figure 9 shows a content schematic diagram of converting a booster station access scheme design report into a markdown format source file, wherein the file contains five first-level titles: “system primary”, “system relay protection and safety automatic device”, “dispatching automation”, “system communication” and “appendix”; and each first-level title content can contain multiple second-level titles and related contents.
[0035] Figure 5 A kind of report content is shown for booster station access scheme design, according to the summary template content schematic diagram generated by the power secondary review expert required, mainly divided into "line protection" "dispatching management" "optical cable construction" three aspects of project construction content. Among them, symbol " [...]" represents the key content needed to be refined from original report.
[0036] Step S2, initially N=1, M=I[i], i=1 is ordered; Specifically, the present application embodiment can create and maintain N, M, i and the like variables in memory or hard disk.
[0037] Further, in the present embodiment, the key information to be extracted is saved in list I, wherein I={“How is the configuration of the line protection of the project?”, “How is the organization of the remote information of the project?”, “How is the optical cable construction scheme of the project?”}.
[0038] Further, in the present embodiment, the second title is set as the maximum matching level of the title, i.e. Nmax=2.
[0039] Step S3, match the N-level title of the text, and split the source file into multiple N-level subfiles; Specifically, Figure 6 A regular expression like "^#\s\d+\s[\u4e00-\u9fa5]+ " is used to match the first-level title of the file content shown in Figure 4 The file splitting method after matching is shown in the figure. Among them, the original text is split into "1-level file 01.md" "1-level file 02.md" "1-level file 03.md" "1-level file 04.md" "1-level file 05.md" and the like 5 1-level subfiles.
[0040] Step S4, using a pre-trained large language model, analyzing the content of the multiple N-level subfiles, positioning the N-level subfile containing the key information M; Specifically, for the present embodiment, when the program is executed for the first time to this step, M=I[1]="How is the configuration of the line protection of the project?"; The above 5 1-level subfiles are sequentially input as context into the pre-trained large language model (such as deepseek-32b), and the prompt word "Does the current subfile content contain the content of the key information M? Please answer 'yes' or 'no'"; Retain the 1-level subfile whose model answer is "yes", and delete the other 1-level subfiles.
[0041] For the 5 1-level subfiles shown in Figure 6 Since only "1-level file 02.md" involves the related content of line protection, only this 1-level subfile is retained.
[0042] Step S5, for the N-level sub-file containing the key information M, let N=N+1, continue to repeat steps S3-S4 until N exceeds the specified maximum value Nmax; Specifically, for this embodiment, N=1+1=2, the program matches and splits the second-level title in the "1-level file 02.md" file using the regular expression "^##\s\d+\.\d+\s[\u4e00-\u9fa5]+", forming 3 second-level sub-files: "2-level sub-file 0201.md", "2-level sub-file 0202.md", and "2-level sub-file 0203.md", as shown in Figure 7 .
[0043] Further, this embodiment inputs the above 3 second-level sub-files into the pre-trained large language model (such as deepseek-32b) as context in turn, and inputs the prompt word "Does the current sub-file content contain the content of the key information M? Please answer 'yes' or 'no'"; keeps the second-level sub-file whose model answer is "yes", and deletes the other second-level sub-files. Since only "2-level file 0202.md" involves related content of line protection, only this second-level sub-file is kept.
[0044] Further, N=2+1=3, at this time N exceeds the specified maximum value Nmax, so the loop is stopped.
[0045] Step S6, using a pre-trained large language model, extracting key information from the final Nmax-level sub-file containing the key information M, and outputting the extracted key information M in the specified format and saving it in a draft file; Specifically, this embodiment extracts the key information M from the "2-level sub-file 0202.md" file, inputs the content of the "2-level file 0202.md" file into the pre-trained large language model (such as deepseek-32b) as context, and inputs the prompt word "Please extract the key information M from the context and arrange the extracted information into a sentence to form the output result of 'System relay protection: …' ", and saves the output result of the above model in a draft file, as shown in Figure 8 .
[0046] Step S7, let i=i+1, N=1, continue to repeat steps S3-S6 until all key information in the list I is traversed; Specifically, for this embodiment, i=1+1=2, N=1, continue to repeat steps S3-S6, and the intermediate results are as follows: Repeat step S3 to split the source file into 6 first-level sub-files as shown in Figure 6 . Repeat the step S4, locate the 1st sub-file containing the key information M=I[2]="How is the organization of the remote information of this project?", keep the 1st sub-file, and delete other 1st sub-files. Specifically, the 1st sub-file "1st sub-file 03.md" is kept, and other 1st sub-files are deleted; Repeat the step S5, split the "1st sub-file 03.md" into 3 2nd sub-files as shown in Figure 9 ; locate and keep the 2nd sub-file containing the key information M, and delete other 2nd sub-files. Specifically, the 2nd sub-file "2nd sub-file 0302.md" is kept, and other 2nd sub-files are deleted; Repeat the step S6, extract the key information M from the "2nd sub-file 0302.md" file, and arrange the extracted information into a sentence to form the output result "Dispatching management settings: ……"; add the model output result to the draft file as shown in Figure 10 .
[0047] Repeat the step S7, for this embodiment, i=2+1=3, N=1, continue to repeat the steps S3-S6, and the intermediate results are as follows: Repeat the step S3, split the source file into 6 1st sub-files as shown in Figure 6 ; Repeat the step S4, locate the 1st sub-file containing the key information M=I[3]="How is the cable construction scheme of this project?", keep the 1st sub-file, and delete other 1st sub-files. Specifically, the 1st sub-file "1st sub-file 04.md" is kept, and other 1st sub-files are deleted; Repeat the step S5, split the "1st sub-file 04.md" into 3 2nd sub-files as shown in Figure 11 ; locate and keep the 2nd sub-file containing the key information M, and delete other 2nd sub-files. Specifically, the 2nd sub-file "2nd sub-file 0402.md" is kept, and other 2nd sub-files are deleted; Repeat the step S6, extract the key information M from the "2nd sub-file 0402.md" file, and arrange the extracted information into a sentence to form the output result "System communication design: ……"; add the model output result to the draft file as shown in Figure 12 .
[0048] Repeat the step S7, since all information in the list I has been traversed, stop the loop; Step S8, use a pre-trained large language model to analyze the content of the draft file, and generate a final summary according to a specified format; Specifically, this embodiment will Figure 12The draft file content shown is used as context, a pre-trained large language model (such as deepseek-32b) is input, and the prompt word "Please carefully understand the context content, and output the final summary according to the'summary template.md' file format, in the'summary template.md', the symbol '…' represents the key information that needs to be filled in." is input. Finally, the model output is saved in the file 'final summary.md', as shown in Figure 13 .
[0049] The embodiment also proposes a device that can be used to execute the method embodiment. For details not disclosed in the device embodiment of the present application, please refer to the method embodiment of the present application.
[0050] Figure 2 is a structural schematic diagram of a device for reporting key information extraction and summary generation in the embodiment of the present application. As shown in Figure 2 , the device comprises: a format conversion module 91 for converting a report file with a fixed template into a source file; a file splitting module 92 for inserting a file splitting point before the searched matching text by performing N-level title regular matching on the source file content, and splitting the source file into multiple content non-overlapping sub-files according to the file splitting point; a file analysis module 93 for inputting the corresponding sub-file content into a large language model, and judging whether the sub-file contains the key information through a suitable prompt word. If the sub-file contains the key information, the current sub-file is positioned. The positioning methods include but are not limited to: recording the path and file name information of the sub-file in the memory or hard disk, writing the sub-file judgment result into a custom data table, adding a specific marker to the sub-file name, etc., the purpose is to determine the sub-file containing the key information among multiple sub-files; a first loop execution module 94 for repeatedly executing the file splitting and file analysis steps until N exceeds the specified maximum value Nmax; an information extraction module 95 for extracting the key information existing in the given file by using the pre-trained large language model, and outputting to the draft file in the specified format; an intermediate storage module 96 for maintaining the draft file, which is used to summarize various key information extracted by the pre-trained large language model; a second loop execution module 97 for repeatedly executing the key information extraction step until all key information is extracted; a summary generation module 98 for analyzing the draft file content by using the pre-trained large language model, and generating the final summary according to the specified format. The final summary content includes but is not limited to: a summary of all key information in the draft file, a calculation analysis of all key information in the draft file, a logical inference of all key information in the draft file, etc.
[0051] The embodiment also provides an electronic device, which can be used to implement the method embodiment.
[0052] Figure 3 FIG. 1 is a structural schematic diagram of an electronic device for reporting key information extraction and summary generation based on a fixed template according to an embodiment of the present application.
[0053] As shown in Figure 3 The electronic device 1000 includes one or more processors 1010 and memory 1020.
[0054] The processor 1010 can be a central processing unit (CPU) or other form of processing unit having data processing and / or instruction executing capabilities, and can control other components in the electronic device 1000 to perform desired functions.
[0055] The memory 1020 can include one or more computer program products, which can include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory, for example, can include random access memory (RAM), cache memory, and / or the like. The non-volatile memory, for example, can include read-only memory (ROM), hard disk, flash memory, and / or the like. One or more computer program instructions can be stored on the computer-readable storage media, and the processor 1010 can run the program instructions to implement the method for reporting key information extraction and summary generation based on a fixed template according to various embodiments of the present application described above and / or other desired functions. Various contents, such as category correspondence, and the like, can also be stored in the computer-readable storage media.
[0056] In one example, the electronic device 1000 can also include an input device 1030 and an output device 1040, which are interconnected through a bus system and / or other forms of connection mechanism (not shown).
[0057] In addition, the input device 1030 can also include, for example, a keyboard, a mouse, an interface, and the like. The output device 1040 can output various information, including analysis results and the like, to the outside. The output device 1040 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and the like.
[0058] Of course, in order to simplify, Figure 3 Only the components of the electronic device related to the present application are shown in the above
[0059] In addition to the above method and device, the example computer program product and computer readable storage medium, the embodiments of the present application can also be a computer program product, which includes computer program instructions, which when executed by a processor, causes the processor to perform the steps of a fixed template based report key information extraction and summary generation method according to various embodiments of the present application described in the above "example method" section of the present specification.
[0060] The computer program product can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++, etc., and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device or entirely on the remote computing device or server.
[0061] In addition, the embodiments of the present application can also be a computer readable storage medium, which stores computer program instructions, which when executed by a processor, causes the processor to perform the steps of a fixed template based report key information extraction and summary generation method according to various embodiments of the present application described in the above "example method" section of the present specification.
[0062] The computer readable storage medium can be any combination of one or more computer readable media. The computer readable medium can be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can include, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any suitable combination of the above. More specific examples (a non-exhaustive list) of the computer readable storage medium include an electrical connection having one or more wires, a portable disc, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.
[0063] The above description is given for illustrative and descriptive purposes. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.
[0064] The above detailed description of the specific embodiments of the present application has been given to understand the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. A method for reporting key information extraction and summary generation, characterized by, The method comprises the following steps: S1. Report format conversion, converting a report file with a fixed template into a source file; S2. Initially, let N=1, M=I[i], i=1, where N is the current title level to be matched, M is the current query key information content, I is a list of all key information, and i is the list index; S3. Match the N-level title of the text, and split the source file into multiple N-level subfiles; S4. Using a pre-trained large language model, analyze the contents of the multiple N-level subfiles to locate the N-level subfile containing the key information M; S5. For the N-level subfile containing the key information M, let N=N+1, and continue to repeat steps S3-S4 until N exceeds the specified maximum value Nmax; S6. Using a pre-trained large language model, extract the key information M from the final Nmax-level subfile containing the key information M, and output the extracted key information M in the specified format and save it in a draft file; S7. Let i=i+1, N=1, and continue to repeat steps S3-S6 until all key information in the list I is traversed; S8. Using a pre-trained large language model, analyze the contents of the draft file, and generate a final summary in the specified format.
2. The method of claim 1, wherein, The text matching in step S3 includes setting a regular matching rule for the specified level of title and performing text matching based on the regular matching rule.
3. The method for reporting key information extraction and summary generation of claim 1, wherein, The file splitting in step S3 includes performing title regular matching on the entire content of the source file, inserting a file splitting point before the searched matching text, and splitting the source file into multiple subfiles with non-overlapping contents according to the file splitting point.
4. The method for reporting key information extraction and summary generation of claim 1, wherein, The analysis of subfile contents using a pre-trained large language model in step S4 includes sequentially inputting the subfile contents split in step 3 into the pre-trained large language model, and determining whether the subfile contains the key information M through appropriate prompt words.
5. The method for reporting key information extraction and summary generation of claim 1, wherein, The implementation of locating the N-level subfile containing the key information M in step S4 is any one of recording the subfile path information containing the key information M in the memory or hard disk, writing the subfile path information containing the key information into a custom data table, and adding a specific marker to the subfile name containing the key information.
6. The method for reporting key information extraction and summary generation of claim 1, wherein, The draft file in step S6 has any one of the following formats: text file format, markdown format, and HTML file format.
7. The method for reporting key information extraction and summary generation of claim 1, wherein, The final summary in step S8 includes at least one of the following: a summary of all key information in the draft file, a calculation analysis of all key information in the draft file, and a logical inference of all key information in the draft file.
8. An apparatus for extracting key information from reports and generating summaries, characterized in that, The device is used to perform any one of the methods of extracting key information and generating summaries in claims 1-7, and comprises the following modules: A format conversion module for converting a report file with a fixed template into a source file; A file splitting module for performing N-level title regular matching on the contents of the source file, inserting a file splitting point before the searched matching text, and splitting the source file into multiple subfiles with non-overlapping contents according to the file splitting point; File analysis module: for inputting the corresponding sub-file content into a large language model, and judging whether the sub-file contains key information through appropriate prompt words. If it contains key information, the current sub-file is located; The first loop execution module is used for repeatedly executing the file splitting and file analysis steps until N exceeds the specified maximum value Nmax; The information extraction module is used for extracting the key information contained in the located file by using a pre-trained large language model, and outputting to a draft file in a specified format; The intermediate storage module is used to maintain the draft file, which is used to summarize all key information extracted by the large language model; The second loop execution module is used for repeatedly executing the key information extraction step until all key information is extracted; The abstract generation module is used to analyze the content of the draft file by using a large language model, and generate a final abstract in a specified format.
9. A storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to realize the method of any one of claims 1-7.
10. An electronic device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program, when executed by the processor, causes the electronic device to perform the method of any one of claims 1-9. The processor executes the computer program to realize the method of any one of claims 1-7.