Method and computing device for displaying patent text based on large language model output

Abstract

A method and a computing device for displaying patent text based on large language model output, wherein the method is performed by the computing device, and the method comprises receiving patent text output by a large language model, wherein the patent text comprises at least one placeholder, each placeholder comprises a general symbol and symbol information, and the symbol information corresponds to a special symbol, backfilling the special symbol having a mapping relationship with the placeholder at the position of the placeholder in the patent text, and displaying the backfilled patent text. The method can realize accurate display of formulas, chemical equations and other special characters.

Description

Technical Field

[0001] This disclosure relates to a method and computing device for displaying patent text output by a large language model, and more specifically to a method and computing device for displaying patent text based on the output of a large language model. Background Technology

[0002] With the rapid development of artificial intelligence technology, Large Language Models (LLMs) have been widely applied in various fields. In the field of intellectual property, LLMs can assist in generating various patent texts, including patent application documents, statements of opinion, defense suggestions, and technical disclosure materials, thereby improving the efficiency and quality of patent drafting and processing.

[0003] However, patent texts typically contain various special symbols, such as chemical formulas, mathematical formulas, chemical equations, scientific symbols, intellectual property symbols, and physical units. These special symbols are crucial for accurately expressing the technical solution and the scope of the claims. For example, patent applications in the field of chemistry may require demonstrating complex molecular structural formulas and chemical reaction equations, while patent applications in the fields of physics or mathematics may contain a large number of mathematical expressions and scientific symbols.

[0004] When using large language models to generate patent text, the handling of special symbols presents several challenges. On one hand, large language models may have limitations in their output capabilities, making it difficult to directly output certain special symbols. On the other hand, web pages or other display interfaces may encounter difficulties rendering certain special characters, leading to errors, garbled text, or incorrect display in the output. These issues can affect the accuracy and readability of the patent text.

[0005] Furthermore, different terminal devices and display environments have varying capabilities in displaying content. In some cases, special symbols may not be displayed correctly on the page, or their display position may be off-center. For patent texts that require precise expression of technical content, these display issues may lead to ambiguity or errors in the description of the technical solution.

[0006] Therefore, improved methods and systems are needed to solve the problem of displaying special symbols in patent texts output by large language models. Summary of the Invention

[0007] To achieve the above objectives, the present invention provides a method and a rapid device for displaying patented text output by a large language model, so as to at least partially solve the above problems.

[0008] According to one aspect of this disclosure, a method for displaying patent text based on the output of a large language model is provided. The method is performed by a computing device and includes receiving the patent text output by the large language model, wherein the patent text includes at least one placeholder, each placeholder including a general symbol and symbol information, and the symbol information corresponding to a special symbol, further including backfilling the positions of the placeholders in the patent text with special symbols that have a mapping relationship with the placeholders, and further including displaying the backfilled patent text.

[0009] In a further implementation, the mapping between placeholders and special symbols is obtained by pre-extracting them from the input text.

[0010] In a further implementation, the placeholders are output in LaTeX format.

[0011] In a further implementation, the special symbols include at least one of the following: formulas, chemical formulas, chemical equations, mathematical symbols, scientific symbols, intellectual property symbols, commercial symbols, chemical units, physical units, drawing characters, images, or emoticons.

[0012] In a further implementation, special symbols are pre-processed into image format.

[0013] In a further embodiment, the method further includes obtaining the position information of the placeholder and the first size information of the special symbol, and determining the second size information of the backfilled special symbol based on the position information and the first size information.

[0014] In a further implementation, determining the second size information takes into account at least one of the following: the font size of the surrounding text, line height, or paragraph width.

[0015] In a further implementation, symbolic information is generated based on the semantic content of special symbols, which can be recognized by a multimodal large language model.

[0016] In a further implementation, when the same special symbol appears multiple times in the patent text, the same placeholder is used for each appearance of the same special symbol.

[0017] In a further implementation, the placeholder is output in at least one of the following formats: MathML format, AsciiMath format, or a custom XML tag format.

[0018] According to another aspect of this disclosure, a computing device is provided. The computing device includes a processor and a memory storing instructions. When the instructions are executed by the processor, the computing device receives patent text output by a large language model, wherein the patent text includes at least one placeholder, each placeholder including a general symbol and symbol information. The computing device also acquires a mapping relationship between the placeholders and special symbols, and fills in the positions of the placeholders in the patent text with special symbols based on the mapping relationship. The instructions further cause the computing device to display the filled-in patent text.

[0019] In a further embodiment, the instructions also cause the computing device to obtain the position information of the placeholder and the first size information of the special symbol, and to determine the second size information of the backfilled special symbol based on the position information, the first size information, and contextual information including at least one of the font size, line height, or paragraph width of the surrounding text.

[0020] In a further implementation, special symbols are preprocessed into at least one of the following: image format, SVG vector graphics format, Unicode encoded string, Base64 encoded data, or font icon format.

[0021] According to another aspect of this disclosure, a non-transitory computer-readable storage medium for storing instructions is provided. When the instructions are executed by a processor, the processor performs the following operations: receiving patent text output by a large language model, wherein the patent text includes at least one placeholder; identifying a mapping relationship between the placeholder and a special symbol, wherein the mapping relationship is obtained by pre-extracting from the input text; and the operation further includes backfilling the positions of the placeholders in the patent text with special symbols that have a mapping relationship with the placeholders, and displaying the backfilled patent text.

[0022] In a further implementation, when the placeholder format in the patent text output by the large language model is incorrect or incomplete, the operation may also include regenerating the patent text.

[0023] The beneficial effects achieved by the present invention using the above structure are as follows: The method disclosed herein for displaying patented text based on a large language model output enables accurate display of formulas, chemical equations, and other special characters.

[0024] To further illustrate the contents of this disclosure, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0025] Figure 1 A system diagram of a computing device for displaying patent text based on a large language model output, according to an embodiment of the present disclosure, is shown.

[0026] Figure 2A flowchart is shown of a method for displaying patent text based on a large language model output according to an embodiment of this disclosure.

[0027] Figure 3 A detailed flowchart of a placeholder backfilling method with defined dimensions according to an embodiment of the present disclosure is shown.

[0028] Figure 4 A node diagram showing the placeholder structure and its mapping relationship with special symbols according to an embodiment of the present disclosure is illustrated.

[0029] Figure 5 A comparison diagram of the patent text before and after backfilling is shown according to an embodiment of the present disclosure.

[0030] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0031] The following description illustrates exemplary aspects of this disclosure. However, it should be understood that this description is not intended to be a limitation on the scope of this disclosure. Rather, the description also covers combinations and modifications of the exemplary aspects described herein.

[0032] As used in this article, the term "Large Language Model" or "LLM" refers to a deep learning-based artificial intelligence model that is trained on large-scale text data and is capable of understanding and generating natural language text. Large Language Models can include, but are not limited to, models based on the Transformer architecture, such as the GPT series, BERT series, or other similar neural network models.

[0033] As used in this paper, the term "placeholder" refers to an alternative marker used in text to indicate the location of a special symbol. A placeholder consists of two parts: a general symbol and symbol information. A general symbol is a character or sequence of characters that can be universally recognized and processed by large language models and display systems. Symbol information is descriptive content associated with the placeholder, used to identify the special symbol corresponding to that placeholder.

[0034] As used in this document, the term "special symbol" refers to a symbol or graphic element that may cause difficulties in regular text processing and display. Special symbols may include, but are not limited to: formulas, chemical formulas, chemical equations, mathematical symbols, scientific symbols, intellectual property symbols, commercial symbols, chemical units, physical units, drawing characters, images, or emoticons.

[0035] The term "backfilling" as used in this document refers to the process of replacing special symbols with placeholders in the patent text. Backfilling is based on the mapping relationship between placeholders and special symbols.

[0036] The term "mapping relationship" used in this article refers to the association between placeholders and their corresponding special symbols. Mapping relationships can be stored in a mapping table and can be obtained by pre-extracting them from the input text.

[0037] This disclosure relates to a method for displaying patent text pages based on Large Language Model (LLM) output. This method solves the display problem of special symbols in patent text through a placeholder backfilling mechanism.

[0038] In the process of generating patent text using large language models, patent texts typically contain various special symbols, such as formulas, chemical formulas, chemical equations, mathematical symbols, scientific symbols, intellectual property symbols, commercial symbols, chemical units, physical units, illustrative characters, images, or emoticons. These special symbols may present display difficulties on a webpage. On one hand, large language models may have limitations in their output capabilities and cannot directly output certain special symbols. On the other hand, webpages or other display interfaces may also encounter difficulties rendering certain special characters, leading to errors, garbled text, or incorrect display in the output text.

[0039] The method disclosed herein employs a placeholder backfilling mechanism to address the aforementioned issues. The core idea of ​​this method includes: first, receiving the patent text output by a large language model, which contains placeholders used to mark the positions of special symbols within the text; second, backfilling the patent text with special symbols that have a mapping relationship to the placeholders at the designated positions; and finally, displaying the backfilled patent text. Through this method, special symbols can be accurately displayed in precise locations, avoiding misalignment or garbled text, and also solving the problem that large language models cannot directly output certain special symbols.

[0040] In some instances, placeholders can comprise two parts: general symbols and symbol information. A general symbol is a character or sequence of characters that can be universally recognized and processed by large language models and display systems. Symbol information is descriptive content associated with the placeholder, used to identify the specific symbol corresponding to that placeholder. The mapping between placeholders and specific symbols can be obtained by pre-extracting them from the input text.

[0041] In some instances, placeholders can be output in LaTeX format. LaTeX format supports the correct output of formulas and chemical formulas, and can be edited after output. In other instances, placeholders can also be output in MathML format, AsciiMath format, or custom XML tag formats.

[0042] In some instances, special characters can be pre-processed into image formats for proper display on web pages or other display interfaces. In other cases, special characters can be pre-processed into SVG vector graphics, Unicode encoded strings, Base64 encoded data, or font icons.

[0043] In some instances, the method may further include obtaining the position information of the placeholder and the first size information of the special symbol, and determining the second size information of the backfilled special symbol based on the position information, the first size information, and context information. The context information may include at least one of the font size, line height, or paragraph width of the surrounding text. In this way, the special symbol can be displayed completely and coherently.

[0044] Figure 1 A system diagram is shown for a computing device 100 for displaying patent text based on Large Language Model (LLM) output. The computing device 100 includes a processor 102, a memory 104, a display 108, a backfill module 110, and a mapping table 112.

[0045] Processor 102 may be communicatively coupled to memory 104. Memory 104 may store instructions that, when executed by processor 102, cause computing device 100 to perform placeholder backfilling operations. In some instances, memory 104 may be a non-transitory computer-readable storage medium that, when executed by processor 102, causes processor 102 to perform operations for displaying patent text with placeholder backfilling.

[0046] The large language model 106 can be external to the computing device 100 and can provide patent text to the computing device 100 via a communication link. The patent text output from the large language model 106 may include multiple placeholders. Each placeholder may include a general symbol and symbol information. The general symbol may be a character or character sequence that can be recognized and processed by the large language model 106 and the computing device 100. The symbol information may correspond to special symbols and can be used to identify the special symbols associated with the placeholders.

[0047] The backfilling module 110 can receive the patent text output from the large language model 106. The backfilling module 110 can communicate with the mapping table 112 to obtain the mapping relationship between placeholders and special symbols. Based on the mapping relationship, the backfilling module 110 can backfill the positions of the placeholders in the patent text with special symbols that have a mapping relationship with the placeholders. After the backfilling operation, the backfilling module 110 can output the patent text to the display 108 to display the backfilled patent text.

[0048] Mapping table 112 can store the mapping relationship between placeholders and special symbols. The mapping relationship can be stored in a local database within computing device 100, or it can be stored on a cloud server accessible via a network connection. The mapping relationship can be stored in a configuration file, such as a JSON or YAML file; or it can be stored in a memory caching system, such as Redis, to improve response speed in high-frequency access scenarios.

[0049] The display 108 can receive the backfilled patent text from the backfill module 110 and can display special symbols at the placeholder locations to render the patent text. The display 108 can be a screen, a monitor, or any other display device capable of rendering text and graphic elements.

[0050] Figure 2 A flowchart illustrating a method for displaying patent text based on the output of a large language model is shown. This can be executed by a computing device, such as the one referenced above. Figure 1 The computing device 100 described.

[0051] In step S202, the computing device 100 may receive patent text output from the large language model. The patent text may include multiple placeholders. Each placeholder may include a general symbol and symbol information. The general symbol may be a character or character sequence that can be recognized and processed by the large language model and the computing device 100. The symbol information may correspond to special symbols and may be used to identify the special symbols associated with the placeholders.

[0052] In step S204, the computing device 100 can fill in the positions of the placeholders in the patent text with special symbols that have a mapping relationship with the placeholders. As described above, the mapping relationship between the placeholders and the special symbols can be obtained by pre-extracting from the input text. Based on the mapping relationship, the computing device 100 can replace each placeholder in the patent text with the corresponding special symbol at the position of the placeholder.

[0053] In step S206, the computing device 100 can display the backfilled patent text. The displayed patent text can include special symbols in the original placeholder positions, thereby achieving accurate display of formulas, chemical equations, mathematical symbols, and other special characters.

[0054] In some instances, the backfilling of special symbols in step S204 can be performed at different stages. In some schemes, backfilling can be performed on the server side before the patent text is sent to the client device. In other schemes, backfilling can be performed on the client side via JavaScript, allowing for real-time backfilling after the client device receives the patent text from the server. In a third variant, a hybrid mode can be used, where simple symbols are processed on the server side and complex symbols are lazily loaded on the client side. The lazy loading method can optimize the initial page load speed, while server-side processing can reduce the computational burden on the client device.

[0055] In some instances, different terminals can be adapted to display the patent text in step S206. These terminals may include web browsers, mobile applications, desktop clients, or PDF export formats. For mobile devices, the display may support touch gesture zooming interactions to view special symbols. For PDF exports, special symbols may be embedded as vector graphics or high-resolution bitmaps to ensure print quality.

[0056] Figure 3 A detailed flowchart of a placeholder backfilling method with defined dimensions is shown. The backfilling method can be executed by a computing device, such as the one referenced above. Figure 1 The computing device 100 described.

[0057] In step S302, the computing device 100 can pre-extract mapping relationships from the input text. The mapping relationship between placeholders and special symbols can be obtained by pre-extracting from the input text. In some instances, the computing device 100 can use third-party tools to identify and extract special symbols in the input text to establish mapping relationships with placeholders. For example, the computing device 100 can use Aspose-Words or similar document processing libraries to parse the input text and identify special symbols, such as formulas, chemical formulas, and mathematical expressions. The extracted special symbols can be associated with corresponding placeholders and stored in mapping table 112.

[0058] In step S304, computing device 100 may receive patent text output from LLM. The patent text may include multiple placeholders. As described above, each placeholder may include general symbols and symbol information.

[0059] In step S306, the computing device 100 can acquire the position information of the placeholders and the first size information of the special symbols. The position information can indicate the position of each placeholder in the patent text, including line number, character offset, or coordinates in the rendered document. The first size information can indicate the original size of the special symbols, such as width and height expressed in pixels or other units of measurement.

[0060] In step S308, the computing device 100 can determine the second size information of the special symbol to be backfilled based on the position information, the first size information, and the context information. The context information may include at least one of the font size, line height, or paragraph width of the surrounding text. The second size information may represent the adjusted size of the special symbol displayed at the placeholder position.

[0061] In some instances, determining the second size information can utilize different size adaptation algorithms. Some schemes may employ a context-based proportional scaling algorithm, where the special character scales proportionally to match the font size of the surrounding text. Other schemes may use an adaptive algorithm based on container width, where the special character is resized to fit the available horizontal space containing the paragraph or column. Some schemes may use a responsive algorithm based on screen resolution, where the special character scales according to the rendering device's display resolution. Some schemes may use a custom scaling algorithm based on user preferences, where the user can specify a preferred display size for different types of special characters. Finally, some schemes may use machine learning models to predict the optimal display size based on context and special character type.

[0062] In some instances, backfilling methods can ensure that special symbols are aligned with the baseline of the surrounding text. The computing device 100 can identify the baseline of the surrounding text and the size of the special symbol to align the special symbol with the text baseline. Baseline alignment can be performed for inline formulas or chemical formulas to ensure that the special symbol is visually consistent with the surrounding text. The baseline alignment process can adjust the vertical position of the special symbol based on the rise and fall heights of the surrounding font.

[0063] In step S310, the computing device 100 can backfill the positions of placeholders in the patent text with special symbols that have a mapping relationship with the placeholders. The backfilling operation can use the second size information determined in step S308 to render the special symbols at an appropriate size. The computing device 100 can replace each placeholder with the corresponding special symbol, applying the determined size and alignment parameters.

[0064] In step S312, the computing device 100 can display the backfilled patent text. The displayed patent text may include rendered special symbols in the original placeholder positions, and the special symbols are resized and aligned according to the context information.

[0065] Figure 4A node graph showing the placeholder structure 400 and its mapping relationship 406 to special symbols 408 is illustrated. The placeholder structure 400 may include a general symbol 402 and symbol information 404. The general symbol 402 may be a character or character sequence that can be recognized and processed by large language models and computing devices. The symbol information 404 may be descriptive content associated with the placeholder structure 400 and may be used to identify the special symbol corresponding to the placeholder structure 400. The mapping relationship 406 connects the placeholder structure 400 to the special symbols 408.

[0066] In some instances, placeholders can be output in LaTeX format. LaTeX format supports the correct output of formulas and chemical formulas and allows for post-output editing. In other instances, placeholders can be output in at least one of the following formats: MathML format, AsciiMath format, or a custom XML tag format. MathML format provides integration with HTML5 standards, while custom XML tag formats offer flexible parsing and processing methods.

[0067] In some instances, placeholder identifiers can use double curly braces such as "{{...}}", prefixed square brackets such as "[SYMBOL:...]", and HTML comment tags such as "<!--SYMBOL:...--> Alternatively, a custom escape sequence such as "\\SPEC{...}" can be used. Different identifier designs can avoid conflicts with the original content in the patent text and can improve parsing accuracy.

[0068] Symbol information 404 can be generated based on the semantic content of special symbols 408. This semantic content can be recognized by a multimodal large language model. In some instances, symbol information 404 can be encoded using different methods besides string format. Symbol information 404 can be encoded using numeric indexes, hash value identifiers, UUID unique identifiers, or hierarchical path identifiers such as "section1 / formula / eq1". Different encoding methods can adapt to different document processing scales; UUID encoding is suitable for symbol management in distributed systems.

[0069] Special symbols 408 may include formulas 410, chemical formulas 412, mathematical symbols 414, and images 416. In some instances, special symbols 408 may include at least one of the following: formulas, chemical formulas, chemical equations, mathematical symbols, scientific symbols, intellectual property symbols, commercial symbols, chemical units, physical units, drawing characters, images, or emoticons.

[0070] In some instances, the special character 408 can be pre-processed into an image format. Pre-processing the special character 408 into an image format facilitates display on web pages or other display interfaces. In some instances, the special character 408 can be pre-processed into at least one of the following formats: image format, SVG vector graphics format, Unicode encoded string, Base64 encoded data, or font icon format. SVG vector graphics format supports lossless scaling and is suitable for scenarios requiring high-definition display. Font icon format maintains consistency with the text style.

[0071] In some instances, special symbols 408 can be pre-generated and stored in batches, rather than being generated in real-time as the user inputs them. Batch pre-generation and storage can improve the processing efficiency of patent texts containing a large number of special symbols.

[0072] For nested special symbols, such as chemical equations containing superscripts, subscripts, and special mathematical symbols, computing device 100 can process nested special symbols either through LaTeX encapsulation or by treating the entire combination as image processing. LaTeX encapsulation preserves the editability of nested special symbols, while image processing simplifies the rendering of complex symbol combinations.

[0073] In some instances, the data structure of mapping relationship 406 may include additional metadata besides the symbol content. This additional metadata may include the original location, symbol type, and rendering parameters. This additional metadata is optional and can be included as needed. When the same special symbol appears multiple times in the patent text, the same placeholder can be used for each occurrence of the same special symbol. Using the same placeholder for repeated special symbols can reduce data redundancy and improve batch processing efficiency.

[0074] Figure 5 The diagram shows a comparison of the patent text before backfilling (502) and after backfilling (504). Before backfilling (502), the patent text can include placeholders, such as {{PLH_1}}, {{PLH_2}}, {{PLH_3}}, and {{PLH_4}}, in the positions where special symbols should appear. The backfilling operation (506) can replace these placeholders with the corresponding special symbols. After backfilling (504), the patent text can display the actual special symbols, including chemical formulas (e.g., H₂O₂) and physical equations (e.g., E=mc²). 2 Mathematical expressions (e.g., ∫f(x)dx) and intellectual property symbols (e.g., ©).

[0075] As described above, the backfilling of placeholders in the patent text with special symbols that have a mapping relationship to the placeholders can convert the patent text from a format containing placeholders to a displayable format. After the backfilling 506 operation, the patent text after backfilling can render the special symbols in the original positions of the placeholders.

[0076] In some instances, when the same special symbol appears multiple times in a patent text, the same placeholder can be used for each occurrence of the same special symbol. For patent texts containing a large number of special symbols, such as hundreds of formulas, using the same placeholder for the same special symbol can optimize batch processing performance. The computing device 100 can process the same special symbol in a single batch operation, rather than processing each instance individually, thereby reducing computational overhead and improving processing efficiency.

[0077] In some instances, when the placeholder format in the patent text output by the large language model is incorrect or incomplete, the computing device 100 can regenerate the patent text. The computing device 100 can detect placeholder format errors by verifying the placeholder structure against an expected pattern. When an error is detected, the computing device 100 can request the LLM to regenerate the patent text with the correct placeholder format.

[0078] In some instances, when special character backfilling fails, computing device 100 may employ different rollback strategies. A first rollback strategy may display the original placeholder text where the special character should appear. A second rollback strategy may display an alternative text description of the special character. A third rollback strategy may display a low-resolution alternative image of the special character. A fourth rollback strategy may display an error message with a manual refresh option, allowing the user to retry the backfilling operation.

[0079] In some instances, the display of the backfilled patent text can offer different interactive modes. A pure display mode renders the patent text with backfilled special symbols but does not allow user modification. An editable mode supports modification of special symbols, allowing users to edit them after backfilling. Annotation mode allows adding annotations to special symbols, enabling users to add notes or comments to specific special symbols in the patent text. A comparison mode displays both the original placeholders and the backfilled results simultaneously, allowing users to view the patent text's 502 status before and 504 after backfilling.

[0080] In some instances, interactive functionality for special symbols, such as clickable expansion of complex formulas, can be achieved through formula editor plugins or chemical formula editor plugins. Formula editor plugins allow users to click on formulas to expand and view detailed information, or to modify formula content. Chemical formula editor plugins provide similar functionality for chemical formulas and equations, enabling users to interact with and modify chemical symbols in patent text.

[0081] Various implementation methods have been described. However, it should be understood that various modifications can be made without departing from the spirit and scope of this disclosure.

Claims

1. A method for displaying patent text based on the output of a large language model, the method being performed by a computing device and comprising: The system receives patent text output by a large language model, wherein the patent text includes at least one placeholder, each placeholder includes a general symbol and symbol information, and the symbol information corresponds to a special symbol; In the patent text, the positions of the placeholders are filled with special symbols that have a mapping relationship with the placeholders; as well as The patent text is displayed after being backfilled.

2. The method according to claim 1, wherein the mapping relationship between the placeholder and the special symbol is obtained by pre-extracting from the input text.

3. The method according to claim 1, wherein the placeholder is output in LaTeX format.

4. The method according to claim 1, wherein the special symbol includes at least one of the following: formula, chemical formula, chemical equation, mathematical symbol, scientific symbol, intellectual property symbol, commercial symbol, chemical unit, physical unit, drawing character, image or emoticon.

5. The method according to claim 1, wherein the special symbol is preprocessed into an image format.

6. The method according to claim 1, further comprising: Obtain the position information of the placeholder and the first size information of the special symbol; as well as The second size information of the special symbol to be backfilled is determined based on the location information and the first size information.

7. The method of claim 6, wherein determining the second size information includes taking into account at least one of the following: the font size of the surrounding text, line height, or paragraph width.

8. The method according to claim 1, wherein the symbol information is generated based on the semantic content of the special symbol, and the semantic content is identified by a multimodal large language model.

9. The method of claim 1, wherein when the same special symbol appears multiple times in the patent text, the same placeholder is used for each occurrence of the same special symbol.

10. The method of claim 1, wherein the placeholder is output in at least one of the following formats: MathML format, AsciiMath format, or custom XML tag format.

11. A computing device, comprising: processor; as well as A memory for storing instructions, which, when executed by the processor, enable the computing device to: Receive patent text output by a large language model, wherein the patent text includes at least one placeholder, and each placeholder includes a general symbol and symbol information; Obtain the mapping relationship between the placeholders and special symbols; Based on the mapping relationship, the special symbol is backfilled into the position of the placeholder in the patent text; and The patent text is displayed after being backfilled.

12. The computing device of claim 11, wherein the instructions further cause the computing device to: Obtain the position information of the placeholder and the first size information of the special symbol; and The second size information of the special symbol to be backfilled is determined based on the location information, the first size information, and contextual information including at least one of the font size, line height, or paragraph width of the surrounding text.

13. The computing device of claim 11, wherein the special symbol is preprocessed into at least one of the following: image format, SVG vector graphics format, Unicode encoded string, Base64 encoded data, or font icon format.

14. A non-transitory computer-readable storage medium storing instructions, which, when executed by a processor, cause the processor to perform operations including: Receive the patent text output by the large language model, wherein the patent text includes at least one placeholder; Identify the mapping relationship between the placeholders and special symbols, wherein the mapping relationship is obtained by pre-extracting from the input text; The placeholders in the patent text are backfilled with special symbols that have the mapping relationship with the placeholders; and The patent text is displayed after being backfilled.

Images