Statement retouching method and system based on large model

Abstract

The invention particularly relates to a statement embellishing method and system based on a large model. According to the statement retouching method based on the large model, massive natural language statements are collected in a user-defined mode and preprocessed, the preprocessed training data and the deep learning technology are utilized to train a large-scale language model based on a Transform architecture, a statement retouching model is built based on the trained large-scale language model, and a plurality of candidate retouching results are generated; establishing an evaluation mechanism in a user-defined manner, sorting from high to low according to scoring results, and selecting a polishing result in a user-defined manner; and outputting the polished statement to the user, and allowing the user to apply the statement to various scenes. According to the statement polishing method and system based on the large model, the input statement is deeply analyzed and processed through the deep learning model, automatic statement polishing is achieved, the quality and efficiency of text polishing are improved, the application field of text polishing is expanded, and the requirement of the modern society for high-quality texts can be met.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular to a sentence polishing method and system based on a large model. Background Art

[0002] With the continuous development of natural language processing technology, sentence polishing has become an important means to improve text quality. However, traditional sentence polishing methods often rely on manually written rules or templates and cannot effectively handle complex language phenomena and semantic relationships.

[0003] Based on the above problems, the present invention proposes a sentence polishing method and system based on a large model, which aims to perform in-depth analysis and processing of input sentences through a deep learning model to achieve automatic sentence polishing. Summary of the invention

[0004] In order to remedy the deficiencies of the prior art, the present invention provides a simple and efficient sentence polishing method and system based on a large model.

[0005] The present invention is achieved through the following technical solutions:

[0006] A sentence polishing method based on a large model includes the following steps:

[0007] Step S1: Data collection and preprocessing

[0008] Customize the collection of massive natural language sentences as training data, including sentences from different fields and styles, to ensure the generalization ability of the model; and perform preprocessing, including word segmentation, part-of-speech tagging, and named entity recognition;

[0009] Step S2: Model training

[0010] Using the preprocessed training data and deep learning technology to train a large-scale language model based on the Transformer architecture, the large-scale language model learns the grammatical rules and semantic information of the language through self-supervised learning;

[0011] In step S2, during the training process, the large-scale language model is trained and optimized using Masked Language Modeling (MLM) and Next Sentence Prediction (NSP) tasks to improve the performance of the model.

[0012] The large-scale language model learns the grammatical rules and semantic information of the language through pre-training tasks, and is fine-tuned in subsequent tasks based on task feedback to adapt to different polishing needs.

[0013] Step S3: Sentence polishing model construction

[0014] Building a sentence polishing model based on the trained large-scale language model, wherein the sentence polishing model receives the sentence to be polished as input and generates multiple candidate polishing results through the encoder and decoder structure of the large-scale language model;

[0015] The large-scale language model integrates the grammatical rules, semantic coherence and fluency of the sentence to generate candidate polishing results;

[0016] Step S4: Candidate result evaluation and selection

[0017] Comprehensively consider the output probability of the sentence polishing model, the fluency and semantic coherence of the sentence, and establish a custom evaluation mechanism to evaluate the generated candidate polishing results, and sort them from high to low according to the scoring results. The user customizes the selection of one or more best results as the final polishing result;

[0018] In step S4, a user feedback mechanism is introduced to allow users to score or select candidate results to further optimize the sentence polishing model.

[0019] Step S5: Result output and application

[0020] The polished sentences are output to users, and users are allowed to customize them for various scenarios, such as writing, translation, advertising copywriting, etc.

[0021] A sentence polishing system based on a large model, comprising:

[0022] The data collection module is responsible for customizing the collection of massive natural language sentences as training data, including sentences from different fields and styles, to ensure the generalization ability of the model;

[0023] The preprocessing module is responsible for preprocessing the collected data, including word segmentation, part-of-speech tagging and named entity recognition;

[0024] The model training module is responsible for training a large-scale language model based on the Transformer architecture using preprocessed training data and deep learning technology. The large-scale language model learns the grammatical rules and semantic information of the language through self-supervised learning;

[0025] During the training process, the model training module uses masked language modeling (MLM) and next sentence prediction (NSP) tasks to train and optimize the large-scale language model to improve the performance of the model.

[0026] The large-scale language model learns the grammatical rules and semantic information of the language through pre-training tasks, and is fine-tuned in subsequent tasks based on task feedback to adapt to different polishing needs.

[0027] A sentence polishing model construction module is responsible for constructing a sentence polishing model based on the trained large-scale language model. The sentence polishing model accepts the sentence to be polished as input and generates multiple candidate polishing results through the encoder and decoder structure of the large-scale language model.

[0028] The large-scale language model integrates the grammatical rules, semantic coherence and fluency of the sentence to generate candidate polishing results;

[0029] The candidate result evaluation and selection module is responsible for comprehensively considering the output probability of the sentence polishing model, the fluency and semantic coherence of the sentence, customizing the evaluation mechanism, evaluating the generated candidate polishing results, and sorting them from high to low according to the scoring results. The user customizes the selection of one or more best results as the final polishing result;

[0030] The candidate result evaluation and selection module introduces a user feedback mechanism to allow users to score or select candidate results to further optimize the sentence polishing model.

[0031] The result output and application module is responsible for outputting the polished sentences to users and allowing users to customize and apply them to various scenarios, such as writing, translation, advertising copy, etc.

[0032] A sentence polishing device based on a large model, characterized in that it includes a memory and a processor; the memory is used to store a computer program, and the processor is used to implement the above method steps when executing the computer program.

[0033] A readable storage medium, characterized in that: a computer program is stored on the readable storage medium, and the computer program implements the above method steps when executed by a processor.

[0034] The beneficial effects of the present invention are as follows: the sentence polishing method and system based on a large model performs in-depth analysis and processing of input sentences through a deep learning model, thereby realizing automated sentence polishing, improving the quality and efficiency of text polishing, expanding the application field of text polishing, and being able to meet the needs of modern society for high-quality texts. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Attached Figure 1 It is a schematic diagram of the sentence polishing method based on the large model of the present invention. DETAILED DESCRIPTION

[0037] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0038] The sentence polishing method based on the large model includes the following steps:

[0039] Step S1: Data collection and preprocessing

[0040] Customize the collection of massive natural language sentences as training data, including sentences from different fields and styles, to ensure the generalization ability of the model; and perform preprocessing, including word segmentation, part-of-speech tagging, and named entity recognition;

[0041] Step S2: Model training

[0042] Using the preprocessed training data and deep learning technology to train a large-scale language model based on the Transformer architecture, the large-scale language model learns the grammatical rules and semantic information of the language through self-supervised learning;

[0043] In step S2, during the training process, the large-scale language model is trained and optimized using Masked Language Modeling (MLM) and Next Sentence Prediction (NSP) tasks to improve the performance of the model.

[0044] The large-scale language model learns the grammatical rules and semantic information of the language through pre-training tasks, and is fine-tuned in subsequent tasks based on task feedback to adapt to different polishing needs.

[0045] Step S3: Sentence polishing model construction

[0046] Building a sentence polishing model based on the trained large-scale language model, wherein the sentence polishing model receives the sentence to be polished as input and generates multiple candidate polishing results through the encoder and decoder structure of the large-scale language model;

[0047] The large-scale language model integrates the grammatical rules, semantic coherence and fluency of the sentence to generate candidate polishing results;

[0048] Step S4: Candidate result evaluation and selection

[0049] Comprehensively consider the output probability of the sentence polishing model, the fluency and semantic coherence of the sentence, and establish a custom evaluation mechanism to evaluate the generated candidate polishing results, and sort them from high to low according to the scoring results. The user customizes the selection of one or more best results as the final polishing result;

[0050] In step S4, a user feedback mechanism is introduced to allow users to score or select candidate results to further optimize the sentence polishing model.

[0051] Step S5: Result output and application

[0052] The polished sentences are output to users, and users are allowed to customize them for various scenarios, such as writing, translation, advertising copywriting, etc.

[0053] At the same time, this sentence polishing method based on a large model can also be combined with other NLP (Natural Language Processing) technologies, such as entity recognition and semantic understanding, to provide more accurate and smoother input data for these technologies.

[0054] The sentence polishing system based on the big model includes:

[0055] The data collection module is responsible for customizing the collection of massive natural language sentences as training data, including sentences from different fields and styles, to ensure the generalization ability of the model;

[0056] The preprocessing module is responsible for preprocessing the collected data, including word segmentation, part-of-speech tagging and named entity recognition;

[0057] The model training module is responsible for training a large-scale language model based on the Transformer architecture using preprocessed training data and deep learning technology. The large-scale language model learns the grammatical rules and semantic information of the language through self-supervised learning;

[0058] During the training process, the model training module uses masked language modeling (MLM) and next sentence prediction (NSP) tasks to train and optimize the large-scale language model to improve the performance of the model.

[0059] The large-scale language model learns the grammatical rules and semantic information of the language through pre-training tasks, and is fine-tuned in subsequent tasks based on task feedback to adapt to different polishing needs.

[0060] A sentence polishing model construction module is responsible for constructing a sentence polishing model based on the trained large-scale language model. The sentence polishing model accepts the sentence to be polished as input and generates multiple candidate polishing results through the encoder and decoder structure of the large-scale language model.

[0061] The large-scale language model integrates the grammatical rules, semantic coherence and fluency of the sentence to generate candidate polishing results;

[0062] The candidate result evaluation and selection module is responsible for comprehensively considering the output probability of the sentence polishing model, the fluency and semantic coherence of the sentence, customizing the evaluation mechanism, evaluating the generated candidate polishing results, and sorting them from high to low according to the scoring results. The user customizes the selection of one or more best results as the final polishing result;

[0063] The candidate result evaluation and selection module introduces a user feedback mechanism to allow users to score or select candidate results to further optimize the sentence polishing model.

[0064] The result output and application module is responsible for outputting the polished sentences to users and allowing users to customize and apply them to various scenarios, such as writing, translation, advertising copy, etc.

[0065] The sentence polishing device based on a large model includes a memory and a processor; the memory is used to store a computer program, and the processor is used to implement the above method steps when executing the computer program.

[0066] The readable storage medium stores a computer program, and when the computer program is executed by a processor, the above method steps are implemented.

[0067] Compared with the existing technology, the sentence polishing method and system based on the large model has the following characteristics:

[0068] First, intelligent text polishing is achieved

[0069] This method can not only perform simple grammar checking and correction on the text, but also perform intelligent polishing based on factors such as the text's theme, style, and sentiment. It can automatically adjust the text's expression and tone according to the user's needs and preferences, making the text more in line with the user's expectations and aesthetic standards.

[0070] Second, it is suitable for large-scale text processing

[0071] With the advent of the information age, the amount of text data has exploded. Traditional text polishing methods are often unable to cope with such a huge amount of data. However, this method can achieve efficient processing of large-scale text by building a large model to meet the needs of modern society.

[0072] Third, optimized text quality

[0073] This method is based on deep learning and natural language processing technology. It can accurately understand the content of the text, automatically detect and correct grammatical errors, unclear semantics, logical confusion and other problems, and can significantly improve the quality of the text, making it more accurate, fluent and easy to understand.

[0074] Fourth, iterative

[0075] The iterative optimization method is adopted to improve the accuracy and efficiency of large model polishing through continuous iteration and optimization. This method can make full use of historical polishing results and context information, continuously optimize the performance of the model, and improve the accuracy and efficiency of polishing.

[0076] Fifth, wide application range

[0077] This method is not only applicable to traditional literary creation and editing, but can also be widely used in news reporting, advertising, social media, education and training, etc. It can provide high-quality and efficient polishing services for various text contents, meeting the needs of text polishing in different fields.

[0078] The embodiment described above is only one specific implementation of the present invention. Common changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. A sentence polishing method based on a large model, characterized by: The following steps are involved: Step S1: Data collection and preprocessing Customize the collection of massive natural language sentences as training data and perform preprocessing, including word segmentation, part-of-speech tagging, and named entity recognition; Step S2: Model training Using the preprocessed training data and deep learning technology to train a large-scale language model based on the Transformer architecture, the large-scale language model learns the grammatical rules and semantic information of the language through self-supervised learning; Step S3: Sentence polishing model construction Building a sentence polishing model based on the trained large-scale language model, wherein the sentence polishing model receives the sentence to be polished as input and generates multiple candidate polishing results through the encoder and decoder structure of the large-scale language model; The large-scale language model integrates the grammatical rules, semantic coherence and fluency of the sentence to generate candidate polishing results; Step S4: Candidate result evaluation and selection Comprehensively consider the output probability of the sentence polishing model, the fluency and semantic coherence of the sentence, and establish a custom evaluation mechanism to evaluate the generated candidate polishing results, and sort them from high to low according to the scoring results. The user customizes the selection of one or more best results as the final polishing result; Step S5: Result output and application Output the polished sentences to the user and allow the user to customize them for various scenarios.

2. The sentence polishing method based on a large model according to claim 1, characterized in that: In step S2, during the training process, the masked language model MLM and the next sentence prediction NSP task are used to train and optimize the large-scale language model to improve the performance of the model.

3. The sentence polishing method based on a large model according to claim 2, characterized in that: The large-scale language model learns the grammatical rules and semantic information of the language through pre-training tasks, and is fine-tuned in subsequent tasks based on task feedback to adapt to different polishing needs.

4. The sentence polishing method based on a large model according to claim 1, characterized in that: In step S4, a user feedback mechanism is introduced to allow users to score or select candidate results to further optimize the sentence polishing model.

5. A sentence polishing system based on a large model, characterized by: include: The data collection module is responsible for customizing the collection of massive natural language sentences as training data; The preprocessing module is responsible for preprocessing the collected data, including word segmentation, part-of-speech tagging and named entity recognition; The model training module is responsible for training a large-scale language model based on the Transformer architecture using preprocessed training data and deep learning technology. The large-scale language model learns the grammatical rules and semantic information of the language through self-supervised learning; A sentence polishing model construction module is responsible for constructing a sentence polishing model based on the trained large-scale language model. The sentence polishing model accepts the sentence to be polished as input and generates multiple candidate polishing results through the encoder and decoder structure of the large-scale language model. The large-scale language model integrates the grammatical rules, semantic coherence and fluency of the sentence to generate candidate polishing results; The candidate result evaluation and selection module is responsible for comprehensively considering the output probability of the sentence polishing model, the fluency and semantic coherence of the sentence, customizing the evaluation mechanism, evaluating the generated candidate polishing results, and sorting them from high to low according to the scoring results. The user customizes the selection of one or more best results as the final polishing result; The result output and application module is responsible for outputting the polished sentences to users and allowing users to customize and apply them to various scenarios.

6. The sentence polishing system based on a large model according to claim 5 is characterized in that: During the training process, the model training module uses the masked language model MLM and the next sentence prediction NSP task to train and optimize the large-scale language model to improve the performance of the model.

7. The sentence polishing system based on a large model according to claim 6 is characterized by: The large-scale language model learns the grammatical rules and semantic information of the language through pre-training tasks, and is fine-tuned in subsequent tasks based on task feedback to adapt to different polishing needs.

8. The sentence polishing system based on a large model according to claim 5, characterized in that: The candidate result evaluation and selection module introduces a user feedback mechanism to allow users to score or select candidate results to further optimize the sentence polishing model.

9. A sentence polishing device based on a large model, characterized in that: The method comprises a memory and a processor; the memory is used to store a computer program, and the processor is used to implement the method according to any one of claims 1 to 4 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images