A Deep Learning-Based Intelligent Writing Assistance System and Method

Abstract

The present invention discloses an intelligent writing assistance system and method based on deep learning, which belongs to the technical field of deep learning intelligent writing. The method specifically includes: collecting different types of text data, preprocessing the collected text data, selecting a deep learning model, and training the selected deep learning model; when a user is writing, obtaining the content of the user's current writing, performing semantic analysis and extracting key information, using the trained deep learning model to generate text, and selecting the generated text with the greatest similarity as the output text, evaluating the quality of the generated text and the deep learning model, adjusting and optimizing the generated text and the deep learning model according to the evaluation results, deploying the optimized deep learning model to practical applications, assisting users in writing, greatly improving the quality and accuracy of text generation, and solving the problem of users' lack of writing materials or inspiration to a certain extent.

Description

Technical Field

[0001] The present invention belongs to the technical field of deep learning intelligent writing, and specifically relates to an intelligent writing assistance system and method based on deep learning. Background Art

[0002] Writing is a creative mental labor process in which people use language and text symbols to reflect things, express thoughts and feelings, transfer knowledge and information, and achieve communication. As a complete system process, writing activities can be roughly divided into three stages: "collection-conception-expression". Different from the free writing of writers and the professional writing of professionals, writing in the sense of Chinese courses is a learning activity in which students create texts in written language according to specific requirements under the guidance of teachers to develop and improve their own writing skills.

[0003] Most of the existing technologies are based on manually collected text resource libraries and small character matching retrieval systems. The generation effect is not good, it cannot fit the specific context, and reduces writing efficiency.

[0004] For example, a Chinese patent with authorization announcement number CN109582839B discloses a writing assistance method and a writing assistance client, which includes: when it is detected that a user needs writing assistance, detecting the voice information input by the user; analyzing the voice information to obtain an analysis result; wherein the analysis result is used to determine the composition material required by the user; according to the analysis result, determining at least one composition material in a composition material database; combining the analysis result with at least one composition material to generate writing assistance information; and outputting the writing assistance information. The implementation of the above method can increase the probability that the composition material recommended by the writing assistance client is adopted by the user.

[0005] For example, a Chinese patent with publication number CN117131871A discloses an intelligent writing assistance system with semantic analysis and an auxiliary writing method, including a database, a subject word extraction module and a writing module. The subject word extraction module includes a semantic analysis unit, a candidate word unit and a sentence output unit. After extracting words from the database, the subject word extraction module uses the semantic analysis unit to analyze the meaning of sentences and words in the text of the extracted words, and extracts the same or similar words and sentences through the candidate word unit. When used, the invention can extract similar and related sentence meanings in the database according to writing needs through semantic analysis. According to the written text, through semantic analysis and the selection of candidate words and sentences, the relationship between the various words required for the writing text can be clarified, so as to find key words for writing.

[0006] The problems existing in the above patents are: 1) When establishing the text library, i.e. the material library, the material library is not processed, resulting in reduced accuracy when assisting writing; 2) When assisting writing, the similarity of words and sentences is first analyzed, which lacks accuracy. Summary of the invention

[0007] In view of the shortcomings of the prior art, the present invention proposes an intelligent writing assistance system and method based on deep learning, which collects different types of text data, preprocesses the collected text data, selects a deep learning model, and trains the selected deep learning model. When the user is writing, the content of the user's current writing is obtained, and semantic analysis and key information extraction are performed. The trained deep learning model is used to generate text, and the generated text with the greatest similarity is selected as the output text. The quality of the generated text and the deep learning model are evaluated. According to the evaluation results, the generated text and the deep learning model are adjusted and optimized, and the optimized deep learning model is deployed in practical applications to assist users in writing.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] An intelligent writing assistance method based on deep learning includes the following specific steps:

[0010] Step S1: collecting different types of text data and preprocessing the collected text data;

[0011] Step S2: Select a deep learning model and train the selected deep learning model;

[0012] Step S3: When the user is writing, the content of the user's current writing is obtained, and semantic analysis and key information extraction are performed, and the trained deep learning model is used to generate text, and the generated text with the greatest similarity is selected as the output text;

[0013] Step S4: Evaluate the quality of the generated text and the deep learning model, and adjust and optimize the generated text and the deep learning model according to the evaluation results;

[0014] Step S5: Deploy the optimized deep learning model to actual applications to assist users in writing.

[0015] Specifically, the different types of text data in step S1 include articles, news and social media posts in a specific field.

[0016] Specifically, the preprocessing in step S1 includes: data cleaning, removing punctuation marks, stop words, spelling errors, and performing word segmentation and stem extraction on text data.

[0017] Specifically, the specific steps of performing word segmentation and stem extraction on text data include:

[0018] Step S101: Set the collected text data set to T, where T = {(t1, l1), (t2, l2), ... (t n ,l n )}, where t n Indicates the nth text data, l n Indicates the label of the nth text data;

[0019] Step S102: using Jieba word segmentation to perform word segmentation processing on the collected text data set T;

[0020] Step S103: construct a stem extraction undirected graph model G, G = (V, E), where V represents the word node set in the stem extraction undirected graph model G, and E represents the set of node edges in the stem extraction undirected graph model G. Calculate the weight value of the i-th node in the stem extraction undirected graph model G, and the calculation formula is:

[0021]

[0022] Among them, ω(V i ) represents the weight value of the i-th node in the stem extraction undirected graph model G, d represents the damping coefficient, J(V i ) represents the set of nodes connected to node i in the stem extraction undirected graph model G, ω(V j ) represents the weight value of the jth node in the stem extraction undirected graph model G, |J(V i )| represents the number of nodes connected to node i in the stem extraction undirected graph model G;

[0023] Step S104: Select m word nodes with the highest weight values ​​as word stems.

[0024] Specifically, the specific steps of step S3 are:

[0025] Step S301: pre-processing the text content written by the user, including removing special characters, punctuation marks, stop words, and performing word segmentation;

[0026] Step S302: The pre-processed text content of the user's writing is used as the trained deep learning model to generate a subsequent auxiliary supplementary content set BCon of the user's current writing content. The formula for generating the supplementary content is: BCon p =S(x w ), among which BCon p represents the pth subsequent auxiliary supplementary content, S(x w ) represents the output of the deep learning model, x wRepresents the output of the projection layer of the deep learning model;

[0027] Step S303: Based on the similarity of text matching, the generated subsequent auxiliary supplementary content of the user's current writing content is matched with the user's current writing content. The calculation formula is:

[0028] sim(tex yx ,BCon p )=P(V k |c)P(tex yx |V k ,c),k∈[1,m],

[0029] Among them, c represents the context information of the kth stem, tex yx Indicates the current writing content of the user, P(V k |c) represents the similarity probability between the context information and the extracted k-th stem, P(tex yx |V k ,c) represents the probability of similarity between the user's current writing content and the extracted context information of the kth stem and the kth stem;

[0030] Step S304: Select max[sim(tex yx ,BCon p )] as the subsequent auxiliary supplementary content of the final output.

[0031] Specifically, the specific steps of step S4 are:

[0032] Step S401: defining indicators for evaluating the deep learning model, including the accuracy, recall rate and F1 score of the auxiliary supplementary content;

[0033] Step S402: The calculation formula for the accuracy of the auxiliary supplementary content is: JZD=TP / (TP+FP), where JZD represents the accuracy of the answer, TP represents the true positive examples, i.e., the number of samples correctly predicted as positive by the model, FP represents the false positive examples, i.e., the number of samples incorrectly predicted as positive by the model, the calculation formula for the recall rate is: Rec=TP / (TP+FN), where Rec represents the recall rate, and the calculation formula for the F1 score is: F1=2(JZD×Rec) / (JZD+Rec), where F1 represents the F1 score.

[0034] An intelligent writing assistance system based on deep learning, comprising: a text data acquisition and preprocessing module, a deep learning model training module, a text generation module, a model evaluation module and a deployment module;

[0035] The text data collection and preprocessing module is used to collect different types of text data and preprocess the collected text data;

[0036] The deep learning model training module is used to select a deep learning model and train the selected deep learning model;

[0037] The text generation module is used to obtain the content of the user's current writing when the user is writing, perform semantic analysis and extract key information, use the trained deep learning model to generate text, and select the generated text with the greatest similarity as the output text;

[0038] The model evaluation module is used to evaluate the quality of the generated text and the deep learning model, and adjust and optimize the generated text and the deep learning model according to the evaluation results;

[0039] The deployment module is used to deploy the optimized deep learning model to practical applications to assist users in writing.

[0040] Specifically, the text data collection and preprocessing module includes: a text data collection unit and a preprocessing unit;

[0041] The text data collection unit is used to collect different types of text data;

[0042] The preprocessing unit is used to preprocess the collected text data, including data cleaning, removing punctuation marks, stop words, spelling errors, and performing word segmentation and stem extraction on the text data.

[0043] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of an intelligent writing assistance method based on deep learning when executing the computer program.

[0044] A computer-readable storage medium stores computer instructions, which, when executed, execute the steps of an intelligent writing assistance method based on deep learning.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. The present invention proposes an intelligent writing assistance system based on deep learning, and optimizes and improves the architecture, operation steps and processes. The system has the advantages of simple process, low investment and operation costs, and low production work cost.

[0047] 2. The present invention proposes an intelligent writing assistance method based on deep learning, which uses the collected data to train the deep learning model so that it can generate high-quality text and text that can be understood by the computer. When text is subsequently generated according to the user's current writing content, the quality and accuracy of text generation are greatly improved.

[0048] 3. The present invention proposes an intelligent writing assistance method based on deep learning, which utilizes trained deep learning to generate a set of subsequent auxiliary supplementary content. Based on the similarity of text matching, the generated subsequent auxiliary supplementary content of the user's current writing content is matched with the user's current writing content for similarity. Compared with first matching the text and then selecting the one with the highest similarity, the quality of the subsequent auxiliary supplementary text is greatly improved, which solves the problem of users' lack of writing materials or inspiration to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a flow chart of an intelligent writing assistance method based on deep learning of the present invention;

[0050] Figure 2 This is an architecture diagram of an intelligent writing assistance system based on deep learning in the present invention;

[0051] Figure 3 This is a diagram of an electronic device for an intelligent writing assistance method based on deep learning according to the present invention. DETAILED DESCRIPTION

[0052] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, in the description of the present invention, it should be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside" and the like indicate the orientation or position relationship based on the orientation or position relationship shown in the accompanying drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "No. 1", "No. 2" and "No. 3" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. The present invention is further explained below in conjunction with specific embodiments.

[0053] Example 1

[0054] See also Figure 1 , an embodiment provided by the present invention: an intelligent writing assistance method based on deep learning, comprising the following specific steps:

[0055] Step S1: collecting different types of text data and preprocessing the collected text data;

[0056] Step S2: Select a deep learning model and train the selected deep learning model;

[0057] Step S3: When the user is writing, the content of the user's current writing is obtained, and semantic analysis and key information extraction are performed, and the trained deep learning model is used to generate text, and the generated text with the greatest similarity is selected as the output text;

[0058] Step S4: Evaluate the quality of the generated text and the deep learning model, and adjust and optimize the generated text and the deep learning model according to the evaluation results;

[0059] Step S5: Deploy the optimized deep learning model to actual applications to assist users in writing.

[0060] The different types of text data in step S1 include articles, news, and social media posts in specific fields.

[0061] The preprocessing in step S1 includes: data cleaning, removing punctuation marks, stop words, spelling errors, and performing word segmentation and stem extraction on text data.

[0062] Step S2: The deep learning model is any neural network for text processing, such as recurrent neural network (RNN), long short-term memory (LSTM) or Transformer, etc. The model is trained using the collected data to enable it to understand and generate high-quality text;

[0063] The specific steps for word segmentation and stem extraction of text data include:

[0064] Step S101: Set the collected text data set to T, where T = {(t1, l1), (t2, l2), ... (t n ,l n )}, where t n Indicates the nth text data, l n Indicates the label of the nth text data;

[0065] Step S102: using Jieba word segmentation to perform word segmentation processing on the collected text data set T;

[0066] Jieba word segmentation uses Python's built-in word segmentation library for word segmentation. Other word segmentation methods include Word2vec, etc.

[0067] Step S103: construct a stem extraction undirected graph model G, G = (V, E), where V represents the word node set in the stem extraction undirected graph model G, and E represents the set of node edges in the stem extraction undirected graph model G. Calculate the weight value of the i-th node in the stem extraction undirected graph model G, and the calculation formula is:

[0068]

[0069] Among them, ω(V i ) represents the weight value of the i-th node in the stem extraction undirected graph model G, d represents the damping coefficient, J(V i) represents the set of nodes connected to node i in the stem extraction undirected graph model G, ω(V j ) represents the weight value of the jth node in the stem extraction undirected graph model G, |J(V i )| represents the number of nodes connected to node i in the stem extraction undirected graph model G;

[0070] Step S104: Select m word nodes with the highest weight values ​​as word stems.

[0071] Here, the m selections in the m word nodes can be 1, 10, or 5, depending on actual needs;

[0072] The specific steps of step S3 are:

[0073] Step S301: pre-processing the text content written by the user, including removing special characters, punctuation marks, stop words, and performing word segmentation;

[0074] Step S302: The pre-processed text content of the user's writing is used as the trained deep learning model to generate a subsequent auxiliary supplementary content set BCon of the user's current writing content. The formula for generating the supplementary content is: BCon p =S(x w ), among which BCon p represents the pth subsequent auxiliary supplementary content, S(x w ) represents the output of the deep learning model, x w Represents the output of the projection layer of the deep learning model;

[0075] Step S303: Based on the similarity of text matching, the generated subsequent auxiliary supplementary content of the user's current writing content is matched with the user's current writing content. The calculation formula is:

[0076] sim(tex yx ,BCon p )=P(V k |c)P(tex yx |V k ,c),k∈[1,m],

[0077] Among them, c represents the context information of the kth stem, tex yx Indicates the current writing content of the user, P(V k |c) represents the similarity probability between the context information and the extracted k-th stem, P(tex yx |V k ,c) represents the probability of similarity between the user's current writing content and the extracted context information of the kth stem and the kth stem;

[0078] Step S304: Select max[sim(tex yx ,BCon p )] as the subsequent auxiliary supplementary content of the final output.

[0079] The trained deep learning model is used to generate a set of subsequent auxiliary supplementary content. Based on the similarity of text matching, the generated subsequent auxiliary supplementary content of the user's current writing content is matched with the user's current writing content for similarity. Compared with first matching the text and then selecting the one with the highest similarity, the quality of the subsequent auxiliary supplementary text is greatly improved. The reason is that the accuracy of the subsequent supplementary content generated by the trained deep learning model is higher than that of the text generated by the similarity calculation, that is, the neural network method is better than the ordinary method. The high-quality supplementary content screened out is then subjected to similarity calculation to calculate the optimal supplementary content, and the accuracy of its content is better than that calculated by the ordinary method.

[0080] The specific steps of step S4 are:

[0081] Step S401: defining indicators for evaluating the deep learning model, including the accuracy, recall rate and F1 score of the auxiliary supplementary content;

[0082] Step S402: The calculation formula for the accuracy of the auxiliary supplementary content is: JZD=TP / (TP+FP), where JZD represents the accuracy of the answer, TP represents the true positive examples, i.e., the number of samples correctly predicted as positive by the model, FP represents the false positive examples, i.e., the number of samples incorrectly predicted as positive by the model, the calculation formula for the recall rate is: Rec=TP / (TP+FN), where Rec represents the recall rate, and the calculation formula for the F1 score is: F1=2(JZD×Rec) / (JZD+Rec), where F1 represents the F1 score.

[0083] Example 2

[0084] See also Figure 2 ,Another embodiment provided by the present invention: An intelligent writing assistance system based on deep learning, comprising: a text data acquisition and preprocessing module, a deep learning model training module, a text generation module, a model evaluation module and a deployment module;

[0085] The text data collection and preprocessing module is used to collect different types of text data and preprocess the collected text data;

[0086] The deep learning model training module is used to select a deep learning model and train the selected deep learning model;

[0087] The text generation module is used to obtain the content of the user's current writing when the user is writing, perform semantic analysis and extract key information, use the trained deep learning model to generate text, and select the generated text with the greatest similarity as the output text;

[0088] The model evaluation module is used to evaluate the quality of the generated text and the deep learning model, and adjust and optimize the generated text and the deep learning model according to the evaluation results;

[0089] The deployment module is used to deploy the optimized deep learning model to practical applications to assist users in writing.

[0090] The text data collection and preprocessing module includes: a text data collection unit and a preprocessing unit;

[0091] The text data collection unit is used to collect different types of text data;

[0092] The preprocessing unit is used to preprocess the collected text data, including data cleaning, removing punctuation marks, stop words, spelling errors, and performing word segmentation and stem extraction on the text data.

[0093] Example 3

[0094] See also Figure 3 , an electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of an intelligent writing assistance method based on deep learning.

[0095] A computer-readable storage medium stores computer instructions, which, when executed, execute the steps of an intelligent writing assistance method based on deep learning.

[0096] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0097] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0098] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0099] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0100] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

[0101] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent writing assistance method based on deep learning, characterized in that: The specific steps include: Step S1: Collect different types of text data and pre-process the collected text data; Step S2: Select a deep learning model and train the selected deep learning model; Step S3: When the user is writing, the content of the user's current writing is obtained, and semantic analysis and key information extraction are performed, and the trained deep learning model is used to generate text, and the generated text with the greatest similarity is selected as the output text; Step S4: Evaluate the quality of the generated text and the deep learning model, and adjust and optimize the generated text and the deep learning model according to the evaluation results; Step S5: deploying the optimized deep learning model to actual applications to assist users in writing; The preprocessing in step S1 includes: data cleaning, removing punctuation marks, stop words, spelling errors, and performing word segmentation and stem extraction on text data; The specific steps of performing word segmentation and stem extraction on text data include: Step S101: Set the collected text data set as T, ,in, Indicates the nth text data, Indicates the label of the nth text data; Step S102: using Jieba word segmentation to perform word segmentation processing on the collected text data set T; Step S103: construct an undirected graph model G for stem extraction. , where V represents the word node set in the stem extraction undirected graph model G, and E represents the node edge set in the stem extraction undirected graph model G. The weight value of the i-th node in the stem extraction undirected graph model G is calculated using the following formula: ， in, represents the weight value of the i-th node in the stem extraction undirected graph model G, d represents the damping coefficient, represents the set of nodes connected to node i in the stem extraction undirected graph model G, represents the weight value of the jth node in the stem extraction undirected graph model G, Represents the number of nodes connected to node i in the stem extraction undirected graph model G; Step S104: Select m word nodes with the highest weight values ​​as word stems; The specific steps of step S3 are: Step S301: pre-processing the text content written by the user, including removing special characters, punctuation marks, stop words, and performing word segmentation; Step S302: Use the pre-processed text content of the user's writing as the trained deep learning model to generate a set of subsequent auxiliary supplementary content for the user's current writing content , the formula for generating supplementary content is: ,in, Indicates the pth subsequent auxiliary supplementary content, represents the output of the deep learning model, Represents the output of the projection layer of the deep learning model; Step S303: Based on the similarity of text matching, the generated subsequent auxiliary supplementary content of the user's current writing content is matched with the user's current writing content. The calculation formula is: ， Among them, c represents the context information of the kth stem, Indicates the current writing content of the user. represents the similarity probability between the context information and the extracted k-th stem, represents the probability of similarity between the user's current writing content and the extracted k-th stem and the context information of the k-th stem; Step S304: Select As subsequent auxiliary supplementary content for the final output.

2. The intelligent writing assistance method based on deep learning as claimed in claim 1, characterized in that: The different types of text data in step S1 include articles, news and social media posts in a specific field.

3. The intelligent writing assistance method based on deep learning as claimed in claim 1, characterized in that: The specific steps of step S4 are: Step S401: defining indicators for evaluating the deep learning model, including the accuracy, recall rate and F1 score of the auxiliary supplementary content; Step S402: The calculation formula of the accuracy of the auxiliary supplementary content is: , where JZD represents the accuracy of the answer, TP represents the number of true positive examples, i.e., the number of samples correctly predicted as positive by the model, and FP represents the number of false positive examples, i.e., the number of samples incorrectly predicted as positive by the model. The calculation formula for recall is: , where Rec represents the recall rate and the calculation formula of F1 score is: , where F1 represents the F1 score.

4. An intelligent writing assistance system based on deep learning, which is implemented based on an intelligent writing assistance method based on deep learning according to any one of claims 1 to 3, characterized in that: include: Text data collection and preprocessing module, deep learning model training module, text generation module, model evaluation module and deployment module; The text data collection and preprocessing module is used to collect different types of text data and preprocess the collected text data; The deep learning model training module is used to select a deep learning model and train the selected deep learning model; The text generation module is used to obtain the content of the user's current writing when the user is writing, perform semantic analysis and extract key information, use the trained deep learning model to generate text, and select the generated text with the greatest similarity as the output text; The model evaluation module is used to evaluate the quality of the generated text and the deep learning model, and adjust and optimize the generated text and the deep learning model according to the evaluation results; The deployment module is used to deploy the optimized deep learning model to practical applications to assist users in writing.

5. The intelligent writing assistance system based on deep learning as claimed in claim 4, characterized in that: The text data acquisition and preprocessing module comprises: a text data acquisition unit and a preprocessing unit; The text data collection unit is used to collect different types of text data; The preprocessing unit is used to preprocess the collected text data, including data cleaning, removing punctuation marks, stop words, spelling errors, and performing word segmentation and stem extraction on the text data.

6. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of an intelligent writing assistance method based on deep learning described in any one of claims 1-3 are implemented.

7. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the steps of an intelligent writing assistance method based on deep learning described in any one of claims 1-3 are executed.

Images