Computer intelligent writing system driven by artificial intelligence

Abstract

The invention provides a computer intelligent writing system driven by artificial intelligence, belongs to the technical field of artificial intelligence and natural language processing, and aims to comprehensively and accurately collect data, cover various data sources, provide rich knowledge basis for writing, enable text content to be rich, diversified, accurate and reliable, and improve writing efficiency. For example, knowledge in multiple aspects can be synthesized during writing of historical and cultural articles, the powerful text analysis module can accurately analyze semantic, emotion, grammar and other elements, text quality is improved, emotion requirements are met, specialty is enhanced, writing styles are diversified, different user scene requirements are met, the application range is widened, the text quality is improved through the perfect content optimization module, and user experience is improved. And the system can adjust styles according to different audiences, has high pertinence and attraction, can meet creation requirements and improve efficiency and quality, and has important practical significance and wide application prospects.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence and natural language processing, and specifically relates to an artificial intelligence driven computer intelligent writing system. Background Art

[0002] In today's information age, the demand for text creation is growing and diverse. Whether it is the news media industry that needs to quickly write a large number of news reports, or the companies that need to write various promotional documents, market analysis reports, etc., the traditional manual writing method is gradually unable to meet the requirements of efficiency and high quality. With the continuous development of information technology, computer technology has made certain progress in text processing. Early word processing software mainly focused on basic text editing functions, such as spelling checking, formatting, etc. Although it improved the standardization and editing efficiency of the text, its ability in content creation was limited. In recent years, artificial intelligence technology, especially machine learning and natural language processing technology, has developed rapidly, providing new opportunities for computer intelligent writing. Researchers have begun to try to use machine learning algorithms to learn large amounts of text data to imitate the way humans write. By analyzing and learning text data from multiple data sources such as Internet pages, e-books, and academic literature, computers can acquire rich language knowledge and writing patterns.

[0003] However, some existing writing assistance tools or systems have many shortcomings. The data collection of some systems is not comprehensive and accurate enough, resulting in the lack of sufficient knowledge support for the generated text content; the analysis capabilities of some text analysis modules are limited, and they cannot accurately understand the semantics, emotions and other key elements of the text, resulting in low quality of the generated text; some writing systems have a single writing style and cannot meet the diverse needs of different users in different scenarios. In addition, the content optimization needs to be further improved, such as grammar checking and sentence fluency optimization, which affects the readability and professionalism of the final text. In summary, the development of a more efficient, accurate and comprehensive artificial intelligence-driven computer intelligent writing system has important practical significance and broad application prospects to meet the growing demand for text creation in actual production and life. Summary of the invention

[0004] The purpose of the present invention is to provide an artificial intelligence driven computer intelligent writing system, aiming to solve the problems in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An artificial intelligence driven computer intelligent writing system, comprising:

[0007] The data collection module is used to collect text data from a variety of data sources, including but not limited to Internet pages, e-books, academic literature, etc. Suppose the total amount of collected data is D, and the data collection rate is V d (Unit: text data volume / time unit), the acquisition time is t, then D = V d ×t;

[0008] The text analysis module can perform semantic analysis, sentiment analysis, and grammatical analysis on the collected data to extract key information and language patterns. For example, in semantic analysis, the word vector model is used to calculate the similarity between words. Suppose the word vectors of word A and word B are and Then their similarity Sab can be obtained by cosine similarity:

[0009] The machine learning module uses deep learning algorithms to learn and train the analyzed text data to continuously optimize its writing ability. In a neural network model (such as a recurrent neural network RNN), the input sequence is X = {x1, x2, ..., x T}, the hidden state sequence is H = {h1, h2, ..., h T}, the output sequence is Y = {y1, y2, ..., y T}, the calculation formula of RNN is: t =tanh(W hh h t-1 +W xh x t +b h ), y t =W hy h t +b y , where W hh , W xh , W hy is the weight matrix, b h 、b y is the bias term;

[0010] The writing generation module generates high-quality text content based on the training results of the machine learning module according to the user's input requirements such as topic, style, and word count. Let the quality evaluation index of the generated text be Q, which is related to grammatical correctness G and semantic coherence S. c , Content Richness C r and other factors, we can set Q = f(G, S c , C r , ...), the specific function form can be further determined and optimized according to actual conditions;

[0011] The content optimization module polishes, proofreads and typesets the generated text to improve the readability and standardization of the text. For example, in typesetting optimization, set the text line spacing to L. s , the character spacing is L w , it can be adjusted according to appropriate typesetting formulas (such as empirical formulas based on visual comfort and reading efficiency) to achieve the best reading effect.

[0012] As a preferred solution of the present invention, the data acquisition module adopts a combination of web crawler technology and data interface call to ensure the comprehensiveness and accuracy of data acquisition. In the calculation of the data capture efficiency of the web crawler, the number of target web pages is assumed to be N. p , the number of web pages successfully crawled is N s , the calculation formula for the crawling success rate is:

[0013] The data collection module also has data screening and classification functions, which can screen and classify the collected data according to the preset standards for subsequent analysis and processing. Let the amount of data that meets the standards after screening be D s , the amount of data before screening is D t , screening ratio P s The calculation formula is:

[0014] As a preferred solution of the present invention, the semantic analysis in the text analysis module adopts the word vector model and semantic network model in natural language processing technology to accurately understand the meaning of the text. In the semantic network, the association strength I between nodes ij (between node i and node j) can be calculated through relevant statistical or machine learning methods, for example, based on factors such as co-occurrence frequency, and the specific formula depends on the algorithm actually adopted.

[0015] Sentiment analysis can determine the emotional tendency expressed in the text, including positive, negative and neutral, and provide emotional reference for the writing generation module. s , whose value range is [-1,1], -1 means completely negative, 1 means completely positive, and 0 means neutral. E is calculated by sentiment analysis algorithm s ,The formulas in the specific algorithm vary depending on the technology used.

[0016] Syntax analysis can detect grammatical errors in text and provide correction suggestions. Let the number of grammatical errors be N e , the text length is L t , grammatical error density D e The calculation formula is:

[0017] As a preferred solution of the present invention, the machine learning module includes a neural network model, such as a recurrent neural network (RNN), a long short-term memory network (LSTM), etc., which can generate text that conforms to language habits and logic through training with a large amount of text data. In LSTM, the calculation formula of the forget gate is f t =σ(W f ·[h t-1 , x t ]+b f ), the input gate calculation formula is i t =σ(W i ·[h t-1 , x t ]+b i ), the candidate cell state calculation formula is: The cell state calculation formula is: The output gate calculation formula is o t =σ(W o ·[h t-1 , x t ]+b o ), the hidden state calculation formula is h t =o t *tanh(C t ), where δ is the sigmoid function, Wf, Wi, Wc, Wo are weight matrices, and bf, bi, bc, bo are bias terms.

[0018] This module also has the function of model update and optimization. With the continuous input of new data, the performance of the writing system can be continuously improved. Let the model performance evaluation index be Pm, and the impact factor of new data on model performance be (the calculation method is determined according to the actual situation), then the updated model performance is, where ΔP is the performance change brought about by the new data.

[0019] As a preferred solution of the present invention, the writing generation module supports multiple writing styles, including news reports, essays, poems, scientific papers, etc. Users can choose according to their needs. Suppose the feature vectors of different writing styles are (k represents different styles, such as news reports Prose for When the user selects a certain style, the writing generation module adjusts the text generation according to the corresponding feature vector. The specific adjustment method can be performed according to the formula involved in the actual algorithm. For example, in the vocabulary selection and sentence structure of the generated sentence, operations are performed according to the association formula between the style feature vector and the text generation rule.

[0020] This module can automatically generate an article outline based on the keywords and topics provided by the user, and fill in and create content according to the outline. Let the keyword set be K = {k1, k2, ..., k n}, the set of concepts related to the topic is T = {t1, t2, ..., t m}, the outline generation algorithm generates an article outline according to the association between keywords and subject concepts (calculated through relevant formulas, such as based on semantic similarity, etc.).

[0021] As a preferred solution of the present invention, the content optimization module includes functions such as spelling check, grammar check, vocabulary replacement, sentence fluency optimization, etc. In vocabulary replacement, let the original vocabulary be W o , replace the word with W r , the evaluation index of the suitability of vocabulary replacement is R v , whose calculation formula can be determined based on factors such as the semantic similarity of words and contextual adaptability, such as in is the semantic similarity between the original word and the replacement word, In order to adapt the replacement vocabulary in the current context, the specific form of the function g is determined according to the actual situation.

[0022] It can also adjust the language style and expression of the text according to different writing purposes and audiences, making the text more targeted and attractive. Suppose the writing purpose and audience feature vectors are The text style adjustment parameter vector is Then the relationship between them can be expressed by a mapping function F: The text style is adjusted according to the mapping relationship.

[0023] As a preferred solution of the present invention, the user inputs the writing topic, style, word count and other requirements through the system interface;

[0024] The data collection module collects relevant text data, and the text analysis module analyzes the data;

[0025] The machine learning module performs learning and training, and the writing generation module generates text content;

[0026] The content optimization module optimizes the text and finally presents the optimized text to the user. The running time of each module in the whole process is T total It can be expressed as the sum of the running time of each module, that is, T total

[0027] =T datacollection +T texctanalysis +T machinelearning +T uritinggeneration +T contentoptimization, where the specific calculation of the running time of each module can be determined based on its internal operations and the formulas involved.

[0028] As a preferred solution of the present invention, the storage medium stores a computer program, and when the computer program is executed by the processor, it implements the functions of the artificial intelligence-driven computer intelligent writing system as described in any one of claims 1 to 6, or implements the method of using the system as described in claim 7.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] First, in terms of data collection, by comprehensively and accurately collecting text data from multiple data sources, including Internet web pages, e-books, academic documents, etc., it can provide a rich and extensive knowledge base for writing. Compared with existing systems, comprehensive data collection avoids the problem of insufficient knowledge support and makes the generated text content richer, more diverse, accurate and reliable. For example, when writing an article about history and culture, the system can obtain professional knowledge from different academic documents, extract vivid story cases from relevant e-books, and understand the latest research trends and public opinions from Internet web pages, thereby creating high-quality texts with both depth and breadth. Secondly, the powerful text analysis module can accurately analyze key elements such as semantics, sentiment, and grammar of the text, which not only helps to improve the quality of the text, but also makes the generated text more in line with the emotional needs of readers. For example, when writing advertising copy, accurate sentiment analysis can make the copy more contagious and attract the attention of consumers, while accurate semantic and grammatical analysis can ensure the logic and regularity of the text. Standardization improves the professionalism of the text. Furthermore, the diversified writing styles meet the needs of different users in different scenarios. Whether it is rapid reporting by news media, corporate promotional copywriting, or paper creation in the academic field, users can choose the appropriate writing style according to specific requirements, which greatly broadens the application scope of the system and improves its applicability and practicality. Finally, the complete content optimization module further improves the quality of the text. Functions such as grammar checking and sentence fluency optimization make the text more fluent and easy to read, enhancing the readability of the text. At the same time, the optimization module can also adjust the language style and expression according to different writing purposes and audiences, making the text more targeted and attractive. For example, the text for children readers will adopt a more concise, easy-to-understand, vivid and interesting language style, while the text for professionals will pay more attention to professionalism and accuracy. In summary, this intelligent writing system can effectively meet the growing demand for text creation in actual production and life, improve writing efficiency and quality, and has important practical significance and broad application prospects. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present invention will be described clearly and completely below in combination with the embodiments of the present invention. Obviously, the described embodiments are only 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 are within the scope of protection of the present invention.

[0032] Example 1

[0033] The present invention provides the following technical solutions:

[0034] An artificial intelligence driven computer intelligent writing system, comprising:

[0035] The data collection module is used to collect text data from a variety of data sources, including but not limited to Internet pages, e-books, academic literature, etc. Suppose the total amount of collected data is D and the data collection rate is V d (Unit: text data volume / time unit), the acquisition time is t, then D = V d ×t;

[0036] The text analysis module can perform semantic analysis, sentiment analysis, and grammatical analysis on the collected data to extract key information and language patterns. For example, in semantic analysis, the word vector model is used to calculate the similarity between words. Suppose the word vectors of word A and word B are and Then their similarity Sab can be obtained by cosine similarity:

[0037] The machine learning module uses deep learning algorithms to learn and train the analyzed text data to continuously optimize its writing ability. In a neural network model (such as a recurrent neural network RNN), the input sequence is X = {x1, x2, ..., x T}, the hidden state sequence is H = {h1, h2, ..., h T}, the output sequence is Y = {y1, y2, ..., y T}, the calculation formula of RNN is: t =tanh(W hh h t-1 +W xh x t +b h ), y t =W hy h t +b y , where W hh , W xh , W hy is the weight matrix, b h 、by is the bias term;

[0038] The writing generation module generates high-quality text content based on the training results of the machine learning module according to the user's input requirements such as topic, style, and word count. Let the quality evaluation index of the generated text be Q, which is related to grammatical correctness G and semantic coherence S. c , Content Richness C r and other factors, we can set Q = f(G, S c , C r , ...), the specific function form can be further determined and optimized according to actual conditions;

[0039] The content optimization module polishes, proofreads and typesets the generated text to improve the readability and standardization of the text. For example, in typesetting optimization, set the text line spacing to L. s , the character spacing is L w , it can be adjusted according to appropriate typesetting formulas (such as empirical formulas based on visual comfort and reading efficiency) to achieve the best reading effect.

[0040] In a specific embodiment of the present invention, firstly, the data acquisition module can collect text data from a variety of rich data sources, which enables the writing system to obtain extensive and comprehensive knowledge and information. In actual production and life, for example, for the news media industry, it can quickly collect real-time news events from Internet pages, obtain professional knowledge from academic literature, etc., and provide a solid foundation for writing high-quality, rich and accurate news articles. The various analysis functions of the text analysis module are of great significance. Semantic analysis can accurately understand the meaning of the text and help to accurately convey concepts when writing. Sentiment analysis plays an obvious role in the field of marketing. For example, when writing advertising copy, you can create more promotional texts that can resonate with consumers based on the results of sentiment analysis of consumer comments. Grammar analysis can also be used to analyze the content of advertisements and to analyze the content of advertisements. Analysis ensures the standardization of the generated text and improves the quality and readability of the text. The machine learning module continuously optimizes writing capabilities through deep learning algorithms. Just like in the content creation platform, as the training continues, the system can generate texts that are more in line with user needs and market trends, such as more exciting novel plots, more professional scientific reports, etc. The writing generation module can generate texts according to the diverse requirements of users, meeting the writing needs of different users in different scenarios, whether it is student composition writing, corporate copywriting planning or personal literary creation, etc. Finally, the content optimization module further improves the quality of the text, making the text finally presented to the user more perfect. In actual applications, it can save users a lot of modification and polishing time and improve work efficiency.

[0041] Specifically, the data collection module uses a combination of web crawler technology and data interface calls to ensure the comprehensiveness and accuracy of data collection. In the calculation of the data capture efficiency of the web crawler, let the number of target web pages be N. p , the number of web pages successfully crawled is N s , the calculation formula for the crawling success rate is:

[0042]

[0043] The data collection module also has data screening and classification functions, which can screen and classify the collected data according to the preset standards for subsequent analysis and processing. Let the amount of data that meets the standards after screening be D s , the amount of data before screening is D t , screening ratio P s The calculation formula is:

[0044] In this embodiment: the web crawler technology can be used to widely capture text data from Internet web pages. The Internet contains a vast amount of information, covering various fields and various types of knowledge. This method can obtain real-time updated news, rich and diverse blog articles, professional academic research results, etc., which greatly enriches the data source of the writing system and lays the foundation for generating high-quality and diverse text content. The data interface call can accurately collect data for specific e-book platforms, academic databases, etc., which is complementary to the web crawler. It can ensure that more authoritative and standardized text resources are obtained. For example, collecting classic literary works from well-known e-book platforms helps the writing system learn excellent writing styles and grammatical structures; obtaining professional knowledge from academic literature databases can enable the writing system to The generated text is more accurate and in-depth when it comes to related fields. The screening and classification functions of the data collection module are also extremely important. In today's era of information explosion, the collected data is often disorganized. The screening function can remove low-quality and irrelevant data according to preset standards, such as text quality, relevance, etc., which can save time and resources for subsequent analysis and processing and improve the efficiency of the system. The classification function classifies the collected data according to different themes, genres, etc., so that when conducting text analysis and machine learning, different types of data can be processed more specifically. For example, news reports and essays can be processed separately, which helps the writing system to better learn their respective characteristics, so that it can be more handy when generating the corresponding type of text and meet the writing needs of users in different scenarios.

[0045] Specifically, the semantic analysis in the text analysis module uses the word vector model and semantic network model in natural language processing technology to accurately understand the meaning of the text. In the semantic network, the association strength between nodes is ij(between node i and node j) can be calculated through relevant statistical or machine learning methods, for example, based on factors such as co-occurrence frequency, and the specific formula depends on the algorithm actually adopted.

[0046] Sentiment analysis can determine the emotional tendency expressed in the text, including positive, negative and neutral, and provide emotional reference for the writing generation module. s , whose value range is [-1,1], -1 means completely negative, 1 means completely positive, and 0 means neutral. E is calculated by sentiment analysis algorithm s ,The formulas in the specific algorithm vary depending on the technology used.

[0047] Syntax analysis can detect grammatical errors in text and provide correction suggestions. Let the number of grammatical errors be N e , the text length is L t , grammatical error density D e The calculation formula is:

[0048] In this embodiment: semantic analysis uses the word vector model and semantic network model in natural language processing technology, which can deeply understand the meaning of the text. The word vector model can convert words into mathematical vectors, and accurately grasp the semantic relationship between words by calculating the distance between vectors. For example, when the writing system wants to generate an article about technology, it can understand the close connection between related terms such as "artificial intelligence" and "machine learning" through the word vector model, so as to accurately apply and explain these concepts in the text, making the content of the article more accurate and professional. The semantic network model further constructs a more complex semantic relationship network between words, which can help the writing system understand the overall semantic structure of the text, such as the logical relationship between sentences, the thematic coherence of paragraphs, etc. When writing a story, the semantic network model can analyze the relationship between the various plots in the story, making the generated story more organized and the plot transition more natural. The sentiment analysis function is also of great significance to the writing system. It can judge the emotional tendency expressed by the text, which is very important in It is very useful in many practical application scenarios. For example, in social media content creation, the writing system can write text that conforms to a specific emotional atmosphere based on the results of sentiment analysis. If it is creating promotional copy for a brand's social media account, when the target audience has a positive emotional tendency towards the brand, the writing system can generate more positive and enthusiastic text to further enhance user favorability; when negative emotions are detected, the writing system can generate text related to explanations or improvement measures to alleviate user dissatisfaction. The grammatical analysis function can detect grammatical errors in the text and provide correction suggestions, which is crucial to improving the quality and standardization of the generated text. In scenarios such as business writing or academic writing, grammatical correctness is a basic requirement. The grammatical analysis function can promptly detect and correct grammatical errors in the text generated by the writing system to avoid affecting the readability and professionalism of the text due to grammatical problems. At the same time, the correction suggestions it provides can also help the writing system to continuously learn and improve, improve its own grammatical level, and thus generate more accurate and fluent text.

[0049] Specifically, the machine learning module includes neural network models, such as recurrent neural network (RNN), long short-term memory network (LSTM), etc., which can generate text that conforms to language habits and logic through training with a large amount of text data. In LSTM, the calculation formula of the forget gate is f t =σ(W f ·[h t-1 , x t ]+b f ), the input gate calculation formula is i t =σ(W i ·[h t-1 , x t ]+b i ), the candidate cell state calculation formula is: The cell state calculation formula is: The output gate calculation formula is o t =σ(W o ·[h t-1 , x t ]+b o ), the hidden state calculation formula is h t =o t *tanh(C t ), where δ is the sigmoid function, Wf, Wi, Wc, Wo are weight matrices, and bf, bi, bc, bo are bias terms.

[0050] This module also has the function of model update and optimization. With the continuous input of new data, the performance of the writing system can be continuously improved. Let the model performance evaluation index be Pm, and the impact factor of new data on model performance be (the calculation method is determined according to the actual situation), then the updated model performance is, where ΔP is the performance change brought about by the new data.

[0051] In this embodiment: the application of neural network models such as recurrent neural networks (RNNs) and long short-term memory networks (LSTMs) brings powerful learning and generation capabilities to the writing system. RNNs can process sequential data and are very suitable for information processing such as texts that have a sequence. They can learn the order of words in sentences and the contextual relationship of sentences in paragraphs. For example, when the writing system generates a novel, RNN can reasonably predict the next content based on the previously generated plots and descriptions, making the development of the story more coherent and natural. LSTMs perform well in processing long-term dependencies. When writing long texts such as scientific papers or historical research reports, it is often necessary to consider a lot of background information and long-term development context. LSTMs can remember long-term information and avoid forgetting information, so as to better integrate and use this information. For example, when writing an article about the development history of a certain technology, LSTMs can accurately remember the technology at different stages. The model update and optimization function is even more critical. With the continuous input of new data, the writing system can continuously improve its performance. In real life, language is constantly developing and changing, and new vocabulary, expressions and hot topics are constantly emerging. The model update function can enable the writing system to adapt to these changes in a timely manner and maintain sensitivity to the latest language trends. For example, when new popular words or new language styles appear on the Internet, the writing system can quickly learn and apply these new elements through updates and optimization to generate text content that is more in line with the characteristics of the current era and meet users' needs for novel and fashionable writing. Moreover, this continuous model update and optimization can also improve the adaptability and versatility of the writing system. It can adapt to the writing needs of different fields and different topics. Whether it is literary creation, commercial copywriting or academic research, it can continuously learn new data to improve the quality and relevance of the generated text and provide users with better writing services.

[0052] Specifically, the writing generation module supports a variety of writing styles, including news reports, essays, poems, scientific papers, etc. Users can choose according to their needs. Suppose the feature vectors of different writing styles are (k represents different styles, such as news reports Prose for When the user selects a certain style, the writing generation module adjusts the text generation according to the corresponding feature vector. The specific adjustment method can be performed according to the formula involved in the actual algorithm. For example, in the vocabulary selection and sentence structure of the generated sentence, operations are performed according to the association formula between the style feature vector and the text generation rule.

[0053] This module can automatically generate an article outline based on the keywords and topics provided by the user, and fill in and create content according to the outline. Let the keyword set be K = {k1, k2, ..., k n}, the set of concepts related to the topic is T = {t1, t2, ..., t m}, the outline generation algorithm generates an article outline according to the association between keywords and subject concepts (calculated through relevant formulas, such as based on semantic similarity, etc.).

[0054] In this embodiment: In actual production and life, different writing scenarios and needs require different writing styles. For example, news reports require conciseness, objectivity and accuracy. The writing generation module can generate relevant texts in accordance with the style of news reports and quickly convey key information. When an enterprise needs to release product news, the module can quickly generate reports that meet news standards to help enterprises promote product information in a timely and effective manner. For prose writing, it can create a beautiful artistic conception and emotional atmosphere. When users create personal prose, the writing generation module can generate infectious prose paragraphs based on the theme and emotional tendency input by the user, helping users to better express their inner feelings. The support for poetry writing style meets the user's needs for literary and artistic creation. It can generate verses with rhythm and rhythm to provide inspiration for poets or assist in creating complete poetry works. The support for scientific and technological paper writing style is of great significance to scientific researchers. It can generate content in accordance with the standards and structure of scientific and technological papers, including accurate Accurate research background introduction, rigorous experimental method description and scientific conclusion analysis, etc., greatly reduce the writing burden of scientific researchers and improve writing efficiency. In addition, the writing generation module can automatically generate an article outline based on the keywords and topics provided by the user, and the function of filling and creating content according to the outline is also very practical. In the writing of corporate planning schemes, users only need to enter relevant keywords and topics, such as "marketing strategy" and "new product promotion", etc., and the writing generation module can quickly generate a reasonable outline, including market analysis, target customer positioning, promotion channels and other aspects of the content framework, and then fill in the content according to the outline to provide the company with a complete and organized draft of the planning scheme, reducing the time and energy consumption of planners from scratch, and improving work efficiency. Moreover, this outline generation and content creation method based on keywords and topics can ensure that the article is closely centered on the core content, avoid problems such as going off topic or loose content, and improve the quality and pertinence of the text.

[0055] Specifically, the content optimization module includes functions such as spelling check, grammar check, vocabulary replacement, and sentence fluency optimization. In vocabulary replacement, let the original vocabulary be W o , replace the word with W r , the evaluation index of the suitability of vocabulary replacement is R v, whose calculation formula can be determined based on factors such as the semantic similarity of words and contextual adaptability, such as in is the semantic similarity between the original word and the replacement word, In order to adapt the replacement vocabulary in the current context, the specific form of the function g is determined according to the actual situation.

[0056] It can also adjust the language style and expression of the text according to different writing purposes and audiences, making the text more targeted and attractive. Suppose the writing purpose and audience feature vectors are The text style adjustment parameter vector is Then the relationship between them can be expressed by a mapping function F: The text style is adjusted according to the mapping relationship.

[0057] In this embodiment: the spelling check function can effectively avoid spelling errors in the text, which is very important for improving the professionalism and readability of the text. In business email communication, a correctly spelled email can leave a good impression on the other party and show the rigor and professionalism of the sender. For example, when doing business with international customers, an email without spelling errors can enhance the customer's trust in the company and promote the smooth development of the business. In addition to improving the standardization of the text as mentioned above, the grammar check function can also help users avoid some common grammatical errors. In academic writing, grammatical accuracy is one of the important criteria for judging the quality of papers. The grammar check function of the content optimization module can detect grammatical problems in complex sentence structures, such as inconsistency between subject and predicate, improper use of clauses, etc., and provide correction suggestions to make academic papers more in line with academic norms and increase the probability of being accepted by academic journals. The vocabulary replacement function can improve the richness and accuracy of the text's expression. In literary creation, the author may encounter a lack of vocabulary or inappropriate use of words. The vocabulary replacement function can provide Appropriate vocabulary replacement suggestions, such as replacing "beautiful" with "colorful" when describing a scene, can make the text more vivid and enhance the artistic appeal of the work. At the same time, for users who write in non-native languages, the vocabulary replacement function can help them find more appropriate vocabulary expressions and improve the quality of writing. The sentence fluency optimization function is committed to making the sentences in the text more fluent and natural. In the creation of advertising copy, fluent sentences can better convey the characteristics and advantages of the product and attract the attention of consumers. This function can adjust the structure and word order of sentences, eliminate awkward or unnatural expressions, and make advertising copy easier to read and understand, thereby improving the publicity effect of advertising. Moreover, the function of adjusting the language style and expression of the text according to different writing purposes and audiences makes the text more targeted and attractive. For example, for storybooks for children readers, the writing system can adjust the language style to be simpler, easier to understand, vivid and interesting; while for technical reports for professionals, a rigorous and accurate language style is adopted to meet the reading needs of different audiences and improve the dissemination and acceptance of the text.

[0058] Specifically, users input writing topics, styles, word counts, and other requirements through the system interface;

[0059] The data collection module collects relevant text data, and the text analysis module analyzes the data;

[0060] The machine learning module performs learning and training, and the writing generation module generates text content;

[0061] The content optimization module optimizes the text and finally presents the optimized text to the user. The running time of each module in the whole process is T total It can be expressed as the sum of the running time of each module, that is, Ttotal =T datacollection +T texctanalysis +T machinelearning +T uritinggeneration +T contentoptimization , where the specific calculation of the running time of each module can be determined based on its internal operations and the formulas involved.

[0062] In this embodiment: From the perspective of user experience, the method in which users input writing topics, styles, word counts, etc. through the system interface greatly improves the convenience of writing. Users no longer need to conceive and create from scratch, but only need to clarify their needs. For example, a corporate publicity staff member wants to write a news report style promotional article about a new product. He only needs to enter the "new product introduction" topic, "news report" style and appropriate word count requirements in the interface, and the system can perform subsequent operations according to its instructions. This saves users a lot of time and energy, allowing them to complete writing tasks more efficiently. Secondly, the data collection module collects relevant text data, which can provide rich materials and references for writing. In today's era of information explosion, In the current era, by collecting data from a variety of data sources such as Internet pages, e-books, and academic literature, the system can obtain extensive and comprehensive knowledge. For example, when writing an article on the development trend of artificial intelligence, the data collection module can collect information such as research results and application cases about artificial intelligence from different fields and different periods, making the generated article content more substantial, accurate, and authoritative. This is very valuable for users who need to have an in-depth understanding of a certain field for writing, and avoids the tedious process of searching and screening a large amount of information by themselves. Furthermore, the steps of analyzing the data by the text analysis module can dig out the key information and language patterns in the data. Through semantic analysis, the system can accurately understand the meaning of the text. , helping users better grasp the core concepts related to the topic, sentiment analysis can inject appropriate emotional colors into the article. For example, when writing a story-telling article, the ups and downs of the plot and the emotional expression of the characters can be adjusted according to the sentiment analysis results to make the article more appealing. Syntax analysis can promptly detect grammatical errors in the text and provide correction suggestions to improve the quality and readability of the article. This is a great help for users who are not very proficient in grammar or want the article to be more standardized, ensuring the accuracy and professionalism of the generated text at the language level. Then, the machine learning module learns and trains to make the text content generated by the writing generation module more high-quality and personalized. The machine learning module uses deep learning algorithms to analyze the text data after analysis. By learning from the existing data, the system can continuously optimize its writing ability and adapt to the needs and style preferences of different users. For example, for users who often write scientific articles, the system will gradually learn the user's word usage habits, sentence structure and other characteristics as they are used, and generate scientific articles that are more in line with their style. At the same time, through continuous learning and training, the system can generate innovative content instead of simply repeating existing texts, providing users with more inspiration and ideas, and promoting the dissemination and innovation of knowledge. Finally, the content optimization module optimizes the text and presents the optimized text to the user. This step further improves the quality and readability of the text. The spelling check function of the content optimization module can avoid the bad reading experience caused by spelling errors.Grammar checking ensures that the sentence structure is correct and smooth, the vocabulary replacement function can make the language of the article richer and more accurate, and the optimization of sentence fluency makes the logic of the article clearer and more coherent. For example, in the process of writing a paper, the content optimization module can replace some overly colloquial words with more professional academic words to make the article more in line with academic norms, and present the optimized text to the user, so that the user can directly obtain high-quality writing results, whether it is used for report writing at work, paper writing in academic research, or personal literary creation, etc., it can greatly improve user satisfaction and work efficiency.

[0063] Specifically, the storage medium stores a computer program, which, when executed by a processor, implements the functions of an artificial intelligence-driven computer intelligent writing system as claimed in any one of claims 1 to 6, or implements a method of using the system as claimed in claim 7.

[0064] In this embodiment: From the perspective of technical implementation and convenience of application, storage media storing computer programs provides a basis for the widespread application of intelligent writing systems. By storing the programs in suitable storage media, such as hard disks, flash memory, etc., the system can be easily installed and run on different devices. This means that users can use the intelligent writing system on personal computers, laptops, or mobile devices such as tablets and smart phones. For example, a reporter who often goes out for interviews can install the system on a smart phone. When he needs to write news articles quickly, he can use the intelligent writing system on the mobile phone to complete the creation and optimization of the manuscript even without a computer, which greatly improves the flexibility of work. Activeness and efficiency. For enterprises and organizations, this storage medium storing computer programs is conducive to the centralized management and sharing of writing resources. Enterprises can deploy storage media storing intelligent writing system programs on internal servers, and employees can access and use the system through the internal network according to their needs. In this way, employees from different departments can use the intelligent writing system to improve writing quality and efficiency when writing various documents, such as market research reports, product manuals, and internal corporate communications. At the same time, the company can uniformly update and maintain the programs in the storage medium to ensure that all employees use the latest version of the system, thereby maintaining the consistency of the entire company's writing style and the advancement of technology, which is very important in teaching. In the field of education, computer programs stored in storage media can provide strong support for teaching and learning. Teachers can use the intelligent writing system to assist in the preparation of teaching materials, such as making lesson plans and writing teaching cases. Students can also use the system to improve their writing skills, especially in writing papers and reports. For example, in university writing courses, teachers can guide students to use the intelligent writing system for creation, and help students correct errors in writing and improve the quality of articles through system grammar checking, content optimization and other functions. At the same time, students can learn the working principles and algorithms of the intelligent writing system, gain a deep understanding of natural language processing and artificial intelligence technology, broaden their knowledge horizons, cultivate innovative thinking and practical ability, and learn from data. From the perspective of security and intellectual property protection, storing computer programs on storage media can better ensure the security of the system and the confidentiality of user data. Storage media can use security technologies such as encryption to prevent programs from being illegally tampered with or stolen. The text content generated by the intelligent writing system can also be protected by the security mechanism of the storage medium to ensure that the user's intellectual property rights are not infringed. For example, a writer uses the intelligent writing system to create a novel. The security measures of the storage medium can prevent others from unauthorized access and plagiarism of his work. At the same time, when companies use intelligent writing systems for commercial writing, they can also protect their own business secrets and core documents to avoid economic losses and competitive disadvantages caused by data leaks. In addition,Storing computer programs in storage media is conducive to the continuous improvement and upgrading of the system. With the continuous development of artificial intelligence and natural language processing technology, R&D personnel can continuously update and optimize the programs of the intelligent writing system. By storing new program versions in storage media, users can easily upgrade and enjoy more advanced functions and better writing experience. For example, when new deep learning algorithms are developed to improve the quality and accuracy of text generation, users can obtain and install updated programs through storage media, so that the intelligent writing system always stays at the forefront of technology, constantly adapts to changes in market and user needs, and provides users with better quality and more efficient writing services.

[0065] The working principle and use process of the present invention:

[0066] First, the data collection module starts working. It uses a combination of web crawler technology and data interface calls to comprehensively and accurately collect text data from multiple data sources such as Internet web pages, e-books, and academic documents. For example, web crawlers will search for relevant web page content on the Internet according to preset rules, and the data interface will connect to e-book platforms and academic databases to obtain data. During the collection process, the data collection module will also screen and classify the collected data, retain and classify the data that meets the preset standards, such as classifying academic documents related to history and culture into one category, and related e-book story cases into another category, etc. Then, the text analysis module takes over. It performs semantic analysis on the collected data and uses word vector models and semantic network models to accurately understand the meaning of the text, just like building a semantic network of knowledge, so that each word and concept has a clear position and association. At the same time, sentiment analysis is performed to judge the emotional tendency of the text and provide emotional reference for subsequent writing, just like giving the text an emotional color. Syntax analysis is also carried out simultaneously to detect grammatical errors and provide correction suggestions to ensure the grammatical correctness of the text. Then, the machine learning module begins to play a role. The deep learning algorithms, such as recurrent neural networks (RNN) and long short-term memory networks (LSTM), are used to learn and train the analyzed text data. Through the training of a large amount of text data, the writing ability is continuously optimized, just like a student who is constantly learning knowledge and improving his writing skills. It will generate texts that conform to language habits and logic based on the learned language patterns and knowledge. After that, the writing generation module will create texts according to the subject, style, word count and other requirements entered by the user. If the user requires to write an article about history and culture and chooses an academic style, it will extract knowledge from the classified historical and cultural related data based on the training results of the machine learning module, automatically generate an article outline, and fill and create content according to the outline, incorporating professional academic views and knowledge. Finally, the content optimization module will polish, proofread and typeset the generated text. It will perform spelling and grammar checks to ensure that there are no errors in the text. It will also perform vocabulary replacement and sentence fluency optimization to make the text more fluent and easy to read. At the same time, it will adjust the language style and expression according to different writing purposes and audiences to make the text more targeted and attractive, and finally present high-quality text to users.

[0067] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An artificial intelligence driven computer intelligent writing system, characterized in that: include: The data collection module is used to collect text data from a variety of data sources, including but not limited to Internet pages, e-books, academic literature, etc. Suppose the total amount of collected data is D, and the data collection rate is V d (Unit: text data volume / time unit), the acquisition time is t, then D = V d ×t; The text analysis module can perform semantic analysis, sentiment analysis, and grammatical analysis on the collected data to extract key information and language patterns. For example, in semantic analysis, the word vector model is used to calculate the similarity between words. Suppose the word vectors of word A and word B are and Then their similarity Sab can be obtained by cosine similarity: The machine learning module uses deep learning algorithms to learn and train the analyzed text data to continuously optimize its writing ability. In a neural network model (such as a recurrent neural network RNN), the input sequence is X = {x1, x2, ..., x T }, the hidden state sequence is H = {h1, h2, ..., h T }, the output sequence is Y = {y1, y2, ..., y T }, the calculation formula of RNN is: t =tanh(W hh h t-1 +W xh x t +b h ), y t =W hy h t +b y , where W hh , W xh , W hy is the weight matrix, b h , b y is the bias term; The writing generation module generates high-quality text content based on the training results of the machine learning module according to the user's input requirements such as topic, style, and word count. Let the quality evaluation index of the generated text be Q, which is related to grammatical correctness G and semantic coherence S. c , Content Richness C r and other factors, we can set Q = f(G, S c , C r , ...), the specific function form can be further determined and optimized according to actual conditions; The content optimization module polishes, proofreads and typesets the generated text to improve the readability and standardization of the text. For example, in typesetting optimization, set the text line spacing to L. s , the character spacing is L w , it can be adjusted according to appropriate typesetting formulas (such as empirical formulas based on visual comfort and reading efficiency) to achieve the best reading effect.

2. The computer intelligent writing system driven by artificial intelligence according to claim 1, characterized in that: The data collection module uses a combination of web crawler technology and data interface calls to ensure the comprehensiveness and accuracy of data collection. In the calculation of the data capture efficiency of the web crawler, the number of target web pages is N. p , the number of web pages successfully crawled is N s , the calculation formula for the crawling success rate is: The data collection module also has data screening and classification functions, which can screen and classify the collected data according to the preset standards for subsequent analysis and processing. Let the amount of data that meets the standards after screening be D s , the amount of data before screening is D t , screening ratio P s The calculation formula is:

3. The computer intelligent writing system driven by artificial intelligence according to claim 2, characterized in that: The semantic analysis in the text analysis module uses the word vector model and semantic network model in natural language processing technology to accurately understand the meaning of the text. In the semantic network, the association strength between nodes is ij (between node i and node j) can be calculated through relevant statistical or machine learning methods, for example, based on factors such as co-occurrence frequency, and the specific formula depends on the algorithm actually adopted. Sentiment analysis can determine the emotional tendency expressed in the text, including positive, negative and neutral, and provide emotional reference for the writing generation module. s , whose value range is [-1,1], -1 means completely negative, 1 means completely positive, and 0 means neutral. E is calculated by sentiment analysis algorithm s ,The formulas in the specific algorithm vary depending on the technology used. Syntax analysis can detect grammatical errors in text and provide correction suggestions. Let the number of grammatical errors be N e , the text length is L t , grammatical error density D e The calculation formula is:

4. The computer intelligent writing system driven by artificial intelligence according to claim 3, characterized in that: The machine learning module includes a neural network model, such as a recurrent neural network (RNN), a long short-term memory network (LSTM), etc., which can generate text that conforms to language habits and logic through training with a large amount of text data. In LSTM, the calculation formula of the forget gate is f t =σ(W f ·[h t-1 , x t ]+b f ), the input gate calculation formula is i t =σ(W i ·[h t-1 , x t ]+b i ), the candidate cell state calculation formula is: The cell state calculation formula is: The output gate calculation formula is o t =σ(W o ·[h t-1 , x t ]+b o ), the hidden state calculation formula is h t =ot*tanh(C t ), where δ is the sigmoid function, Wf, Wi, Wc, Wo are weight matrices, and bf, bi, bc, bo are bias terms. This module also has the function of model update and optimization. With the continuous input of new data, the performance of the writing system can be continuously improved. Let the model performance evaluation index be Pm, and the impact factor of new data on model performance be (the calculation method is determined according to the actual situation), then the updated model performance is, where ΔP is the performance change brought about by the new data.

5. The computer intelligent writing system driven by artificial intelligence according to claim 4, characterized in that: The writing generation module supports a variety of writing styles, including news reports, essays, poems, scientific papers, etc. Users can choose according to their needs. Suppose the feature vectors of different writing styles are (k represents different styles, such as news reports Prose for When the user selects a certain style, the writing generation module adjusts the text generation according to the corresponding feature vector. The specific adjustment method can be performed according to the formula involved in the actual algorithm. For example, in the vocabulary selection and sentence structure of the generated sentence, operations are performed according to the association formula between the style feature vector and the text generation rule. This module can automatically generate an article outline based on the keywords and topics provided by the user, and fill in and create content according to the outline. Let the keyword set be K = {k1, k2, ..., k n }, the set of concepts related to the topic is T = {t1, t2, ..., t m }, the outline generation algorithm generates an article outline according to the association between keywords and subject concepts (calculated through relevant formulas, such as based on semantic similarity, etc.).

6. The computer intelligent writing system driven by artificial intelligence according to claim 5, characterized in that: The content optimization module includes functions such as spelling check, grammar check, vocabulary replacement, sentence fluency optimization, etc. In vocabulary replacement, let the original vocabulary be W o , replace the word with W r , the evaluation index of the suitability of vocabulary replacement is R v , whose calculation formula can be determined based on factors such as the semantic similarity of words and contextual adaptability, such as in is the semantic similarity between the original word and the replacement word, In order to adapt the replacement vocabulary in the current context, the specific form of the function g is determined according to the actual situation. It can also adjust the language style and expression of the text according to different writing purposes and audiences, making the text more targeted and attractive. Suppose the writing purpose and audience feature vectors are The text style adjustment parameter vector is Then the relationship between them can be expressed by a mapping function F: The text style is adjusted according to the mapping relationship.

7. An artificial intelligence driven computer intelligent writing system according to claims 1-6, characterized in that: The user enters the writing topic, style, word count and other requirements through the system interface; The data collection module collects relevant text data, and the text analysis module analyzes the data; The machine learning module performs learning and training, and the writing generation module generates text content; The content optimization module optimizes the text and finally presents the optimized text to the user. The running time of each module in the whole process is T total It can be expressed as the sum of the running time of each module, that is, T total =T datacollection +T texctanalysis +T machinelearning +T uritinggeneration +T contentoptimization , where the specific calculation of the running time of each module can be determined based on its internal operations and the formulas involved.

8. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the functions of the artificial intelligence-driven computer intelligent writing system as described in any one of claims 1 to 6, or implements the method of using the system as described in claim 7.

Images