Intelligent form creation system and method

Abstract

The invention discloses an intelligent form creation system and method, and relates to the technical field of information processing. The method comprises the following steps: receiving a creation instruction of a current user terminal, and extracting industry information of a user from a database according to the creation instruction; substituting the industry information into the feature conversion model to obtain text features; obtaining each historical text feature corresponding to the historical user, sequentially performing similarity matching on the text feature and each historical text feature to obtain a similar user set, determining a similar form set according to the similar user set, and generating a historical form according to the similar form set; substituting the text features into a target generation model to generate an SQL statement, and correcting a historical form through the SQL statement to obtain a current latest form; updating the historical form by using the current newest form; and sending the current newest form to the current user terminal and confirming the current newest form. According to the invention, the efficient multiplexing and intelligent updating of the form are realized, the overuse of the server is reduced, and the form accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, specifically to an intelligent form creation system and method. Background Technology

[0002] In many industry application scenarios, forms are an important tool for information collection and display. Traditional electronic forms rely on manual dragging and dropping of controls and configuring validation rules item by item. This requires repeated verification of field definitions and business logic, resulting in long cycles and high error rates.

[0003] Existing technology (CN114492351A) discloses a method and system for constructing intelligent encrypted forms for business process processing. The method includes: constructing a model for business process processing with a three-layer logical structure of template-block-field; creating several data blocks, each containing several fields; connecting the data blocks to determine an intelligent form template; during the business process flow, filling in data according to the intelligent form template, and at each stage determining whether the value corresponding to the field in the data block can be modified; if the value corresponding to the field can be modified, performing digital signature and data encryption calculations. This invention, in the process of creating intelligent forms, creates several data blocks and connects them to form a form template, thereby achieving the purpose of easy creation, adjustment, and modification, accelerating business process flow, and improving the organization's agile management needs.

[0004] However, in practical applications, the above solution mainly addresses the issue of changing form requirements. When a large number of user-requested forms are created, it can overload the server, reduce the accuracy of form creation, and consequently decrease the user experience. Summary of the Invention

[0005] The purpose of this invention is to solve the problem by proposing an intelligent form creation system and method.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] First, a smart form creation method is proposed, which includes:

[0008] Receive the creation instruction of the current user terminal, extract the user's industry information from the database according to the creation instruction; convert the industry information into industry text, and substitute the industry text into the feature transformation model to obtain text features;

[0009] Obtain the historical text features corresponding to each historical user, and perform similarity matching between the text features and each historical text feature in turn to obtain each similar user.

[0010] A set of similar users is determined based on each similar user, and a set of similar forms is determined based on the set of similar users. The set of similar forms consists of forms generated by each user. Historical forms are generated based on the set of similar forms.

[0011] Text features are fed into the target generation model to generate SQL statements. Historical forms are corrected using SQL statements to obtain the latest current form. The latest current form is then used to update the historical forms.

[0012] The latest form is sent to the current user's terminal, where the user confirms the latest form.

[0013] Optionally, substituting the industry text into the feature transformation model to obtain text features includes:

[0014] Preprocessing is performed on the industry text to obtain individual subtexts, and each subtext is defined as a text set; the text sets are then subjected to text optimization operations to obtain an optimized text sequence.

[0015] The optimized text sequence is imported into the word embedding layer to obtain the feature vector. The feature vector is then processed through a convolutional layer, an activation layer, and a GRU layer to obtain the first feature vector, which is denoted as the text feature.

[0016] Optionally, similar users can be obtained by sequentially performing similarity matching between the text features and various historical text features, including:

[0017] S1: Calculate the similarity value between the text features and historical text features using cosine similarity;

[0018] S2: If the similarity value is greater than the preset similarity threshold, then obtain the user corresponding to the historical text feature and record it as a similar user;

[0019] S3: Based on steps S1 and S2, iterate through each historical text feature to obtain each similar user that meets the judgment condition.

[0020] Optionally, incorporating text features into the target generation model to generate SQL statements includes:

[0021] Extract keywords from text features, predict key columns based on keywords and column name information of data tables, and obtain information corresponding to key columns; the column name information of the data tables is obtained from the database.

[0022] The information corresponding to the key columns is imported into the variable suggestion function to obtain variable suggestion content. The variable suggestion content is then converted into JSON format, and an SQL statement that meets the requirements is automatically generated based on the JSON format.

[0023] Optionally, updating historical forms with the latest form includes:

[0024] The current form is merged into a set of similar forms, and the column name information of each form in the set of similar forms is extracted; the keywords corresponding to each column name information are obtained; and each keyword is put into a sequence to obtain a key sequence.

[0025] Keyword sequences with similar occurrences are categorized to obtain classified key sequences;

[0026] The number of occurrences of each keyword in the categorized key sequence is obtained by statistical analysis.

[0027] The percentage of each keyword can be obtained by comparing the number of occurrences of each keyword with the total number of forms.

[0028] The key percentage is compared with the preset key percentage threshold in turn. If the key percentage is greater than the preset key percentage threshold, the column name information corresponding to the keyword is merged into the historical form, and the historical form is finally obtained.

[0029] The process of categorizing key sequences by grouping similar keywords into key sequences includes the following steps:

[0030] A1: Use the first keyword in the key sequence as the initial keyword;

[0031] A2: Perform similarity algorithms on the initial keyword and other keywords in turn to obtain the fusion degree; if the fusion degree is greater than the fusion threshold, then classify the initial keyword and the compared keyword into the same category, take the next keyword of the key sequence as the initial keyword, and return to step A1;

[0032] If the fusion degree is less than or equal to the fusion threshold, then the next keyword of the key sequence is taken as the initial keyword, and the process returns to step A1.

[0033] A3: When all keywords in the key sequence have been traversed, the traversal ends and the categorized key sequence is obtained.

[0034] A smart form creation system is proposed, comprising:

[0035] Text Feature Module: Receives the creation instruction from the current user terminal, extracts the user's industry information from the database according to the creation instruction, converts the industry information into industry text, and substitutes the industry text into the feature conversion model to obtain text features;

[0036] User matching module: Obtain the historical text features corresponding to historical users, and perform similarity matching between the text features and each historical text feature in turn to obtain each similar user;

[0037] Historical form generation module: Determines a set of similar users based on each similar user, determines a set of similar forms based on the set of similar users, the set of similar forms consists of forms generated by each user, and generates historical forms based on the set of similar forms.

[0038] Latest Form Module: Inputs text features into the target generation model to generate SQL statements, corrects historical forms using SQL statements to obtain the current latest form, and updates historical forms with the current latest form;

[0039] User confirmation module: Sends the latest form to the current user's terminal, where the user confirms the latest form.

[0040] Optionally, the text feature module includes: an optimized text sequence module and a text feature module.

[0041] The optimized text sequence module is used to perform preprocessing operations on industry text to obtain individual sub-texts, and to determine each sub-text as a text set; the text set is then subjected to text optimization operations to obtain an optimized text sequence.

[0042] The text feature module is used to import the optimized text sequence into the word embedding layer to obtain a feature vector, and then process the feature vector through a convolutional layer, an activation layer, and a GRU layer to obtain a first feature vector, which is denoted as the text feature.

[0043] Optional, the user matching module includes:

[0044] S1: Calculate the similarity value between the text features and historical text features using cosine similarity;

[0045] S2: If the similarity value is greater than the preset similarity threshold, then obtain the user corresponding to the historical text feature and record it as a similar user;

[0046] S3: Based on steps S1 and S2, iterate through each historical text feature to obtain each similar user that meets the judgment condition.

[0047] Optional, the latest form modules include:

[0048] Extract keywords from text features, predict key columns based on keywords and column name information of data tables, and obtain information corresponding to key columns; the column name information of the data tables is obtained from the database.

[0049] The information corresponding to the key columns is imported into the variable suggestion function to obtain variable suggestion content. The variable suggestion content is then converted into JSON format, and an SQL statement that meets the requirements is automatically generated based on the JSON format.

[0050] Optionally, the user confirmation module includes:

[0051] The current form is merged into a set of similar forms, and the column name information of each form in the set of similar forms is extracted; the keywords corresponding to each column name information are obtained; and each keyword is put into a sequence to obtain a key sequence.

[0052] Keyword sequences with similar occurrences are categorized to obtain classified key sequences;

[0053] The number of occurrences of each keyword in the categorized key sequence is obtained by statistical analysis.

[0054] The percentage of each keyword can be obtained by comparing the number of occurrences of each keyword with the total number of forms.

[0055] The key percentage is compared with the preset key percentage threshold in turn. If the key percentage is greater than the preset key percentage threshold, the column name information corresponding to the keyword is merged into the historical form, and the historical form is finally obtained.

[0056] The process of categorizing key sequences by grouping similar keywords into key sequences includes the following steps:

[0057] A1: Use the first keyword in the key sequence as the initial keyword;

[0058] A2: Perform similarity algorithms on the initial keyword and other keywords in turn to obtain the fusion degree; if the fusion degree is greater than the fusion threshold, then classify the initial keyword and the compared keyword into the same category, take the next keyword of the key sequence as the initial keyword, and return to step A1;

[0059] If the fusion degree is less than or equal to the fusion threshold, then the next keyword of the key sequence is taken as the initial keyword, and the process returns to step A1.

[0060] A3: When all keywords in the key sequence have been traversed, the traversal ends and the categorized key sequence is obtained.

[0061] The beneficial effects of this invention are:

[0062] This invention proposes an intelligent form creation method. It receives a creation instruction from the current user terminal, extracts the user's industry information from a database based on the instruction, inputs the industry information into a feature transformation model to obtain text features, acquires historical text features corresponding to historical users, performs similarity matching between the text features and each historical text feature sequentially to obtain a set of similar users, determines a set of similar forms based on the set of similar users, and generates historical forms based on the set of similar forms. It then inputs the text features into a target generation model to generate an SQL statement, modifies the historical forms using the SQL statement to obtain the latest current form, updates the historical forms with the latest current form, and sends the latest current form to the current user terminal for confirmation. This invention achieves efficient form reuse and intelligent updating, reduces excessive server usage, improves form accuracy, and enhances the user experience. Attached Figure Description

[0063] Figure 1 A flowchart illustrating an intelligent form creation method provided in an embodiment of the present invention;

[0064] Figure 2 This is a framework diagram of an intelligent form creation system provided in an embodiment of the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] This invention provides a method for creating intelligent forms. See also... Figure 1 , Figure 1 A flowchart illustrating a smart form creation method provided in an embodiment of the present invention. The method includes the following steps:

[0067] Receive the creation instruction of the current user terminal, extract the user's industry information from the database according to the creation instruction; convert the industry information into industry text, and substitute the industry text into the feature transformation model to obtain text features;

[0068] Obtain the historical text features corresponding to each historical user, and perform similarity matching between the text features and each historical text feature in turn to obtain each similar user.

[0069] A set of similar users is determined based on each similar user; a set of similar forms is determined based on the set of similar users; the set of similar forms consists of forms generated by each user; and historical forms are generated based on the set of similar forms.

[0070] Text features are fed into the target generation model to generate SQL statements. Historical forms are corrected using SQL statements to obtain the latest current form. The latest current form is then used to update the historical forms.

[0071] The latest form is sent to the current user's terminal, where the user confirms the latest form.

[0072] It should be noted that industry information includes, but is not limited to, various fields such as education and training, finance, healthcare, education, and e-commerce, and each industry information has a corresponding set of forms; text features include key information corresponding to each industry; information for each user corresponding to each industry; and SQL statements are usually used specifically for managing data in relational databases.

[0073] Based on the intelligent form creation method provided by the embodiments of the present invention, the method extracts industry text features and matches similar users, generates historical forms from a set of similar forms, and then outputs SQL statements through the generation model to accurately correct the forms. This achieves efficient form reuse and intelligent updates, reduces excessive server usage, improves form accuracy, and enhances user experience.

[0074] In one implementation, the text features obtained by substituting industry text into a feature transformation model include:

[0075] Preprocessing is performed on the industry text to obtain individual subtexts, and each subtext is defined as a text set; the text sets are then subjected to text optimization operations to obtain an optimized text sequence.

[0076] The optimized text sequence is imported into the word embedding layer to obtain the feature vector. The feature vector is then processed through a convolutional layer, an activation layer, and a GRU layer to obtain the first feature vector, which is denoted as the text feature.

[0077] In one implementation, the text optimization operation is specifically calculated as follows:

[0078]

[0079] Among them, WF i This represents the term frequency minus the inverse text frequency (IQF) of the i-th word, where N represents the total number of words in the text. i W represents the number of times word i appears in the text set, and W represents the total number of texts. i S represents the number of texts containing the word i; i D represents the score corresponding to the i-th word. i The value represents the category density corresponding to the i-th word; M represents the total number of text categories, n k i This represents the number of times word i appears in category k; it should be noted that D iThe larger the value, the more dispersed the distribution of the word, and the more obvious its characteristics.

[0080] In one implementation, industry texts are segmented through preprocessing operations, and the various texts are combined into a text set. Then, text optimization, such as Chinese word segmentation, obfuscation mapping, and stop word removal, is performed to ensure the semantic accuracy of subsequent inputs. The word embedding layer maps discrete characters into dense vectors, which compresses dimensionality while preserving industry word meanings. The convolutional layer captures key phrase patterns with local receptive fields, and the activation layer enhances nonlinear expression. The GRU layer further models long-range dependencies and outputs a first feature vector that combines local details and global context.

[0081] In one implementation, similar users are obtained by sequentially performing similarity matching between text features and various historical text features, including:

[0082] S1: Calculate the similarity value between the text features and historical text features using cosine similarity;

[0083] S2: If the similarity value is greater than the preset similarity threshold, then obtain the user corresponding to the historical text feature and record it as a similar user;

[0084] S3: Based on steps S1 and S2, iterate through each historical text feature to obtain each similar user that meets the judgment condition.

[0085] In one implementation, cosine similarity is used to quickly compare text features with historical data to accurately identify highly similar users, avoiding the inefficiency and errors of manual screening and improving the accuracy of similar users.

[0086] In one implementation, substituting text features into the target generation model to generate SQL statements includes:

[0087] Extract keywords from text features, predict key columns based on keywords and column name information of data tables, and obtain information corresponding to key columns; column name information of data tables is obtained from the database.

[0088] The information corresponding to the key columns is imported into the variable suggestion function to obtain variable suggestion content. The variable suggestion content is then converted into JSON format, and an SQL statement that meets the requirements is automatically generated based on the JSON format.

[0089] In one implementation, the variable hint function is specifically as follows:

[0090]

[0091] Where Pv represents variable hint content, colname represents the information corresponding to the key column, I represents the specific description of the data table column name information; f() represents the combination of the information corresponding to the key column and the specific description of the data table column name information to generate variable hint content.

[0092] In one implementation, key columns are predicted based on the keywords and column names of the data table. Specifically, the key columns are obtained by importing the keyword and column name information into the output of a BERT fine-tuned classification model. The BERT fine-tuned classification model is based on a stack of multiple Transformer architecture encoder blocks, each of which contains a multi-head self-attention mechanism and a feedforward neural network. The multi-head self-attention mechanism allows the model to capture the relationships between words in the text from different perspectives, thereby learning rich semantic information. The feedforward neural network further transforms the features output by the self-attention mechanism, enhancing the expressive power of the features.

[0093] In one implementation, the variable prompt content describes the natural language instructions for querying key columns; the efficient JSON format is preferred as the primary output medium, which helps reduce information redundancy and improve data readability and operability. Specifically, the JSON format can provide a clear definition and order for key columns, determining the specific position of key columns in the SQL statement.

[0094] One implementation method for updating historical forms with the latest form includes:

[0095] The current form is merged into a set of similar forms, and the column name information of each form in the set of similar forms is extracted; the keywords corresponding to each column name information are obtained; and each keyword is put into a sequence to obtain a key sequence.

[0096] Keyword sequences with similar occurrences are categorized to obtain classified key sequences;

[0097] The number of occurrences of each keyword in the categorized key sequence is obtained by statistical analysis.

[0098] The percentage of each keyword can be obtained by comparing the number of occurrences of each keyword with the total number of forms.

[0099] The key percentage is compared with the preset key percentage threshold in turn. If the key percentage is greater than the preset key percentage threshold, the column name information corresponding to the keyword is merged into the historical form, and the historical form is finally obtained.

[0100] The process of categorizing key sequences by grouping similar keywords into key sequences includes the following steps:

[0101] A1: Use the first keyword in the key sequence as the initial keyword;

[0102] A2: Perform similarity algorithms on the initial keyword and other keywords in turn to obtain the fusion degree; if the fusion degree is greater than the fusion threshold, then classify the initial keyword and the compared keyword into the same category, take the next keyword of the key sequence as the initial keyword, and return to step A1;

[0103] If the fusion degree is less than or equal to the fusion threshold, then the next keyword of the key sequence is taken as the initial keyword, and the process returns to step A1.

[0104] A3: When all keywords in the key sequence have been traversed, the traversal ends and the categorized key sequence is obtained.

[0105] In one implementation, similar forms are automatically merged, keywords are extracted and categorized uniformly, and matching forms are quickly located; this standardizes the form structure, significantly reduces manual processing costs, and improves data accuracy.

[0106] Based on the same inventive concept, this invention also provides an intelligent form creation system. See [link to related document]. Figure 2 , Figure 2 A framework diagram of an intelligent form creation system provided in an embodiment of the present invention includes:

[0107] The intelligent form creation system provided by this invention extracts industry text features and matches similar users, generates historical forms from a set of similar forms, and then outputs SQL statements through a generation model to accurately correct the forms. This achieves efficient form reuse and intelligent updates, reduces excessive server usage, improves form accuracy, and enhances the user experience.

[0108] It should be noted that, in this document, terms such as “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0109] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for creating intelligent forms, characterized in that, The method includes: Receive the creation instruction of the current user terminal, extract the user's industry information from the database according to the creation instruction; convert the industry information into industry text, and substitute the industry text into the feature transformation model to obtain text features; Obtain the historical text features corresponding to each historical user, and perform similarity matching between the text features and each historical text feature in turn to obtain each similar user. A set of similar users is determined based on each similar user, and a set of similar forms is determined based on the set of similar users. The set of similar forms consists of forms generated by each user. Historical forms are generated based on the set of similar forms. Text features are fed into the target generation model to generate SQL statements. Historical forms are corrected using SQL statements to obtain the latest current form. The latest current form is then used to update the historical forms. The latest form is sent to the current user's terminal, where the user confirms the latest form.

2. The intelligent form creation method according to claim 1, characterized in that, The step of substituting the industry text into the feature transformation model to obtain text features includes: Preprocessing is performed on the industry text to obtain individual subtexts, and each subtext is defined as a text set; the text sets are then subjected to text optimization operations to obtain an optimized text sequence. The optimized text sequence is imported into the word embedding layer to obtain the feature vector. The feature vector is then processed through a convolutional layer, an activation layer, and a GRU layer to obtain the first feature vector, which is denoted as the text feature.

3. The intelligent form creation method according to claim 1, characterized in that, The step of sequentially matching text features with various historical text features to obtain similar users includes: S1: Calculate the similarity value between the text features and historical text features using cosine similarity; S2: If the similarity value is greater than the preset similarity threshold, then obtain the user corresponding to the historical text feature and record it as a similar user; S3: Based on steps S1 and S2, iterate through each historical text feature to obtain each similar user that meets the judgment condition.

4. The intelligent form creation method according to claim 1, characterized in that, The step of substituting text features into the target generation model to generate SQL statements includes: Extract keywords from text features, predict key columns based on keywords and column name information of data tables, and obtain information corresponding to key columns; the column name information of the data tables is obtained from the database. The information corresponding to the key columns is imported into the variable suggestion function to obtain variable suggestion content. The variable suggestion content is then converted into JSON format, and an SQL statement that meets the requirements is automatically generated based on the JSON format.

5. The intelligent form creation method according to claim 1, characterized in that, The step of updating historical forms with the latest form includes: The current form is merged into a set of similar forms, and the column name information of each form in the set of similar forms is extracted; the keywords corresponding to each column name information are obtained; and each keyword is put into a sequence to obtain a key sequence. Keyword sequences with similar occurrences are categorized to obtain classified key sequences; The number of occurrences of each keyword in the categorized key sequence is obtained by statistical analysis. The percentage of each keyword can be obtained by comparing the number of occurrences of each keyword with the total number of forms. The key percentage is compared with the preset key percentage threshold in turn. If the key percentage is greater than the preset key percentage threshold, the column name information corresponding to the keyword is merged into the historical form, and the historical form is finally obtained. The process of categorizing key sequences by grouping similar keywords into key sequences includes the following steps: A1: Use the first keyword in the key sequence as the initial keyword; A2: Perform similarity algorithms on the initial keyword and other keywords in turn to obtain the fusion degree; if the fusion degree is greater than the fusion threshold, then classify the initial keyword and the compared keyword into the same category, take the next keyword of the key sequence as the initial keyword, and return to step A1; If the fusion degree is less than or equal to the fusion threshold, then the next keyword of the key sequence is taken as the initial keyword, and the process returns to step A1. A3: When all keywords in the key sequence have been traversed, the traversal ends and the categorized key sequence is obtained.

6. An intelligent form creation system, characterized in that, The system includes: Text Feature Module: Receives the creation instruction from the current user terminal, extracts the user's industry information from the database according to the creation instruction, converts the industry information into industry text, and substitutes the industry text into the feature conversion model to obtain text features; User matching module: Obtain the historical text features corresponding to historical users, and perform similarity matching between the text features and each historical text feature in turn to obtain each similar user; Historical form generation module: Determines a set of similar users based on each similar user, determines a set of similar forms based on the set of similar users, the set of similar forms consists of forms generated by each user, and generates historical forms based on the set of similar forms. Latest Form Module: Inputs text features into the target generation model to generate SQL statements, corrects historical forms using SQL statements to obtain the current latest form, and updates historical forms with the current latest form; User confirmation module: Sends the latest form to the current user's terminal, where the user confirms the latest form.

7. The intelligent form creation system according to claim 6, characterized in that, The text feature module includes: an optimized text sequence module and a text feature module. The optimized text sequence module is used to perform preprocessing operations on industry text to obtain individual sub-texts, and to determine each sub-text as a text set; the text set is then subjected to text optimization operations to obtain an optimized text sequence. The text feature module is used to import the optimized text sequence into the word embedding layer to obtain a feature vector, and then process the feature vector through a convolutional layer, an activation layer, and a GRU layer to obtain a first feature vector, which is denoted as the text feature.

8. The intelligent form creation system according to claim 6, characterized in that, The user matching module includes: S1: Calculate the similarity value between the text features and historical text features using cosine similarity; S2: If the similarity value is greater than the preset similarity threshold, then obtain the user corresponding to the historical text feature and record it as a similar user; S3: Based on steps S1 and S2, iterate through each historical text feature to obtain each similar user that meets the judgment condition.

9. The intelligent form creation system according to claim 6, characterized in that, The latest form module includes: Extract keywords from text features, predict key columns based on keywords and column name information of data tables, and obtain information corresponding to key columns; the column name information of the data tables is obtained from the database. The information corresponding to the key columns is imported into the variable suggestion function to obtain variable suggestion content. The variable suggestion content is then converted into JSON format, and an SQL statement that meets the requirements is automatically generated based on the JSON format.

10. The intelligent form creation system according to claim 6, characterized in that, The user confirmation module includes: The current form is merged into a set of similar forms, and the column name information of each form in the set of similar forms is extracted; the keywords corresponding to each column name information are obtained; and each keyword is put into a sequence to obtain a key sequence. Keyword sequences with similar occurrences are categorized to obtain classified key sequences; The number of occurrences of each keyword in the categorized key sequence is obtained by statistical analysis. The percentage of each keyword can be obtained by comparing the number of occurrences of each keyword with the total number of forms. The key percentage is compared with the preset key percentage threshold in turn. If the key percentage is greater than the preset key percentage threshold, the column name information corresponding to the keyword is merged into the historical form, and the historical form is finally obtained. The process of categorizing key sequences by grouping similar keywords into key sequences includes the following steps: A1: Use the first keyword in the key sequence as the initial keyword; A2: Perform similarity algorithms on the initial keyword and other keywords in turn to obtain the fusion degree; if the fusion degree is greater than the fusion threshold, then classify the initial keyword and the compared keyword into the same category, take the next keyword of the key sequence as the initial keyword, and return to step A1; If the fusion degree is less than or equal to the fusion threshold, then the next keyword of the key sequence is taken as the initial keyword, and the process returns to step A1. A3: When all keywords in the key sequence have been traversed, the traversal ends and the categorized key sequence is obtained.

Images