Method for creating item matching condition control model generation result based on large model

Abstract

The invention provides a method for creating an item matching condition control model generation result based on a large model. The method comprises an item capturing module, an item splitting module, a condition expansion module, a model synchronization module, a reasoning control module, a result output module, a timing updating module and a reasoning monitoring module. A full-privatized operation ecology is created, user demand statements are analyzed through an artificial intelligence technology, a model is accurately matched, then completed items are pointed to corresponding item sources, real contents and intentions conveyed by the statements are accurately grasped, and convenience is brought to affair handling of the user; standard files are automatically collected through the RPA robot and added into the knowledge base, and therefore the knowledge base content of the government affair question-answering system is continuously maintained and enriched; according to the method, service items are collected, a service item list is formed, conditional expansion is carried out on the service items, conditional nature and description are carried out on the items, and therefore accurate matching and efficient management of service item conditions are achieved.

Description

Technical Field

[0001] The present invention relates to the field of large models, and in particular to a method for controlling model generation results based on large model creation event matching conditions. Background Art

[0003] With the continuous breakthroughs in natural language processing and large language model technology, the field of intelligent question answering has received widespread attention. Intelligent question answering is an important application in the field of artificial intelligence. It can automatically reply to questions raised by users and help users quickly obtain the information or answers they need.

[0004] In the existing technology, government question-and-answer robots can only make some inferences about documents and knowledge content, but cannot guarantee the correctness of the documents; the retrieval of documents is still at a relatively rough "keyword" matching and retrieval level, and cannot perform accurate retrieval through semantic analysis and understanding; there is no accurate retrieval and guidance for service items. Summary of the invention

[0005] The purpose of the present invention is to provide a method for controlling model generation results based on large model creation event matching conditions, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for creating a result of a control model based on a large model to match an event, characterized in that it includes an event capture module, an event splitting module, a condition expansion module, a model synchronization module, an inference control module, a result output module, a timing update module, and an inference monitoring module. The item capture module is used to set different RPA robots, and the RPA robots capture the data of service items according to the different set tasks; The item splitting module is used to set the splitting rules of service items and split the service items according to different splitting rules; The condition expansion module is used to analyze service items, refine and explain the specific meaning and content of the items, and form a series of corresponding condition contents; The model synchronization module synchronizes the maintained items into the reasoning model; The inference control module performs semantic analysis on the input content, extracts the content and matches it with the item list; The result output module displays the details of the matched items; The timing update module regularly obtains a list of service items, disassembles the service items, compares them with items in the database, and updates the changed items to the model module; The inference monitoring module verifies each item after the content of the timing update module is updated.

[0007] Preferably, the content captured by the matter capture module includes the category to which the matter belongs, the name of the matter, the description of the matter, the content of the web page, and the corresponding link address.

[0008] Preferably, the splitting rules of the matter splitting module include scope, establishment basis, approval conditions, application materials, handling method, handling process, and handling time limit. The names of the split matters are named in the order of original matter name + splitting rules to form a new matter name, and the original web page content is saved and stored in the database together with the content of the question and answer of the matter name.

[0009] Preferably, the condition expansion module analyzes the content of the service items including the specific meaning of the items, qualification conditions, legal and compliance conditions, economic ability conditions, service demand conditions, technical conditions, time and location conditions, and other specific conditions.

[0010] Preferably, when the model synchronization module synchronizes the split content of the service items, the conditional content after the item name is disassembled is disassembled according to ';' and organized into multiple matching item information, and the spliced ​​items and the content of the items are synchronized to the large model.

[0011] Preferably, the inference control module extracts content and matches it with the item list, including the following steps: Step 1: Perform semantic analysis on the input content to extract the keywords in the sentence and the importance between the keywords; Step 2: Match the extracted important content with the item list. After matching, arrange the matched items from large to small according to the item threshold, where threshold = title similarity*M+content similarity*N. The user adjusts the accuracy of item matching by setting the value of the reference threshold K. After matching, compare the maximum threshold of the matched service items with the reference threshold K. If the threshold>K, the content of the web page is directly displayed; if the threshold is less than K, the service item with the largest title similarity is used as a reference recommendation item.

[0012] Preferably, the result output module arranges the matched results from high to low according to the similarity, and attaches the web link corresponding to the content after the displayed content. Clicking the link directly jumps to the designated item on the government service website to directly apply for and handle the business.

[0013] Preferably, when the timed update module disassembles the service item and compares it with the items in the database, it performs MD5 on the title and content of the service item to obtain its HASH value, and compares it with the items in the database. If they are different, they are updated to the item library and the version number + 1 is stored as the new version. After the synchronous collection is completed, the items with higher version numbers are updated to the model module.

[0014] Preferably, the inference monitoring module sets a verification list of high-frequency used items for the item services in the system according to the daily matching frequency. When the content of the scheduled update module is updated, the updated items are verified one by one. After the verification is successful, the results are compared to verify whether the data is synchronized successfully and whether the model is running normally.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention creatively stores the comparison items locally, creating a fully private operation ecosystem. It uses artificial intelligence technology to analyze user demand statements, and after accurate model matching, it directs the completed items to the corresponding item sources, accurately grasping the real content and intention conveyed by the statement, so as to facilitate users to do things; (2) The present invention expands and interprets localized items, creates unique matching content for each item, and expands the scope and meaning of item description to the greatest extent, including: time, place, name (alias), type, condition, qualification, compliance, ability, demand, technology, and other specific content categories. This enables rapid learning and accurate understanding of the semantic understanding of the large model, greatly improving the item comparison effect.

[0016] (3) The present invention automatically collects standard documents through RPA robots and adds them to the knowledge base, thereby continuously maintaining and enriching the knowledge base content of the government question and answer system; (4) The present invention collects service items, forms a service item list, and expands the conditions of the service items, and performs conditional properties and descriptions on the items, thereby achieving accurate matching and efficient management of service item conditions. DETAILED DESCRIPTION

[0017] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solution 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.

[0018] The method for creating a result of a conditional control model based on a large model consists of a matter capture module, a matter splitting module, a conditional expansion module, a model synchronization module, a reasoning control module, a result output module, a timed update module, and a reasoning monitoring module. The matter capture module is used to set different RPA robots, and the RPA robots capture the data of service matters according to the different tasks set; The item splitting module is used to set the splitting rules of service items and split the service items according to different splitting rules; The condition expansion module is used to analyze service items, refine and explain the specific meaning and content of the items, and form a series of corresponding condition contents; The model synchronization module synchronizes the maintained items to the inference model; The reasoning control module performs semantic analysis on the input content, extracts the content and matches it with the list of items; The result output module displays the details of the matched items; The scheduled update module obtains the service item list regularly, disassembles the service items and compares them with the items in the database, and updates the changed items to the model module; The reasoning monitoring module verifies each item after the content of the timing update module is updated.

[0019] The implementation steps of the method for controlling the model generation results based on the matching conditions of the large model creation items are as follows: Step S1: The matter capture module sets different RPA robots. The RPA robots capture the data of service matters according to the different set tasks. The captured content includes the category to which the matter belongs, the name of the matter, the description of the matter, the content of the web page, and the corresponding link address.

[0020] Step S2: The item splitting module sets the splitting rules for service items, and splits the service items according to different splitting rules. The item splitting rules include scope, establishment basis, approval conditions, application materials, handling method, handling process, and handling time limit. The name of the item after splitting is named in the order of original item name + splitting rules to form a new item name, and the original web page content is saved and stored in the database together with the content of the question and answer of the item name.

[0021] Step S3: The condition expansion module analyzes the service items, including the specific meaning of the items, qualification conditions, legal and compliance conditions, economic ability conditions, service demand conditions, technical conditions, time and location conditions, and other specific conditions. The specific meaning and content of the items are refined and explained to form a series of corresponding conditional contents; the localized items are expanded and explained, and unique matching content is created for each item, which maximizes the scope and meaning of the item description, including: time, place, name (alias), type, conditions, qualifications, compliance, ability, demand, technology, and other specific content categories, thereby achieving rapid learning and accurate understanding of the semantic understanding of the large model, greatly improving the item comparison effect.

[0022] Step S4: The model synchronization module synchronizes the maintained items to the inference model. When synchronizing the split content of the service items, the conditional content after the item name is disassembled is disassembled according to ';' and organized into multiple matching item information, and the spliced ​​items and their contents are synchronized to the large model.

[0023] Step S5: The reasoning control module performs semantic analysis on the input content, extracts the content and matches it with the item list. When matching, the input content is firstly semantically analyzed to extract the keywords in the sentence and the importance between the keywords; The extracted important content is then matched with the item list. After the match, the matched items are arranged from large to small according to the item threshold, where threshold = title similarity * M + content similarity * N. The user adjusts the accuracy of item matching by setting the value of the reference threshold K. After the match, the maximum threshold of the matched service items is compared with the reference threshold K. If the threshold > K, the content of the web page is directly displayed; if the threshold is less than K, the service item with the largest title similarity is used as a reference recommendation item.

[0024] Step S6: The result output module displays the detailed contents of the matched items, arranges the matched results from high to low according to the similarity, and attaches the corresponding web link to the displayed content. Click the link to jump directly to the designated item on the government service website to apply for and handle the business directly.

[0025] Step S7: The scheduled update module periodically obtains the service item list, disassembles the service item and compares it with the items in the database, performs MD5 on the title and content of the service item to obtain its HASH value, and compares it with the items in the database. If they are different, it will be updated to the item library and the version number +1 will be stored as the new version. After the synchronous collection is completed, the items with higher version numbers will be updated to the model module.

[0026] Step S8: The reasoning monitoring module verifies the items one by one after the content of the scheduled update module is updated, and sets a verification list of high-frequency items for the item services in the system according to the daily matching frequency. When the content of the scheduled update module is updated, the updated items are verified one by one. After the verification is successful, the results are compared to verify whether the data is synchronized successfully and whether the model is running normally.

[0027] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present invention.

Claims

1. A method for controlling model generation results based on a large model creation event matching condition, characterized in that: It includes event capture module, event splitting module, condition expansion module, model synchronization module, reasoning control module, result output module, timed update module and reasoning monitoring module. The item capture module is used to set different RPA robots, and the RPA robots capture the data of service items according to the different set tasks; The item splitting module is used to set the splitting rules of service items and split the service items according to different splitting rules; The condition expansion module is used to analyze service items, refine and explain the specific meaning and content of the items, and form a series of corresponding condition contents; The model synchronization module synchronizes the maintained items into the reasoning model; The inference control module performs semantic analysis on the input content, extracts the content and matches it with the item list; The result output module displays the details of the matched items; The timing update module regularly obtains a list of service items, disassembles the service items, compares them with items in the database, and updates the changed items to the model module; The inference monitoring module verifies each item after the content of the timing update module is updated.

2. The method for controlling model generation results based on large model creation items matching conditions according to claim 1, characterized in that: The content captured by the item capture module includes the category to which the item belongs, the name of the item, the description of the item, the content of the web page, and the corresponding link address.

3. The method for controlling model generation results based on large model creation items matching conditions according to claim 1, characterized in that: The splitting rules of the matter splitting module include scope, establishment basis, approval conditions, application materials, handling method, handling process, and handling time limit. The names of the split matters are named in the order of original matter name + splitting rules to form a new matter name, and the original web page content is saved and stored in the database together with the content of the question and answer of the matter name.

4. The method for controlling model generation results based on large model creation items matching conditions according to claim 1, characterized in that: The condition expansion module analyzes the content of the service items including the specific meaning of the items, qualification conditions, legal and compliance conditions, economic ability conditions, service demand conditions, technical conditions, time and location conditions, and other specific conditions.

5. The method for controlling model generation results based on large model creation items matching conditions according to claim 1, characterized in that: When the model synchronization module synchronizes the content of the service item split, the conditional content after the item name is split is split according to ';' and organized into multiple matching item information, and the spliced ​​items and the content of the items are synchronized to the large model.

6. The method for controlling model generation results based on large model creation items matching conditions according to claim 1, characterized in that: The inference control module extracts content and matches it with the item list, including the following steps: Step 1: Perform semantic analysis on the input content to extract the keywords in the sentence and the importance between the keywords; Step 2: Match the extracted important content with the item list. After matching, arrange the matched items from large to small according to the item threshold, where threshold = title similarity*M+content similarity*N. The user adjusts the accuracy of item matching by setting the value of the reference threshold K. After matching, compare the maximum threshold of the matched service items with the reference threshold K. If the threshold > K, the content of the web page is directly displayed; if the threshold is less than K, the service item with the largest title similarity is used as a reference recommendation item.

7. The method for controlling model generation results based on large model creation items matching conditions according to claim 1, characterized in that: The result output module arranges the matched results from high to low according to the similarity, and attaches the web link corresponding to the content after the displayed content. Clicking the link directly jumps to the designated item on the government service website to directly apply for and handle the business.

8. The method for controlling model generation results based on large model creation items matching conditions according to claim 1, characterized in that: When the scheduled update module disassembles the service items and compares them with the items in the database, it performs MD5 on the title and content of the service items to obtain their HASH values, and compares them with the items in the database. If they are different, they are updated to the item library and the version number + 1 is stored as the new version. After the synchronous collection is completed, the items with higher version numbers are updated to the model module.

9. The method for controlling model generation results based on large model creation items matching conditions according to claim 1, characterized in that: The reasoning monitoring module sets a verification list of high-frequency items for the item services in the system according to the daily matching frequency. When the content of the scheduled update module is updated, the updated items are verified one by one. After the verification is successful, the results are compared to verify whether the data is synchronized successfully and whether the model is running normally.