CN117788163A - Verification method and device for trade background, computer equipment and storage medium - Google Patents
Verification method and device for trade background, computer equipment and storage medium Download PDFInfo
- Publication number
- CN117788163A CN117788163A CN202311843787.3A CN202311843787A CN117788163A CN 117788163 A CN117788163 A CN 117788163A CN 202311843787 A CN202311843787 A CN 202311843787A CN 117788163 A CN117788163 A CN 117788163A
- Authority
- CN
- China
- Prior art keywords
- verification
- trade
- question
- document
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012795 verification Methods 0.000 title claims abstract description 368
- 238000000034 method Methods 0.000 title claims abstract description 105
- 230000008569 process Effects 0.000 claims abstract description 55
- 238000012216 screening Methods 0.000 claims abstract description 17
- 238000004590 computer program Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 3
- 230000001172 regenerating effect Effects 0.000 claims 1
- 238000013461 design Methods 0.000 description 7
- 230000005856 abnormality Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000012015 optical character recognition Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- PCTMTFRHKVHKIS-BMFZQQSSSA-N (1s,3r,4e,6e,8e,10e,12e,14e,16e,18s,19r,20r,21s,25r,27r,30r,31r,33s,35r,37s,38r)-3-[(2r,3s,4s,5s,6r)-4-amino-3,5-dihydroxy-6-methyloxan-2-yl]oxy-19,25,27,30,31,33,35,37-octahydroxy-18,20,21-trimethyl-23-oxo-22,39-dioxabicyclo[33.3.1]nonatriaconta-4,6,8,10 Chemical compound C1C=C2C[C@@H](OS(O)(=O)=O)CC[C@]2(C)[C@@H]2[C@@H]1[C@@H]1CC[C@H]([C@H](C)CCCC(C)C)[C@@]1(C)CC2.O[C@H]1[C@@H](N)[C@H](O)[C@@H](C)O[C@H]1O[C@H]1/C=C/C=C/C=C/C=C/C=C/C=C/C=C/[C@H](C)[C@@H](O)[C@@H](C)[C@H](C)OC(=O)C[C@H](O)C[C@H](O)CC[C@@H](O)[C@H](O)C[C@H](O)C[C@](O)(C[C@H](O)[C@H]2C(O)=O)O[C@H]2C1 PCTMTFRHKVHKIS-BMFZQQSSSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 230000008521 reorganization Effects 0.000 description 1
- 238000012502 risk assessment Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
- 108010064245 urinary gonadotropin fragment Proteins 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a verification method, a verification device, computer equipment and a storage medium of trade background, which are used for solving the problem that a large number of process nodes cannot be accurately and automatically verified under variable systems or regulations and complex verification processes. The method comprises the following steps: screening out trade documents related to verification targets, and extracting document elements in the trade documents; constructing a question-answer task containing document elements according to the verification target; acquiring a system text related to the verification target through a verification system corresponding to the verification target; and inputting the system text and the question-answering task into a first question-answering model based on prompt learning to obtain a verification conclusion of the verification target.
Description
Technical Field
The present invention relates to the field of financial science and technology, and in particular, to a verification method, apparatus, computer device and storage medium for trade background.
Background
The supply chain finance (Supply Chain Finance) is triggered integrally from the supply chain industry chain, integrates information such as logistics, fund flow, information flow and the like by using financial technology means, and constructs a financial supply system and a risk assessment system of which a core enterprise and an upstream enterprise are dominant in the supply chain under the background of real transaction so as to quickly respond to comprehensive requirements such as settlement, financing, financial management and the like of the enterprises on the supply industry chain. Thus, in the process of verifying trade background in supply chain finance, various complex supply and demand relationships, numerous business requirements and verification files with different formats need to be considered, which results in that the verification will have a complex verification process and a large number of process nodes. In addition, due to policy and environmental changes and adjustments, the associated regulations and regulations change, which results in various unpredictable problems such as obsolescence of verification files.
Because of the problems with the verification process, process nodes and regulations, the prior art performs verification on the trade background in real time, and usually needs to manually verify the process nodes one by one according to the existing regulations or regulations, and although part of verification functions can be realized by automated software, the software can only be used for assistance, and the verification of the whole trade background cannot be completed completely by automation.
Disclosure of Invention
The embodiment of the invention provides a verification method, a verification device, computer equipment and a storage medium for trade background, which are used for solving the problem that a large number of process nodes cannot be accurately and automatically verified under variable systems or regulations and complex verification processes.
In a first aspect of the present invention, there is provided a method of verifying trade context, comprising:
determining verification targets of trade backgrounds, and screening out trade documents related to the verification targets;
extracting document elements in the trade document;
constructing a question-answer task containing the document elements according to the verification target;
acquiring a system text related to the verification target through a verification system corresponding to the verification target;
and inputting the system text and the question-answering task into a first question-answering model based on prompt learning to obtain a verification conclusion of the verification target.
In one possible design, the extracting document elements in the trade document includes:
analyzing the trade document, and selecting a core segment of the trade document;
constructing a prompting word of the element extraction class;
and inputting the core segment and the prompt word into a second question-answer model based on prompt learning to obtain document elements in the trade document.
In one possible design, after the inputting the system text and the question-answering task into a first question-answering model based on prompt learning to draw a verification conclusion of the verification object, the method further includes:
obtaining a verification rule of the verification target according to the system text;
generating a verification flow conforming to the verification rule;
when any verification node in the verification process is modified, the subsequent verification process of the verification node is regenerated according to the verification rule.
In one possible design, the constructing a question-answer task including the document element according to the verification target includes:
constructing a judging question sentence whether the verification target is reached or not by using the document element;
and taking the judging question sentence as the question answering task.
In one possible design, before the acquiring, by the verification system corresponding to the verification target, the system text related to the verification target, the method further includes:
classifying all verification system documents to obtain classified documents belonging to different business categories;
according to the service type, screening out similar documents from the classified documents through similarity calculation;
recombining prompt words of different targets according to the prompt word template, the similar documents and the different verification targets;
based on the business type, training the first question-answering model based on prompt learning by using a large language model and prompt words of different targets.
In one possible design, the reorganizing the alert words of different targets according to the alert word template, the similar document and the different verification targets includes:
acquiring different versions of the similar document, and acquiring version documents in the different versions;
comparing the version documents in the different versions to obtain difference contents of the different versions;
and recombining prompt words of different targets according to the prompt word template, the difference content and the different verification targets.
In one possible design, after the alert words of different targets are recombined according to the alert word template, the difference content and the different verification targets, the method further includes:
respectively obtaining verification conclusions corresponding to different versions of documents according to each verification target;
and obtaining the conclusion differences of different verification conclusions through the verification conclusions corresponding to the difference content and the different version documents.
In a second aspect, there is provided a verification apparatus for trade contexts, comprising:
the screening module is used for determining a verification target of the trade background and screening out trade documents related to the verification target;
the extraction module is used for extracting document elements in the trade document;
the construction module is used for constructing a question-answer task containing the document elements according to the verification target;
the acquisition module is used for acquiring a system text related to the verification target through a verification system corresponding to the verification target;
and the output module is used for inputting the system text and the question-answer task into a first question-answer model based on prompt learning to obtain a verification conclusion of the verification target.
In a third aspect, a computer device is provided comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the verification method of trade background described above when the computer program is executed.
In a fourth aspect, a computer readable storage medium is provided, the computer readable storage medium storing a computer program which, when executed by a processor, performs the steps of the verification method of trade background described above.
The verification method, the verification device, the computer equipment and the storage medium of the trade background are characterized in that firstly, trade documents related to verification targets are screened out, and document elements in the trade documents are extracted, so that the abnormality of a final verification conclusion caused by error judgment on the document elements during manual verification is prevented. Then, constructing a question-answer task containing document elements according to the verification target, and then obtaining a system text related to the verification target through a verification system corresponding to the verification target so as to prevent the fact that the verification process is difficult to update in real time due to the changeable verification system, thereby ensuring that the verification system accords with the current verification system and ensuring the accuracy and the high efficiency of a verification conclusion obtained subsequently. Finally, inputting the system text and the question-answering task into a first question-answering model based on prompt learning to obtain a verification conclusion of the verification target, and directly obtaining the related system text through the verification system related to the verification target, thereby preventing the difficulty in efficiently and accurately setting a verification process manually caused by the complicated verification system; the verification method, the verification device, the computer equipment and the storage medium of the trade background enable the verification process to conform to the verification system, and further the verification conclusion is obtained efficiently and accurately.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments of the present invention will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic illustration of an application environment of a verification method for trade background in an embodiment of the invention;
FIG. 2 is a flow chart of a verification method for trade background in an embodiment of the invention;
FIG. 3 is a schematic diagram of a verification device for trade background in an embodiment of the invention;
FIG. 4 is a schematic diagram of a computer device in accordance with an embodiment of the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In supply chain finance, trade background verification is carried out because complex supply-demand relations and business requirements exist, and formats of different verification documents are inconsistent, so that in a program of trade background verification automation, coupling of the business processes is serious, logic splitting of the program cannot be carried out, and a large amount of manual intervention is needed, so that verification is completed. In addition, due to the change of policies and environments, the system files on which verification is based often change, so that verification errors are often caused due to untimely updating of the system or regulations in the verification process, and verification efficiency and accuracy are affected.
The automatic trade background verification system on the market at present mainly relates to management, verification management and conclusion management of trade files, but does not solve the verification detail problem, and does not realize the index of verification system files. Although some of the systems also generate partial verification rules according to the regulations or regulations, the data volume of the regulations or regulations is too large, so that the verification rules are difficult to automatically generate, and even if the verification rules are generated, the verification rules often need to be manually checked one by one to prevent errors of verification procedures.
Based on the above problems, the verification method of trade context provided by the embodiment of the invention, that is, based on the semantic understanding capability of the large model, completes the verification of trade context, including but not limited to automatic verification, verification system management and flow visualization management. The verification method is applicable in an application environment as in fig. 1, where the terminal device communicates with the server via a network. The verification target can be obtained through the terminal equipment, a large number of trade documents and trade systems are obtained, so that the trade documents and the verification systems are associated through the verification target at the server, and an accurate verification conclusion is obtained. The terminal device may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices. The server may be implemented as a stand-alone server or as a server cluster composed of a plurality of servers.
In one embodiment, as shown in fig. 2, a verification method for trade background is provided, and the method is applied to the server in fig. 1 for illustration, and includes the following steps:
s10: determining verification targets of trade backgrounds, and screening out trade documents related to the verification targets;
where trade documents refer to all documents used in connection with trade context verification, including but not limited to contracts, invoices, and the like. Verification targets include, but are not limited to, confirming background qualification of a company, verifying a ticket, validating, etc.
S20: and extracting document elements in the trade document.
Where document elements refer to information elements obtained by mining and/or extracting trade documents, including but not limited to text entities, text relationships, and the like. Verification targets may be entered by the user in text format or by the user uploading relevant verification target documents via a page, not limited herein.
For example, in the verification target of contract verification, the trade document is a contract file uploaded by the user, and at this time, the entities of both the contract are required to be extracted from the contract file as document elements, and then the document elements are finally obtained as follows: party a, party B.
Notably, step S20 essentially automatically extracts document elements in the trade document. Because the trade background has responsible supply and demand relations and the number of verification files in different formats is numerous, if the document elements of the trade documents are screened manually, on one hand, the document elements which are finally obtained are caused by different people understanding the business, so that the process is automated, and the abnormality of the follow-up verification conclusion caused by the problems can be prevented; on the other hand, compared with manual operation, the automatic extraction effectively improves the efficiency of extracting the document elements.
S30: and constructing a question-answer task containing the document elements according to the verification targets.
The question-answering task refers to a question-answering task containing document elements, and is used for judging whether the document elements meet a verification target.
For example, in the verification target of whether the contract is qualified, the trade document is a contract document uploaded by the user, and the document elements are: party a, party B. At this time, if the question-answer task is that the contract first party element is a, the second party element is B, and the contract is qualified?
It is noted that step S30 is mainly used for the first question-answer model of the subsequent step S50, so that the verification conclusion can be accurately reached by the question-answer task based on the first question-answer model. Thus, the accuracy of the question-and-answer task at step S30 will directly affect the accuracy of the final verification conclusion. The step utilizes the document elements and verification targets to construct the question-answering task, and can ensure the accuracy of the question-answering task to the greatest extent.
S40: and acquiring a system text related to the verification target through a verification system corresponding to the verification target.
The verification system includes, but is not limited to, policy files, system files and the like related to the verification target. The system text is the text with the highest correlation degree with the verification target in the content of the verification system.
For example, if the target is that the names of the parties a and b in the contract are in compliance and the system is that the contract law, step S40 obtains the terms related to the names of the parties a and b from the contract law as the system text.
It is worth noting that in the prior art, the system texts are often screened out from the verification system by manpower, but the finally obtained system texts are often different due to different understanding of different people on the verification system, so that the accuracy of the obtained system texts is effectively improved by automatically obtaining the system texts through the verification target, and the accuracy of the verification conclusion obtained later is ensured. In addition, since the verification system is updated frequently due to the change of the policy or the environment, in this step, the verification system is acquired according to the verification target in real time every time the core, so that the use of the expired verification system is effectively prevented, and the abnormality of the subsequent verification result is caused.
S50: and inputting the system text and the question-answering task into a first question-answering model based on prompt learning to obtain a verification conclusion of the verification target.
Wherein Prompt Learning (Prompt Learning) is capable of changing downstream tasks to text generation tasks by adding Prompt words to the input without significantly changing the pre-trained language model structure and parameters. And generating a task through the text to obtain a task target. In step S50, the text generation task is generated by the system text and the question-answer task, and the task target is the verification target.
For example, if the target is that the names of the first party and the second party in the contract are in compliance, the question-answering task is that the name A of the first party and the name B of the second party are in compliance, and the system text is the part of the terms of the contract law, the content is input into a first question-answering model, wherein the text generating task is as follows: in the provision of the contract law section clause, the first party name a and the second party name B are compliant. The final verification conclusion is compliance or non-compliance.
In this embodiment, first, document elements in a trade document are extracted from the trade document corresponding to a verification target, so as to prevent abnormality of a final verification conclusion caused by error judgment on the document elements during manual verification. Then, constructing a question-answer task containing document elements according to the verification target, and then obtaining a system text related to the verification target through a verification system corresponding to the verification target so as to prevent the fact that the verification process is difficult to update in real time due to the changeable verification system, thereby ensuring that the verification system accords with the current verification system and ensuring the accuracy and the high efficiency of a verification conclusion obtained subsequently. And finally, inputting the system text and the question-answering task into a first question-answering model based on prompt learning to obtain a verification conclusion of the verification target, and directly passing through the system text related to the verification target, thereby preventing the complex verification system from causing the difficulty that a verification process cannot be set manually, efficiently and accurately, enabling the verification process to be more in accordance with the verification system, and further obtaining the verification conclusion efficiently and accurately.
In one embodiment, in step S20, extracting document elements in the trade document specifically includes the following steps:
s21: and analyzing and verifying the trade document corresponding to the target, and selecting out the core fragment of the trade document.
S22: and constructing prompt words of the element extraction class.
S23: and inputting the core segment and the prompt word into a second question-answer model based on prompt learning to obtain document elements in the trade document.
The prompt word refers to prompt and is used for guiding generation of a second question-answer model based on prompt learning. The core segment refers to the content segment with the highest correlation degree with the verification target in the content of the trade document. Since the hint terms in this embodiment are extracted element classes, the final step S23 will be able to extract document elements in the trade document according to the second question-answer model.
For example, in the verification target of contract analysis, firstly, a contract file is read, a prompt of element extraction class is constructed, and then, through the processes of loading trade documents, document segmentation, similarity calculation and the like, the most relevant contract segment (i.e. core segment) is finally selected and sent to a second question-answer model together with the prompt, so that extraction of the first party/second party elements is realized.
It should be noted that, in this embodiment, the document elements in the trade document are accurately extracted substantially through prompt learning, and the extraction of the core segment in step S21 can make the finally obtained document elements more targeted and accurate, so that the verification conclusion finally obtained through the document elements is more accurate.
In one embodiment, after step S50, the system text and the question-answer task are input into a first question-answer model based on prompt learning, and after the verification conclusion of the verification object is obtained, the verification method further includes the following steps:
s91: and according to the system text, obtaining the verification rule of the verification target.
S92: a verification flow is generated that conforms to the verification rules.
S93: when any verification node in the verification process is modified, the subsequent verification process of the verification node is regenerated according to the verification rule.
S94: and updating the verification conclusion according to the modified verification node and the updated verification process.
The embodiment provides an intervention method for manually finding abnormal verification flow in an automatic verification process. Where verification flow refers to a flow that must be completed in a system or code associated with verification that includes one or more verification nodes.
For example, a certain verification target must actually pass through the verification of four process nodes "ABCD" in sequence, but the verification process obtained by the current verification method passes through four process nodes "AECF" in sequence, and after the user finds the abnormality, the node "E" is modified, and the process is changed to "ABCF". When the user completes the change, the subsequent verification process "CF" is automatically updated to "CD".
It should be noted that, this embodiment is essentially to prevent all verification processes from being automatically performed by the first question-answer model, so that the verification processes cannot be manually supervised. Thus, steps S91-S92 essentially show the user a visual representation of the verification process completed by the first question-and-answer model, and step S93 is when the user changes the verification process, the first question-and-answer model updates the subsequent verification process to thereby arrive at a new verification conclusion. The embodiment effectively ensures the manual supervision and manual intervention in the whole verification process, thereby preventing the situation that the manual work cannot be processed in time when unpredictable abnormality occurs in the full-automatic verification process. Due to the existence of such manual supervision, the accuracy of the whole verification process is more effectively ensured. In addition, since the subsequent verification process is automatically updated after the user modifies any node of the verification process, the accuracy and the generation efficiency of the subsequent verification process are effectively improved.
In one embodiment, in step S30, a question-answer task including the document element is constructed according to the verification target, which specifically includes the following steps:
s31: disassembling the verification object into a plurality of sub-objects;
s32: screening document elements corresponding to each sub-target to serve as document sub-elements;
s33: respectively constructing a question-answer subtask corresponding to each sub-target according to each document sub-element;
s34: and summarizing all the question-answer subtasks into the question-answer tasks.
For example, if the current verification target is to verify the trade background of the company a, all the contract files of the company a and other companies are obtained, and a sub-target corresponding to each contract file is generated, wherein the sub-target is whether the current contract is compliant. At this time, a question-answer subtask corresponding to each sub-target can be constructed, and the question-answer subtask of the contract A is: "determine if contract A is compliant" the question-answer subtask of contract B is: "determine if contract B is compliant". Summarizing the two subtasks to obtain a question-answer task: "judging whether contract a is compliant? If so, judging whether the contract B is compliant.
It should be noted that, in this embodiment, the verification target is substantially disassembled, a large verification target is disassembled into a plurality of small sub-targets, and then a plurality of question-answer sub-tasks are generated according to the sub-targets, so that the accuracy and pertinence of the verification process are ensured, and the accuracy of the verification conclusion finally obtained according to the question-answer tasks is further effectively ensured.
In one embodiment, in step S30, a question-answer task including the document element is constructed according to the verification target, which specifically includes the following steps:
s35: and constructing a judging question sentence whether the verification target is reached or not by using the document element.
S36: and taking the judging question sentence as the question answering task.
Step S35 is a process of generating a question for judgment including document elements, the question indicating whether the answer mode is yes or no. Steps S35-S36 may be used in the process of generating a question-answer subtask from a document sub-element, in addition to generating a question-answer task from a document element.
It should be noted that, in this embodiment, the accuracy of the verification conclusion obtained after the question-answering task is input into the first question-answering model in the subsequent steps is essentially guaranteed. Because the judging question sentence is more convenient for a machine to understand the problem, the complexity of the problem is prevented from influencing the accuracy of the verification conclusion. Thus, the present embodiment further ensures the accuracy of the verification conclusion.
In an embodiment, after step S33, that is, after each question-answer subtask corresponding to each sub-object is respectively constructed according to each document sub-element, different service interfaces corresponding to the question-answer subtask can be called by means of semantic understanding capability of a large model, so as to obtain sub-prompt words for different subtasks, for example, task 1 is contract analysis, at this time, a contract file uploaded by a user is input, and sub-prompt words corresponding to analysis services are obtained through calling interfaces.
In one embodiment, before step S40, that is, before the system text related to the verification target is obtained through the verification system corresponding to the verification target, the verification method further includes the following steps:
s81: and classifying all the verification system documents to obtain classified documents belonging to different business categories.
S82: and according to the service type, screening out similar documents from the classified documents through similarity calculation.
S83: and recombining the prompt words of different targets according to the prompt word template, the similar documents and the different verification targets.
S84: based on the business type, training the first question-answering model based on prompt learning by using a large language model and prompt words of different targets.
Wherein the Prompt word (Prompt) is used for guiding and generating a subsequent first question-answer model. The prompting word template refers to a template for reserving the position of the prompting word. A large language model (english: large Language Model, abbreviated LLM) is used to understand and generate human language, and the use of the large language model in step S84 enables the final first question-answering model to understand the human language-based question-answering task and to draw a human language-based verification conclusion.
Specifically, all verification system documents under different service types are vectorized and stored in a vector database. And carrying out text vectorization on the verification target, and further carrying out similarity calculation on vectors of all documents under the service type corresponding to the verification target, so as to obtain similar documents. Since these similar documents carry index information, the user can efficiently inquire about these documents in accordance with the index information. And then recombining the similar documents and verification targets according to the prompt word templates to recombine different prompt words of different verification targets.
It should be noted that this embodiment essentially constructs a knowledge base of verification documents. The knowledge base is classified by service category, and all similar texts are associated by similarity calculation. The knowledge base effectively ensures the management of the verification system with complex changes, thereby preventing the untimely updating of the verification system document caused by the policy or environment changes from affecting the subsequent verification conclusion, i.e. better ensuring the accuracy of the verification conclusion.
The verification system documents are classified according to the service types, the similarity calculation is carried out on the verification system documents, and all similar texts are obtained, so that the accuracy of the verification conclusion is effectively improved, the retrieval efficiency of the verification system documents is improved, and the high efficiency of the whole verification method is ensured. In addition, in the embodiment, the similar documents are utilized to construct the prompt words, so that all relevant documents are considered under the current service type, and the accuracy and precision of the prompt words are ensured.
In one embodiment, in step S83, the prompt words of different targets are recombined according to the prompt word template, the similar document and the different verification targets, and the method specifically includes the following steps:
s831: and acquiring different versions of the similar document, and acquiring version documents in the different versions.
S832: and comparing the version documents in the different versions to obtain the difference content of the different versions.
S833: and recombining prompt words of different targets according to the prompt word template, the difference content and the different verification targets.
Wherein different versions refer to different updated versions of the same system document, for example, the A regulation issues an updated version on 1 month 1 day and an updated version on 2 months 1 day, then 1 month 1 day and 2 months 1 day are two different versions of the A regulation.
In step S833, the prompt word template is a method for constructing the prompt word in prompt learning, and in this embodiment, the prompt word is recombined in a targeted manner mainly according to the differences between different versions of the current verification system. For example, the A-rule in the previous version is "verify conclusions by the A element", while the next version "verify conclusions by both A and B elements together". Therefore, the prompt word of the previous version only needs to guide the first question-answer model through the A to obtain a result, and the next version needs to consider the conclusion deduced by the elements of the A and the B, so that attention needs to be paid to the difference of the related verification contents under the influence of the common factors of the A and the B, namely the difference in the system or the rule, and the prompt word needs to guide the model effectively and accurately according to the difference in the system or the rule so that the model obtains an accurate conclusion.
It should be noted that, based on the relevant system documents in the trade background, unpredictable update iterations, i.e. version iterations, are often generated due to system or environment changes, thus providing a verification method in this embodiment. By understanding and comparing verification systems of different versions, the accuracy of the finally recombined prompt words is ensured through difference contents. The step simulates the thinking process that the artificial nucleus compares the verification systems of different versions in real time, thereby ensuring that the finally recombined prompt words more accord with the verification target, and the subsequently obtained system text also more accords with the current verification scene, and further ensuring the accuracy of the verification method.
In an embodiment, after step S833, that is, after the alert words of different objects are recombined according to the alert word template, the difference content and the different verification objects, the verification method further includes the following steps:
s61: and respectively obtaining verification conclusion corresponding to the documents of different versions according to each verification target.
S62: and judging whether the prompt words of the different targets accord with the difference content or not through the verification conclusion corresponding to the documents of different versions.
S63: and if the difference content is not matched with the difference content, comparing the version documents in the different versions again so as to reorganize the prompt words through the reacquired difference content.
It should be noted that, because of the update iteration of the system document, it is possible that the content of the previous version of the document passes verification, and the content of the new version of the document cannot pass verification. Therefore, the verification conclusions of different versions need to be compared, and differences are analyzed, so that whether the different verification conclusions accord with the difference content of the document itself is judged, whether an automatic verification process is accurate is monitored, and the accuracy and precision of the verification method are improved.
In an embodiment, after step S61, after verification conclusions corresponding to different version documents are obtained, version document links corresponding to the verification conclusions are generated, so that the corresponding version documents are accessed through the version document links, and supervision comparison is performed on different very-fit conclusions manually, so that accuracy of finally obtaining prompt words is ensured.
In summary, the core design of the present invention is to disassemble and reassemble the task of verification targets, which includes numerous verification sub-targets, and besides the large model based on semantic understanding (prompt learning model), a plurality of technologies such as OCR (optical character recognition, which is called Optical character recognition) recognition, form reorganization and the like can be combined to complete different verification targets.
Since the whole verification process is completed through the system text automation, the verification process is not required to be manually configured or designed, and the verification process is visualized, so that the verification node cannot be manually corrected due to the excessively high degree of automation. In addition, due to the application of the prompt learning model, the retrieval and understanding capability of the verification method on the verification system files is ensured, so that the information retrieval efficiency is improved, the difference of verification conclusions caused by different people understanding the same verification system files is avoided, and the verification risk is effectively prevented.
It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic, and should not limit the implementation process of the embodiment of the present invention.
In one embodiment, a verification device for trade background is provided, and the verification device for trade background corresponds to the verification method for trade background in the above embodiment one by one. As shown in fig. 3, the verification device for trade background includes a screening module 10, an extraction module 20, a construction module 30, an acquisition module 40, and an output module 50. The functional modules are described in detail as follows:
a screening module 10, configured to determine a verification target of a trade background, and screen out a trade document related to the verification target;
an extraction module 20, configured to extract document elements in the trade document;
a construction module 30 for constructing a question-answer task including the document element according to the verification target;
an obtaining module 40, configured to obtain a system text related to the verification target through a verification system corresponding to the verification target;
and the output module 50 is used for inputting the system text and the question-answer task into a first question-answer model based on prompt learning to obtain a verification conclusion of the verification target.
The specific definition of the verification device for trade background can be found in the definition of the verification method for trade background hereinabove, and will not be described in detail herein. The various modules in the verification device of the trade background described above may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 4. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used to store all data generated in the verification method implementing the trade background described above. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by the processor to implement a verification method of trade context.
In one embodiment, a computer device is provided comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of when executing the computer program:
determining verification targets of trade backgrounds, and screening out trade documents related to the verification targets;
extracting document elements in the trade document;
constructing a question-answer task containing the document elements according to the verification target;
acquiring a system text related to the verification target through a verification system corresponding to the verification target;
and inputting the system text and the question-answering task into a first question-answering model based on prompt learning to obtain a verification conclusion of the verification target.
In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of:
determining verification targets of trade backgrounds, and screening out trade documents related to the verification targets;
extracting document elements in the trade document;
constructing a question-answer task containing the document elements according to the verification target;
acquiring a system text related to the verification target through a verification system corresponding to the verification target;
and inputting the system text and the question-answering task into a first question-answering model based on prompt learning to obtain a verification conclusion of the verification target.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions.
The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and are intended to be included in the scope of the present invention.
Claims (10)
1. A method of verifying trade context, comprising:
determining verification targets of trade backgrounds, and screening out trade documents related to the verification targets;
extracting document elements in the trade document;
constructing a question-answer task containing the document elements according to the verification target;
acquiring a system text related to the verification target through a verification system corresponding to the verification target;
and inputting the system text and the question-answering task into a first question-answering model based on prompt learning to obtain a verification conclusion of the verification target.
2. The method for verifying a trade background of claim 1, wherein said extracting document elements in said trade document comprises:
analyzing the trade document, and selecting a core segment of the trade document;
constructing a prompting word of the element extraction class;
and inputting the core segment and the prompt word into a second question-answer model based on prompt learning to obtain document elements in the trade document.
3. The method for verifying a trade background according to claim 1, wherein said inputting said system text and said question-answering task into a first question-answering model based on prompt learning, after deriving a verification conclusion of said verification target, further comprises:
obtaining a verification rule of the verification target according to the system text;
generating a verification flow conforming to the verification rule;
when any verification node in the verification process is modified, regenerating the subsequent verification process of the verification node according to the verification rule;
and updating the verification conclusion according to the modified verification node and the updated verification process.
4. The method for verifying the trade background of claim 1, wherein said constructing a question-answer task containing said document element according to said verification target comprises:
disassembling the verification object into a plurality of sub-objects;
screening document elements corresponding to each sub-target to serve as document sub-elements;
respectively constructing a question-answer subtask corresponding to each sub-target according to each document sub-element;
and summarizing all the question-answer subtasks into the question-answer tasks.
5. The method for verifying a trade background according to claim 1, wherein said method further comprises, before said acquiring a system text related to said verification target by a verification system corresponding to said verification target:
classifying all verification system documents to obtain classified documents belonging to different business categories;
according to the service type, screening out similar documents from the classified documents through similarity calculation;
recombining prompt words of different targets according to the prompt word template, the similar documents and the different verification targets;
based on the business type, training the first question-answering model based on prompt learning by using a large language model and prompt words of different targets.
6. The method for verifying a trade background according to claim 5, wherein said reorganizing the prompter of different targets according to the prompter template, the similar document, and the different verification targets comprises:
acquiring different versions of the similar document, and acquiring version documents in the different versions;
comparing the version documents in the different versions to obtain difference contents of the different versions;
and recombining prompt words of different targets according to the prompt word template, the difference content and the different verification targets.
7. The method for verification of trade background according to claim 6, wherein said method further comprises, after said reassembling the prompter of different targets according to the prompter template, said difference content and different said verification targets:
respectively obtaining verification conclusions corresponding to different versions of documents according to each verification target;
judging whether the prompt words of different targets accord with the difference content or not through verification conclusions corresponding to different versions of documents;
and if the difference content is not matched with the difference content, comparing the version documents in the different versions again so as to reorganize the prompt words through the reacquired difference content.
8. A verification apparatus for a trade background, comprising:
the screening module is used for determining verification targets of trade backgrounds and screening out trade documents related to the verification targets;
the extraction module is used for extracting document elements in the trade document;
the construction module is used for constructing a question-answer task containing the document elements according to the verification target;
the acquisition module is used for acquiring a system text related to the verification target through a verification system corresponding to the verification target;
and the output module is used for inputting the system text and the question-answer task into a first question-answer model based on prompt learning to obtain a verification conclusion of the verification target.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the verification method of trade context according to any one of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the verification method of trade background according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311843787.3A CN117788163A (en) | 2023-12-28 | 2023-12-28 | Verification method and device for trade background, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311843787.3A CN117788163A (en) | 2023-12-28 | 2023-12-28 | Verification method and device for trade background, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117788163A true CN117788163A (en) | 2024-03-29 |
Family
ID=90383388
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311843787.3A Pending CN117788163A (en) | 2023-12-28 | 2023-12-28 | Verification method and device for trade background, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117788163A (en) |
-
2023
- 2023-12-28 CN CN202311843787.3A patent/CN117788163A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111061833B (en) | Data processing method and device, electronic equipment and computer readable storage medium | |
| CN108256074A (en) | Method, apparatus, electronic equipment and the storage medium of checking treatment | |
| CA2907208C (en) | System and method for developing business rules for decision engines | |
| US11367008B2 (en) | Artificial intelligence techniques for improving efficiency | |
| US11907299B2 (en) | System and method for implementing a securities analyzer | |
| EP3640814A1 (en) | User-friendly explanation production using generative adversarial networks | |
| Gupta et al. | Reducing user input requests to improve IT support ticket resolution process | |
| CN111177307A (en) | Test scheme and system based on semantic understanding similarity threshold configuration | |
| US11055200B2 (en) | Systems and methods for validating domain specific models | |
| US20220292268A1 (en) | Smart contract generation system and methods | |
| Kulkarni et al. | Toward automated regulatory compliance | |
| CN116484025A (en) | Vulnerability knowledge graph construction method, vulnerability knowledge graph evaluation equipment and storage medium | |
| CN117788163A (en) | Verification method and device for trade background, computer equipment and storage medium | |
| US20220215142A1 (en) | Extensible Agents in Agent-Based Generative Models | |
| US11847390B2 (en) | Generation of synthetic data using agent-based simulations | |
| US20220215262A1 (en) | Augmenting Datasets with Synthetic Data | |
| CN117940890A (en) | Collaborative industrial integrated development and execution environment | |
| EP4275343A1 (en) | Generation and evaluation of secure synthetic data | |
| US20220309335A1 (en) | Automated generation and integration of an optimized regular expression | |
| Antonelli et al. | Early identification of crosscutting concerns with the Language Extended Lexicon | |
| CN111240846A (en) | Data auditing method and device, computer equipment and storage medium | |
| Arcega et al. | Feature location through the combination of run-time architecture models and information retrieval | |
| Kumar et al. | Natural Language Generation and Artificial Intelligence in Financial Reporting: Transforming Financial Data into Strategic Insights for Executive Leadership | |
| CN115249017B (en) | Text labeling method, training method of intention recognition model and related equipment | |
| CN115543432B (en) | Universal type Git code hosting platform system and implementation method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |