An intelligent reconciliation method and system for a procurement principal

Abstract

The application discloses an intelligent account checking method and system for a procurement subject, comprising the following steps: firstly, obtaining an archiving library of a target procurement subject and an initial account checking request; then, loading the request and corresponding first and second account checking prompt words into corresponding models respectively to obtain a first information item missing in the initial request and a second information item with ambiguity; determining a plurality of pending confirmation information according to the second information item and the archiving library; and according to the supplementary content of the first information item and a confirmation instruction for the pending confirmation information, optimizing the initial request to obtain a target account checking request, and performing a query in the archiving library according to the target account checking request, so that intelligent and accurate account checking operation is realized, and the account checking efficiency and quality are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to an intelligent reconciliation method and system for purchasing entities. Background Technology

[0002] In procurement operations, reconciliation is crucial. Traditional reconciliation methods targeting specific procurement entities rely on manual verification of large volumes of procurement data, which is inefficient and prone to errors. Furthermore, initial reconciliation requests often contain missing or ambiguous information, leading to inaccuracies or delays. Faced with massive amounts of procurement contracts and transaction records, existing simple query methods are insufficient to meet complex and ever-changing reconciliation needs, failing to quickly and accurately obtain the required reconciliation information. Therefore, there is an urgent need for an intelligent reconciliation method to improve efficiency and accuracy. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent reconciliation method and system for purchasing entities.

[0004] In a first aspect, embodiments of the present invention provide an intelligent reconciliation method for a purchasing entity, comprising:

[0005] Obtain the target procurement entity archive database and an initial reconciliation request for querying the target procurement entity archive database;

[0006] The initial reconciliation request and the first reconciliation prompt word corresponding to the initial reconciliation request are loaded into the first information item requirement parsing model to obtain the first information item, wherein the first information item is a missing mandatory information item in the initial reconciliation request;

[0007] The initial reconciliation request and the second reconciliation prompt corresponding to the initial reconciliation request are loaded into the second information item requirement parsing model to obtain the second information item, wherein the second information item is the ambiguous information item in the initial reconciliation request;

[0008] Based on the second information item and the target procurement entity archive, determine multiple pending confirmation information items corresponding to the second information item;

[0009] Based on the required supplementary content for the first information item and the confirmation instructions for the multiple pending confirmation information, the initial reconciliation request is optimized to obtain the target reconciliation request, and a query is performed in the target procurement entity archive based on the target reconciliation request.

[0010] In one possible implementation, before loading the initial reconciliation request and the first reconciliation prompt word corresponding to the initial reconciliation request into the first information item demand parsing model to obtain the first information item, the method further includes:

[0011] Based on the initial reconciliation request, query the target procurement entity's archive database to obtain the initial reconciliation results;

[0012] The initial reconciliation request, the initial reconciliation result, and the third reconciliation prompt corresponding to the initial reconciliation request and the initial reconciliation result are loaded into the reconciliation complexity assessment model to obtain the inferred reconciliation complexity assessment result.

[0013] If the reconciliation complexity assessment result indicates that the initial reconciliation request does not require complex reconciliation processing, the execution of the query based on the initial reconciliation request will be terminated.

[0014] In one possible implementation, the step of querying the target procurement entity's archive database to obtain the initial reconciliation result based on the initial reconciliation request includes:

[0015] Obtain the rules for extracting contract elements;

[0016] According to the contract element extraction rules, extract contract elements from the initial reconciliation request;

[0017] The coverage of the contract elements relative to the contract elements is determined based on the number of times the contract elements are matched in each related transaction instance in the target procurement entity's archive.

[0018] Based on the number of related transaction instances containing the contract element in the target procurement entity's archive, the strength of the related transaction instance's clause constraint relative to the contract element is determined.

[0019] Based on the coverage and binding strength of the terms, a set of write-off candidates is determined from each related transaction instance in the target procurement entity's archive, which serves as the initial reconciliation result.

[0020] In one possible implementation, the third reconciliation prompt is obtained by the following process, including:

[0021] Obtain the first business rule preset conditions used for determining complex reconciliation processing;

[0022] Based on the preset conditions of the first business rule, a first logic determination component is constructed. The first logic determination component is configured to determine whether the initial reconciliation request requires complex reconciliation processing based on the preset conditions of the first business rule.

[0023] Construct an execution strategy component and an archiving strategy component. The execution strategy component configures the handling when the initial reconciliation request requires complex reconciliation processing, and the archiving strategy component configures the handling when the initial reconciliation request does not require complex reconciliation processing.

[0024] The third reconciliation prompt word is constructed based on the first logic determination component, the execution strategy component, and the archiving strategy component;

[0025] The first reconciliation prompt is obtained through the following process:

[0026] Obtain the first contract terms and first clause performance examples for each first information item;

[0027] Based on the first contract terms and the first performance example, a second logic determination component is constructed. The second logic determination component is configured to determine whether the initial reconciliation request contains the first information item based on the first contract terms and the first performance example.

[0028] Construct a clause matching prompt component and a clause missing prompt component. The clause matching prompt component is configured to provide a contract clause association prompt when the initial reconciliation request includes the first information item. The clause missing prompt component is configured to provide a contract element missing warning when the initial reconciliation request does not include the first information item.

[0029] Based on the second logic determination component, the clause matching prompt component, and the clause missing prompt component, the first reconciliation prompt word is constructed;

[0030] The construction of the clause matching prompt component and the clause missing prompt component includes:

[0031] The contract terms parsing process, terms matching status code, and terms performance description are obtained. The contract terms parsing process is the parsing process of the initial reconciliation request containing the first information item. The terms matching status code indicates that the initial reconciliation request contains the first information item. The terms performance description is the text information of the initial reconciliation request containing the first information item.

[0032] Based on the contract terms parsing process, the terms matching status codes, and the terms performance descriptions, construct the terms matching prompt component.

[0033] Obtain the contract reconciliation anomaly analysis process, element missing identifier, and element missing reason explanation. The contract reconciliation anomaly analysis process is the analysis process in which the initial reconciliation request does not contain the first information item. The element missing identifier indicates that the initial reconciliation request does not contain the first information item, and the element missing reason explanation indicates that the first information item is missing in the initial reconciliation request.

[0034] Based on the contract reconciliation anomaly analysis process, the missing element identifier, and the explanation of the missing element reasons, construct the missing clause prompt component;

[0035] The second reconciliation prompt key is obtained through the following process:

[0036] Obtain the second contract terms specifications and second contract performance examples for each second information item;

[0037] Based on the second contract terms specification and the second terms performance example, a third logic determination component is constructed. The third logic determination component is configured to extract the second information item from the initial reconciliation request based on the second contract terms specification and the second terms performance example.

[0038] Based on the priority rules for the reconciliation elements extracted from the second information item, a fourth logical judgment component is constructed; the priority rules for the reconciliation elements include the principle of mandatory parsing of amount elements, the principle of conflict handling of timeliness elements, and the principle of verification of signature validity.

[0039] The second reconciliation prompt is constructed based on the third and fourth logic determination components.

[0040] In one possible implementation, the target procurement entity archive contains a procurement contract terms knowledge graph, which contains multiple contract terms nodes.

[0041] The step of determining multiple pending confirmation information corresponding to the second information item based on the second information item and the target procurement entity archive includes:

[0042] For each of the second information items, query the first associated clause instance corresponding to the second information item in the plurality of contract clause nodes;

[0043] If there are multiple instances of the first associated clause retrieved, add all instances of the first associated clause to the procurement term feature library;

[0044] Referring to the aforementioned procurement terminology feature library, the second information item is subjected to contract element structured parsing to obtain performance element atomic items;

[0045] Determine the multiple pending confirmation information corresponding to the atomic items of the performance element.

[0046] In one possible implementation, the procurement contract terms knowledge graph also includes multiple clause element constraints corresponding to the contract terms nodes;

[0047] The determination of the multiple pending confirmation information corresponding to the atomic item of the performance element includes:

[0048] Based on the atomic items of the performance elements, query the knowledge graph of the procurement contract terms;

[0049] If a second associated clause instance corresponding to the atomic item of the performance element is obtained by querying the knowledge graph of the procurement contract terms, and the second associated clause instance corresponds to multiple second clause element constraints in the knowledge graph of the procurement contract terms, then the multiple pending confirmation information is constructed based on the multiple second clause element constraints.

[0050] In one possible implementation, if querying the procurement contract terms knowledge graph yields a second associated clause instance corresponding to the atomic item of the performance element, and in the procurement contract terms knowledge graph, the second associated clause instance corresponds to multiple second clause element constraints, then constructing the multiple pending confirmation information based on the multiple second clause element constraints includes:

[0051] It is determined that querying the procurement contract terms knowledge graph yields a second associated clause instance corresponding to the atomic item of the performance element, and in the procurement contract terms knowledge graph, the second associated clause instance corresponds to multiple second clause element constraints;

[0052] The performance element atomic item and the fifth reconciliation prompt word corresponding to the performance element atomic item are loaded into the clause element constraint inference model to obtain at least one supplementary clause element constraint condition corresponding to the performance element atomic item.

[0053] Based on the constraints of the multiple second clause elements and the constraints of the at least one supplementary clause element, the multiple pending confirmation information is constructed;

[0054] After querying the knowledge graph of the procurement contract terms based on the atomic items of the performance elements, the method further includes:

[0055] If the knowledge graph of the procurement contract terms does not yield a second associated clause instance corresponding to the atomic item of the performance element, the atomic item of the performance element and the fourth reconciliation prompt word corresponding to the atomic item of the performance element are loaded into the contract element compliance verification model.

[0056] If the contract element compliance verification model identifies that the performance element atomic item is not a standard contract atomic item, a contract element update instruction is constructed, which is used to indicate the update of the performance element atomic item;

[0057] The fourth reconciliation prompt key is obtained through the following process:

[0058] Based on the contract terms specification of the standard contract atomic item, a fifth logic determination component is constructed. The fifth logic determination component is configured to determine whether the performance element atomic item is the standard contract atomic item based on the contract terms specification of the standard contract atomic item.

[0059] Based on the compliance verification rules of the standard contract atomic items, a compliance verification rule determination component is constructed;

[0060] The fourth reconciliation prompt is constructed based on the fifth logic determination component and the compliance verification rule determination component.

[0061] In one possible implementation, optimizing the initial reconciliation request based on the required supplementary content for the first information item and the confirmation instructions for the plurality of pending confirmation information to obtain the target reconciliation request includes:

[0062] Receive the required supplementary content for the first information item loaded by the current administrator and the confirmation instruction to select the target confirmation information from the plurality of pending confirmation information;

[0063] Based on the required supplementary information, supplement the first information item in the initial reconciliation request;

[0064] Based on the target confirmation information, the second information item in the initial reconciliation request is optimized, and the initial reconciliation request after supplementing the first information item and optimizing the second information item is taken as the target reconciliation request.

[0065] In one possible implementation, optimizing the initial reconciliation request based on the required supplementary content for the first information item and the confirmation instructions for the plurality of pending confirmation information to obtain the target reconciliation request includes:

[0066] Obtain the first contract classification identifier and the first transaction scenario characteristics of the initial reconciliation request;

[0067] Obtain the set of past reconciliation requests of the current administrator, where the current administrator is the administrator who initiated the initial reconciliation request;

[0068] For each past reconciliation request in the set of past reconciliation requests, obtain the second contract classification identifier and the second transaction scenario feature of the past reconciliation request;

[0069] If the second contract classification identifier of a past reconciliation request in the set of past reconciliation requests matches the first contract classification identifier, and the second transaction scenario feature of the past reconciliation request matches the first transaction scenario feature, then the first past information item and the second past information item in the past reconciliation request are obtained.

[0070] Based on the first past information item, the required supplementary content is constructed, and based on the second past information item, the confirmation instruction is constructed;

[0071] Based on the required supplementary information and the confirmation instruction, the initial reconciliation request is optimized to obtain the target reconciliation request.

[0072] In a second aspect, embodiments of the present invention provide a server system, including a server, the server being used to execute the method described in the first aspect.

[0073] Compared to existing technologies, the beneficial effects of this invention include: Employing the intelligent reconciliation method and system disclosed herein for procurement entities, the method obtains the target procurement entity's archive database and initial reconciliation request. Then, the request and corresponding first and second reconciliation prompts are loaded into the respective models to obtain the missing mandatory first information item and ambiguous second information item from the initial request. Based on the second information item and the archive database, multiple pending confirmation information items are determined. Then, based on the supplementary content of the first information item and the confirmation instructions for the pending confirmation information, the initial request is optimized to obtain the target reconciliation request. A query is then performed in the archive database accordingly, achieving intelligent and accurate reconciliation operations and improving reconciliation efficiency and quality. Attached Figure Description

[0074] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0075] Figure 1 A flowchart illustrating the steps of the intelligent reconciliation method for a purchasing entity provided in an embodiment of the present invention;

[0076] Figure 2 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0078] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0079] In order to solve the technical problems mentioned in the background art Figure 1This is a flowchart illustrating the intelligent reconciliation method for a purchasing entity provided in this embodiment of the disclosure. The intelligent reconciliation method for a purchasing entity will be described in detail below.

[0080] Step S201: Obtain the target procurement entity archive database and an initial reconciliation request for querying the target procurement entity archive database;

[0081] Step S202: Load the initial reconciliation request and the first reconciliation prompt word corresponding to the initial reconciliation request into the first information item requirement parsing model to obtain the first information item, wherein the first information item is a missing mandatory information item in the initial reconciliation request;

[0082] Step S203: Load the initial reconciliation request and the second reconciliation prompt corresponding to the initial reconciliation request into the second information item demand parsing model to obtain the second information item, wherein the second information item is the ambiguous information item in the initial reconciliation request;

[0083] Step S204: Based on the second information item and the target procurement entity archive, determine multiple pending confirmation information corresponding to the second information item;

[0084] Step S205: Based on the required supplementary content for the first information item and the confirmation instructions for the multiple pending confirmation information, optimize the initial reconciliation request to obtain the target reconciliation request, and perform a query in the target procurement entity archive database according to the target reconciliation request.

[0085] In this embodiment of the invention, for example, in today's digital business environment, enterprise procurement activities are frequent, involving a large amount of contract and transaction data, making reconciliation work extremely complex. This intelligent reconciliation method for procurement entities, leveraging advanced technology, can efficiently handle the reconciliation process, improving work accuracy and efficiency. The following uses a server as the execution entity to illustrate each step of this intelligent reconciliation method with detailed scenario examples. Assume the server serves a large chain supermarket group that has business dealings with numerous suppliers. The server stores relevant data for each procurement entity (i.e., supplier), which is integrated to form a target procurement entity archive. This archive contains massive amounts of information, such as detailed terms of procurement contracts, transaction records, delivery information, payment details, etc. One day, the group's procurement reconciliation staff discovers a discrepancy in accounts with a certain supplier and wants to verify the accounts for a recent batch of food purchases. Therefore, the staff sends an initial reconciliation request to the server through the group's internal reconciliation system. This request might be expressed as: "Query reconciliation information with [supplier name] regarding recent food purchases." Upon receiving the request, the server simultaneously retrieves the corresponding target procurement entity archive, which contains all procurement data records related to that supplier, preparing to begin the subsequent reconciliation process. In the scenario of the aforementioned supermarket group, the generation of the first reconciliation prompt follows a specific process. The server first obtains the first contract terms specifications and first clause performance examples corresponding to each potentially missing first information item (such as the procurement contract number, specific procurement date range, etc.). For example, the first contract terms specification for the procurement contract number clearly states that it should have a specific format, consisting of a combination of numbers and letters, and should be prominently displayed at the beginning of the contract; the first clause performance examples are the actual presentation styles of that number in numerous past contracts. Based on these, the server constructs a second logical judgment component to determine whether the initial reconciliation request contains a certain first information item. Simultaneously, it constructs a clause matching prompt component and a clause missing prompt component. The clause matching prompt component is configured to provide a contract clause association prompt when the request contains a certain first information item, such as providing the specific clause location of the information item in the contract and related explanations; the clause missing prompt component, when the request does not contain a certain first information item, provides a warning about missing contract elements, such as "The key procurement contract number is missing, which may affect accurate reconciliation." The server loads the initial reconciliation request, "Query reconciliation information regarding recent food purchases with [Supplier Name]," and the corresponding first reconciliation prompt into the first information item requirement parsing model. After analysis, the model determines that the request lacks the required purchase contract number, and therefore outputs "Purchase Contract Number" as the first information item. Similarly, using a supermarket group as an example, the acquisition of the second reconciliation prompt follows a specific process. The server retrieves the second contract terms and second contract performance examples for each potentially ambiguous second information item (such as the specific time range of "recent," the specific category definition of "food," etc.).For example, regarding the definition of the "recent" time range, the second contract clause might stipulate that 30 days from the contract signing date constitutes "recent"; the second clause performance examples demonstrate the actual time span corresponding to the "recent" expression in past contracts. Based on this, the server constructs a third logical judgment component to extract the second information item from the initial reconciliation request. Simultaneously, a fourth logical judgment component is constructed based on the priority rules for the reconciliation elements used to extract the second information item (including the mandatory parsing principle for monetary elements, the conflict handling principle for time-related elements, and the signature validity verification principle). For example, the mandatory parsing principle for monetary elements requires prioritizing the clarification of ambiguous information related to monetary amounts; the conflict handling principle for time-related elements specifies the handling method when there are conflicts in time definitions. The server loads the initial reconciliation request and the second reconciliation prompt into the second information item requirement parsing model. After model analysis, it is found that the expression "recent" is ambiguous because it does not specify a concrete time range, so the second information item is output as "'recent' time range". Taking the supermarket group as an example again, the target procurement entity archive contains a procurement contract clause knowledge graph, which contains multiple contract clause nodes. For the second information item, "'Recent' Time Range," the server queries multiple contract clause nodes for corresponding first-related clause instances. Assuming the server retrieves multiple instances of first-related clauses defining the "recent" time range (e.g., some contracts specify "recent" as within 15 days of contract signing, others within 30 days), the server adds these instances to the procurement terminology feature library. Next, referring to the procurement terminology feature library, the server performs structured parsing of the contract elements for "'Recent' Time Range," obtaining performance element atomic items, such as the smallest units defining time ranges like "within 15 days" and "within 30 days." Then, based on the multiple clause element constraints corresponding to the contract clause nodes in the procurement contract clause knowledge graph, if the performance element atomic items related to the "recent" time range correspond to multiple clause element constraints (e.g., different restrictions on the "recent" time range under different contract scenarios), the server constructs multiple pending confirmation information based on these conditions, which might be "within 15 days (applicable to fresh food procurement scenarios)" or "within 30 days (applicable to ambient temperature food procurement scenarios)," etc. In a supermarket group scenario, the staff member responsible for reconciliation, acting as the current administrator, sees a system prompt indicating that the first information item is missing the purchase contract number. They then enter the required supplementary information into the system interface, such as "[Specific Contract Number]". Simultaneously, for multiple pending confirmation information, such as "Within 15 days (applicable to fresh produce purchases)" and "Within 30 days (applicable to ambient temperature food purchases)," the staff member, based on the fact that this reconciliation involves ambient temperature food purchases, selects "Within 30 days (applicable to ambient temperature food purchases)" as the confirmation instruction.After receiving this content, the server supplements the first information item in the initial reconciliation request based on the required supplementary information, and optimizes the second information item in the initial reconciliation request based on the target confirmation information. The supplemented and optimized initial reconciliation request is then used as the target reconciliation request, which may become "Query reconciliation information for room temperature food purchases within the last 30 days between contract number [specific contract number] and [supplier name]". Then, the server executes a query in the target purchase entity's archive database based on this target reconciliation request, accurately filtering out matching reconciliation information from massive amounts of data and feeding it back to the staff. The server obtains the first contract category identifier (e.g., "food purchase contract" category identifier) ​​and the first transaction scenario feature (e.g., "room temperature food purchase scenario" feature) of the initial reconciliation request. Simultaneously, it obtains the set of past reconciliation requests from the staff member (current administrator) who initiated the initial reconciliation request. Within the set of past reconciliation requests, for each past request, it obtains its second contract category identifier and second transaction scenario feature. Suppose the server discovers that the second contract category identifier of a past reconciliation request and the first contract category identifier of the current request both belong to the "Food Procurement Contract" category, and both the second and first transaction scenario features are "Ambient Temperature Food Procurement Scenario." Then, the server retrieves the first past information item (such as the procurement contract number entered at the time) and the second past information item (such as the confirmation information for the "recent" time range selected at the time) from the past reconciliation request. Based on the first past information item, it constructs required supplementary content, such as "[Past Contract Number]"; based on the second past information item, it constructs a confirmation instruction, such as selecting the same "within 30 days (applicable to ambient temperature food procurement scenario)" as before. The server optimizes the initial reconciliation request based on this constructed content to obtain the target reconciliation request, such as "Query reconciliation information for ambient temperature food procurement within the last 30 days with contract number [Past Contract Number] and [Supplier Name]," and executes the query in the target procurement entity's archive database, providing staff with accurate reconciliation data.

[0086] In this embodiment of the invention, before loading the initial reconciliation request and the first reconciliation prompt word corresponding to the initial reconciliation request into the first information item demand parsing model to obtain the first information item, this embodiment of the invention provides the following implementation method.

[0087] Based on the initial reconciliation request, query the target procurement entity's archive database to obtain the initial reconciliation results;

[0088] The initial reconciliation request, the initial reconciliation result, and the third reconciliation prompt corresponding to the initial reconciliation request and the initial reconciliation result are loaded into the reconciliation complexity assessment model to obtain the inferred reconciliation complexity assessment result.

[0089] If the reconciliation complexity assessment result indicates that the initial reconciliation request does not require complex reconciliation processing, the execution of the query based on the initial reconciliation request will be terminated.

[0090] In this embodiment of the invention, for example, the server first receives an initial reconciliation request from a staff member, such as "query reconciliation information regarding recent food purchases with [supplier name]", and then queries the target purchasing entity's archive database based on this request. The server obtains contract element extraction rules, such as specifying that key information such as supplier name, purchase category, and time should be extracted from the request as contract elements. According to these rules, contract elements such as "[supplier name]", "food", and "recent" are extracted from the initial reconciliation request. Next, the server counts the number of times these contract elements match in each related transaction instance in the target purchasing entity's archive database, thereby determining the coverage of related transaction instances relative to the contract elements. At the same time, the strength of the clause constraints is determined based on the number of related transaction instances containing these contract elements. For example, if most related transaction instances mention the supplier and food purchases, the clause coverage and constraint strength are relatively high. Based on these two indicators, the server determines the reconciliation candidate set as the initial reconciliation result, and may draw a preliminary conclusion: there are a large number of recent food purchase transactions involving this supplier, and the time range of some transaction records needs to be further clarified. Next, the server retrieves the first business rule preset conditions for determining complex reconciliation processing, such as stipulating that complex reconciliation processing is required if the transaction involves a large amount of money, a long time span, or complex contract terms. Based on this, a first logic judgment component is constructed to determine whether the initial reconciliation request requires complex reconciliation processing. Simultaneously, an execution strategy component and an archiving strategy component are constructed; the former configures the specific process for complex reconciliation processing, and the latter sets the operations when complex reconciliation processing is not required. Combining these components, a third reconciliation prompt is generated. The server loads the initial reconciliation request, the initial reconciliation result, and the third reconciliation prompt into the reconciliation complexity assessment model. After analysis, if the model concludes that the reconciliation complexity assessment result indicates that the initial reconciliation request does not require complex reconciliation processing—for example, if the query only involves routine food purchases and the amount and time are within a simple range—the server terminates the query based on the initial reconciliation request, avoids subsequent complex processes, and directly feeds back the existing results to the staff, improving reconciliation efficiency.

[0091] In this embodiment of the invention, the step of querying the target procurement entity archive database according to the initial reconciliation request to obtain the initial reconciliation result can be implemented through the following example.

[0092] Obtain the rules for extracting contract elements;

[0093] According to the contract element extraction rules, extract contract elements from the initial reconciliation request;

[0094] The contract elements are loaded into the target procurement entity archive database, and the initial reconciliation results are obtained using the target query algorithm.

[0095] In this embodiment of the invention, exemplarily, the server receives an initial reconciliation request from the purchasing department, the content of which is "to query the reconciliation information regarding the purchase of snacks with Huihuang Foods supplier in October 2024". First, the server obtains the contract element extraction rules. These rules are pre-defined, for example, specifying that the supplier name, purchase time, and purchase category should be extracted as key contract elements from the request. Next, the server parses the initial reconciliation request according to the contract element extraction rules. From "to query the reconciliation information regarding the purchase of snacks with Huihuang Foods supplier in October 2024", it successfully extracts "Huihuang Foods supplier" as the supplier name element, "October 2024" as the purchase time element, and "snacks" as the purchase category element. Then, the server loads these extracted contract elements into the target purchasing entity archive. This archive stores a large number of purchase contracts and transaction records between the supermarket and numerous suppliers. The server uses a target query algorithm to search within the archive. The target query algorithm performs searches according to a specific logic and order. For example, it first locates all records related to the supplier Huihuang Foods based on the supplier name, then filters these records to find those from October 2024, and finally identifies the content related to the purchase of snack foods within these records. After the target query algorithm's search, the server obtains initial reconciliation results. For instance, the initial reconciliation results show that in October 2024, the supermarket had three snack food purchase transactions with the supplier Huihuang Foods. One transaction amounted to 5,000 yuan, and the purchased snacks included potato chips and nuts. The other two transactions also had corresponding detailed information, such as delivery time and payment status. These results provide the basic data for subsequent reconciliation operations.

[0096] In this embodiment of the invention, the step of loading the contract elements into the target procurement entity archive and obtaining the initial reconciliation result using a target query algorithm can be implemented through the following example.

[0097] The coverage of the contract elements relative to the contract elements is determined based on the number of times the contract elements are matched in each related transaction instance in the target procurement entity's archive.

[0098] Based on the number of related transaction instances containing the contract element in the target procurement entity's archive, the strength of the related transaction instance's clause constraint relative to the contract element is determined.

[0099] Based on the coverage and binding strength of the terms, a set of write-off candidates is determined from each related transaction instance in the target procurement entity's archive, which serves as the initial reconciliation result.

[0100] In this embodiment of the invention, an exemplary example is taken: a server of a large chain supermarket group processing a reconciliation request with the supplier "Meiweiyuan Food Factory". The initial reconciliation request is "to query the reconciliation information with Meiweiyuan Food Factory regarding biscuit purchases in the second half of 2024". The server has already extracted the contract elements "Meiweiyuan Food Factory", "second half of 2024", and "biscuits". The server loads these contract elements into the target purchasing entity archive and begins to calculate the clause coverage. The archive contains numerous related transaction instances, for example, 100 transaction instances involving Meiweiyuan Food Factory. For the contract element "biscuits", the server iterates through the clauses of each related transaction instance. If 80 related transaction instances explicitly mention clauses related to "biscuit" purchases, and these clauses vary in detail (some only mention the category, while others include specific biscuit brands, specifications, etc.), the server assigns different weights based on the detail and matching degree of the clauses. After calculation, the score corresponding to the number of clause matches for the "biscuit" contract element in these related transaction instances is obtained, and thus the clause coverage of the "biscuit" element is determined to be 80% (assuming the result after score conversion). Similarly, the coverage of the terms for "Meiweiyuan Food Factory" and "Second Half of 2024" is calculated. Next, the server determines the strength of the terms. For the "biscuit" contract element, there are 60 related transaction instances containing "biscuit" procurement content in the target procurement entity's archive, while the total number of related transaction instances is 500. The server calculates the proportion of related transaction instances containing the "biscuit" element to the total number of instances, concluding that the term strength of the "biscuit" element is 12% (60 ÷ 500). The term strength for "Meiweiyuan Food Factory" and "Second Half of 2024" is calculated in the same way. Finally, the server determines the write-off candidate set based on the coverage and strength of the terms. The server sets certain screening criteria, such as related transaction instances with a coverage of 60% and a strength of 10% being eligible for the write-off candidate set. Following this standard, the server filters through the related transaction instances in the target procurement entity's archive, identifying those that meet the criteria as the write-off candidate set; this is the initial reconciliation result. For example, 30 related transaction instances were selected, which recorded in detail the biscuit procurement transactions of Meiyuan Food Factory in the second half of 2024, including information such as purchase quantity, price, and delivery date. These instances constituted the initial reconciliation results and provided a basis for further reconciliation.

[0101] In this embodiment of the invention, the third reconciliation prompt is obtained by the following process, which can be implemented through the following example.

[0102] Obtain the first business rule preset conditions used for determining complex reconciliation processing;

[0103] Based on the preset conditions of the first business rule, a first logic determination component is constructed. The first logic determination component is configured to determine whether the initial reconciliation request requires complex reconciliation processing based on the preset conditions of the first business rule.

[0104] Construct an execution strategy component and an archiving strategy component. The execution strategy component configures the handling when the initial reconciliation request requires complex reconciliation processing, and the archiving strategy component configures the handling when the initial reconciliation request does not require complex reconciliation processing.

[0105] The third reconciliation prompt is constructed based on the first logic determination component, the execution strategy component, and the archiving strategy component.

[0106] In this embodiment of the invention, an exemplary example is still the server of a large chain supermarket group. When the server receives an initial reconciliation request from the purchasing department, such as "Querying reconciliation information regarding the purchase of imported fruits with [supplier name] in November 2024," it begins to obtain the third reconciliation prompt. First, the server obtains the first business rule preset conditions for determining complex reconciliation processing. For example, the preset conditions stipulate that complex reconciliation processing is required if the purchase involves goods with multiple different tax rates, the purchase amount exceeds 500,000 yuan, or the goods involve goods from multiple different origins. Next, based on these first business rule preset conditions, the server constructs a first logical determination component. This component analyzes the initial reconciliation request according to the above preset conditions. For example, for the request "Querying reconciliation information regarding the purchase of imported fruits with [supplier name] in November 2024," it searches the target purchasing entity's archive database to determine whether the imported fruit purchase involves goods with multiple different tax rates, whether the purchase amount exceeds 500,000 yuan, and whether the goods involve goods from multiple different origins, thereby determining whether the initial reconciliation request requires complex reconciliation processing. Next, the server constructs an execution strategy component and an archiving strategy component. The execution strategy component configures the specific handling method when the initial reconciliation request requires complex reconciliation processing. For example, if it is determined that complex reconciliation is required, the execution strategy component will arrange a dedicated finance team to conduct multiple rounds of detailed accounting, request suppliers to provide more detailed cost structures and pricing basis, and conduct sampling inspections of purchased goods, etc. The archiving strategy component, on the other hand, configures the handling method when the initial reconciliation request does not require complex reconciliation processing. For example, if it is determined that complex reconciliation is not required, the archiving strategy component will directly organize and archive the data related to this reconciliation according to the regular process, preparing for subsequent simple audits. Finally, based on the first logic judgment component, the execution strategy component, and the archiving strategy component, the server constructs a third reconciliation prompt. This prompt integrates the logic and processing methods of the above components, clearly indicating how the subsequent process should proceed. For example, the third reconciliation prompt might be stated as: "If this imported fruit purchase involves multiple commodities with different tax rates, the purchase amount exceeds 500,000 yuan, or involves commodities from multiple different origins, a complex reconciliation process will be initiated, with a dedicated finance team conducting detailed accounting and other operations; if none of the above situations apply, the standard process will be followed for filing." In this way, the third reconciliation prompt provides clear guidance for subsequent judgment and processing of the initial reconciliation request.

[0107] In this embodiment of the invention, the first reconciliation prompt is obtained by the following process, which can be implemented through the following example.

[0108] Obtain the first contract terms and first clause performance examples for each first information item;

[0109] Based on the first contract terms and the first performance example, a second logic determination component is constructed. The second logic determination component is configured to determine whether the initial reconciliation request contains the first information item based on the first contract terms and the first performance example.

[0110] Construct a clause matching prompt component and a clause missing prompt component. The clause matching prompt component is configured to provide a contract clause association prompt when the initial reconciliation request includes the first information item. The clause missing prompt component is configured to provide a contract element missing warning when the initial reconciliation request does not include the first information item.

[0111] The first reconciliation prompt word is constructed based on the second logic determination component, the clause matching prompt component, and the clause missing prompt component.

[0112] In this embodiment of the invention, taking, for example, a server of a large chain supermarket group, when the server receives the initial reconciliation request "Query reconciliation information with [supplier name] regarding recent food purchases," it begins to obtain the first reconciliation prompt. First, the server obtains the first contract terms specification and first clause performance examples for each first information item. For example, for the potentially missing first information item "Purchase Contract Number," the first contract terms specification clearly stipulates that the purchase contract number should consist of 10 digits, with the first 4 digits representing the year and the last 6 digits being a sequential number, and prominently marked in the upper left corner of the purchase contract's homepage. The first clause performance examples are actual purchase contract numbers extracted from a large number of past purchase contracts, such as "2023000001" and "2024000005," etc. Next, based on these first contract terms specification and first clause performance examples, the server constructs a second logical judgment component. This component acts like a precise detector, capable of meticulously analyzing the initial reconciliation request based on the first contract terms specification and first clause performance examples to determine whether the request contains the first information item "Purchase Contract Number." For example, it checks if the request text contains a string that matches a 10-digit number and has a specific format to determine if the initial reconciliation request includes this information item. Next, the server constructs a clause matching prompt component and a clause missing prompt component. The clause matching prompt component is configured to provide contract clause association prompts when the initial reconciliation request detects that it contains the first information item, "Purchase Contract Number." For example, it might prompt, "In the contract corresponding to Purchase Contract Number [specific number], Article 3 stipulates the product quality standards, and Article 5 clarifies the payment method," helping staff quickly understand the key clauses of the contract with that number. The clause missing prompt component is configured to provide a warning about missing contract elements when the initial reconciliation request does not contain "Purchase Contract Number," such as, "Your reconciliation request is missing the Purchase Contract Number. This information is a key element for accurate reconciliation; please complete it," reminding staff to pay attention and supplement the missing information. Finally, the server constructs the first reconciliation prompt based on the second logic judgment component, the clause matching prompt component, and the clause missing prompt component. This initial reconciliation prompt integrates the functions of the aforementioned components. It might be expressed as: "If the request contains a 10-digit purchase contract number in the correct format (the first 4 digits represent the year, and the last 6 digits are the sequential number), you can view the corresponding key terms of the contract; if it does not contain this, the purchase contract number, a key reconciliation element, is missing. Please provide it." In this way, the initial reconciliation prompt provides clear and explicit guidance for staff to process the initial reconciliation request.

[0113] In this embodiment of the invention, the construction of the clause matching prompt component and the clause missing prompt component can be implemented through the following examples.

[0114] The contract terms parsing process, terms matching status code, and terms performance description are obtained. The contract terms parsing process is the parsing process of the initial reconciliation request containing the first information item. The terms matching status code indicates that the initial reconciliation request contains the first information item. The terms performance description is the text information of the initial reconciliation request containing the first information item.

[0115] Based on the contract terms parsing process, the terms matching status codes, and the terms performance descriptions, construct the terms matching prompt component.

[0116] Obtain the contract reconciliation anomaly analysis process, element missing identifier, and element missing reason explanation. The contract reconciliation anomaly analysis process is the analysis process in which the initial reconciliation request does not contain the first information item. The element missing identifier indicates that the initial reconciliation request does not contain the first information item, and the element missing reason explanation indicates that the first information item is missing in the initial reconciliation request.

[0117] Based on the contract reconciliation anomaly analysis process, the missing element identifier, and the explanation of the missing element reason, the missing clause prompt component is constructed.

[0118] In this embodiment of the invention, an exemplary example is taken: a large chain supermarket group's server processes reconciliation requests. Assume the initial reconciliation request is "querying reconciliation information with [supplier name] regarding this month's grain and oil purchases." The server then constructs relevant components for the potentially missing first information item, "purchase contract number." First, it obtains the content required to construct the clause matching prompt component. The contract clause parsing process stipulates that if the initial reconciliation request includes "purchase contract number," the number must be extracted first, then the corresponding contract must be located in the contract database based on the number, and then each clause of the contract must be parsed. The clause matching status code is set to "01," indicating that the initial reconciliation request includes "purchase contract number." The clause performance description is text information such as "The purchase contract number is used to trace the detailed content of the purchase contract, including performance information such as product specifications, price, and delivery time." Next, the clause matching prompt component is constructed based on this information. This component integrates the contract terms parsing process, terms matching status codes, and terms performance descriptions. When it detects that the initial reconciliation request contains a "purchase contract number," it provides a prompt such as "Status code 01: Purchase contract number identified. Please follow the parsing process to first extract the number to locate the contract, and then understand the performance information such as product specifications, price, and delivery time," guiding staff to proceed. Then, it obtains information to construct a missing terms notification component. The contract reconciliation anomaly analysis process indicates that if the initial reconciliation request does not contain a "purchase contract number," the current reconciliation process must be paused, staff notified to supplement the information, and the anomaly recorded. The element missing identifier is set to "M001," representing that the initial reconciliation request does not contain a "purchase contract number." The reason for element missing is explained as "This reconciliation request lacks a purchase contract number used for accurate contract location and account verification." Finally, based on this information, a missing terms notification component is constructed. This component integrates the contract reconciliation anomaly analysis process, missing element identifiers, and explanations of missing element reasons. When it detects that the initial reconciliation request does not contain the "Purchase Contract Number," it will display the message: "Identifier M001: Missing Purchase Contract Number. Because this number is used to accurately locate contracts and verify accounts, reconciliation is now suspended. Please supplement the number and resubmit the request. The anomaly has been recorded." This clearly informs staff of the problem and the handling method. Through the above steps, the server completes the construction of the clause matching prompt component and the clause missing prompt component, providing comprehensive and clear guidance for processing initial reconciliation requests.

[0119] In this embodiment of the invention, the second reconciliation prompt is obtained by the following process, which can be implemented through the following example.

[0120] Obtain the second contract terms specifications and second contract performance examples for each second information item;

[0121] Based on the second contract terms specification and the second terms performance example, a third logic determination component is constructed. The third logic determination component is configured to extract the second information item from the initial reconciliation request based on the second contract terms specification and the second terms performance example.

[0122] Based on the priority rules for the reconciliation elements extracted from the second information item, a fourth logical judgment component is constructed; the priority rules for the reconciliation elements include the principle of mandatory parsing of amount elements, the principle of conflict handling of timeliness elements, and the principle of verification of signature validity.

[0123] The second reconciliation prompt is constructed based on the third and fourth logic determination components.

[0124] In this embodiment of the invention, taking, for example, a server of a large chain supermarket group, when the server receives the initial reconciliation request "Query reconciliation information with [supplier name] regarding recent large-volume beverage purchases," it begins to obtain the second reconciliation prompt. First, the server obtains the second contract terms specification and second clause performance examples for each second information item. For example, regarding the potentially ambiguous second information item "recently," the second contract terms specification stipulates that "recently" generally refers to within 30 days from the contract signing date, but for seasonal goods, the range can be appropriately adjusted according to seasonal characteristics. The second clause performance examples are practical application examples of "recently" collected from past contracts, such as in a beverage purchase contract where "recently" is explicitly defined as within 20 days before a summer promotional event. Next, based on these second contract terms specification and second clause performance examples, the server constructs a third logical judgment component. This component can filter and analyze the initial reconciliation request based on the second contract terms specification and second clause performance examples, extracting second information items such as "recently" from the request. For example, it analyzes the expression of "recently" in the request, combining the contract terms specification and performance examples to determine its potential meaning in the current context. Then, the server constructs a fourth logical judgment component based on the priority rules for the reconciliation elements extracted from the second information item. The priority rules for reconciliation elements include the mandatory parsing principle for monetary elements, the conflict resolution principle for timeliness elements, and the signature validity verification principle. Taking the second information item "recent" as an example, the mandatory parsing principle for monetary elements requires that if the calculation of beverage purchase amounts is related to the "recent" time range, the accurate time range corresponding to the amount calculation must be clearly defined first. The conflict resolution principle for timeliness elements stipulates that if the time definition of "recent" conflicts with the timeliness stipulated in other contracts, it will be handled according to specific rules, such as taking the special timeliness of seasonal goods as the standard. The signature validity verification principle emphasizes that if there is a dispute over the interpretation of "recent," the final interpretation right must be confirmed based on the signature on the contract. Finally, the server constructs a second reconciliation prompt based on the third and fourth logical judgment components. This second reconciliation prompt combines the functions of the two components and might be expressed as: "For the phrase 'recent,' we will extract and analyze information based on contract terms and past performance examples. If monetary calculations are involved, we will prioritize clarifying the corresponding time frame; if there is a conflict with other timeframes, we will handle it according to the special timeframes for seasonal goods; if there is a dispute over the interpretation, we will confirm the right of interpretation with a signature." In this way, the second reconciliation prompt provides staff with detailed and orderly processing guidance when dealing with ambiguous information items in the initial reconciliation request.

[0125] In this embodiment of the invention, the target procurement entity archive includes a procurement contract terms knowledge graph, which contains multiple contract terms nodes.

[0126] The step of determining multiple pending confirmation information corresponding to the second information item based on the second information item and the target procurement entity archive can be implemented through the following example.

[0127] For each of the second information items, query the first associated clause instance corresponding to the second information item in the plurality of contract clause nodes;

[0128] If there are multiple instances of the first associated clause retrieved, add all instances of the first associated clause to the procurement term feature library;

[0129] Referring to the aforementioned procurement terminology feature library, the second information item is subjected to contract element structured parsing to obtain performance element atomic items;

[0130] Determine the multiple pending confirmation information corresponding to the atomic items of the performance element.

[0131] In this embodiment of the invention, taking the server processing reconciliation business of a large chain supermarket group as an example, the target procurement entity archive managed by the server contains a procurement contract terms knowledge graph, which is filled with multiple contract terms nodes. Assume the initial reconciliation request received by the server is "querying reconciliation information with [supplier name] regarding recent large-scale purchases of snacks," and analysis reveals "recent" as the second information item. First, for this second information item, "recent," the server searches through multiple contract terms nodes in the procurement contract terms knowledge graph to find corresponding first-related clause instances. Since the knowledge graph covers numerous contract terms, the server discovers multiple first-related clause instances related to "recent" during the query process. For example, in some snack procurement contract terms, "recent" is defined as within 15 days after the contract is signed, used for emergency replenishment scenarios; in other contracts, "recent" refers to within one month, applicable to the regular promotion preparation phase. Because multiple first-related clause instances are found, the server adds these instances to a procurement term feature library. This procurement term feature library contains the definitions and application examples of various procurement terms in different contract contexts. Next, the server refers to the procurement terminology feature library to perform structured parsing of the second information item, "Recently," into contract elements. Based on different definitions and examples in the feature library, it breaks down the broad expression "Recently" into more specific performance element atoms. For example, it parses performance element atoms such as "Within 15 days (emergency replenishment scenario)" and "Within one month (regular promotion preparation stage)." Finally, the server determines multiple pending confirmation information corresponding to these performance element atoms. Based on the previously parsed performance element atoms, combined with factors such as the supermarket's and supplier's past transaction habits, current market conditions, and the overall contract background, multiple pending confirmation information are determined. For example, it might determine pending confirmation information such as "Within 15 days (considering the recent large fluctuations in market demand for snacks, possibly for emergency replenishment)" and "Within one month (based on the past regular promotion preparation cycle)," providing multiple possibilities for further clarifying the precise meaning of "Recently," allowing staff to make a final confirmation based on the actual situation.

[0132] In this embodiment of the invention, the procurement contract terms knowledge graph also includes multiple clause element constraints corresponding to the contract terms node;

[0133] The determination of the multiple pending confirmation information corresponding to the atomic item of the performance element can be implemented through the following example.

[0134] Based on the atomic items of the performance elements, query the knowledge graph of the procurement contract terms;

[0135] If a second associated clause instance corresponding to the atomic item of the performance element is obtained by querying the knowledge graph of the procurement contract terms, and the second associated clause instance corresponds to multiple second clause element constraints in the knowledge graph of the procurement contract terms, then the multiple pending confirmation information is constructed based on the multiple second clause element constraints.

[0136] In this embodiment of the invention, taking, for example, a server of a large chain supermarket group, the knowledge graph of procurement contract terms in its target procurement entity archive not only has multiple contract term nodes, but each node also corresponds to multiple clause element constraints. Taking the initial reconciliation request of "querying reconciliation information with [supplier name] regarding recent large-scale purchases of snacks" as an example, the server has parsed out the atomic items of the "recent" performance element, such as "within 15 days (emergency replenishment scenario)" and "within one month (regular promotion preparation stage)". For the atomic item of the performance element "within 15 days (emergency replenishment scenario)", the server queries the procurement contract term knowledge graph based on it. In the graph, the server obtains the second associated clause instance corresponding to this atomic item of the performance element. For example, a contract clause is found stipulating that in an emergency replenishment scenario, if the purchase amount exceeds 50,000 yuan, the replenishment must be completed within 15 days, and the supplier must provide expedited delivery service; if the purchase amount is less than 50,000 yuan, although the replenishment must also be completed within 15 days, the delivery service is carried out according to the regular process. Here, "whether the purchase amount exceeds 50,000 yuan" is one of the multiple second clause element constraints. Because of these second clause element constraints, the server constructs multiple pending confirmation messages based on them. Specifically, if the emergency replenishment purchase amount for snacks with [supplier name] exceeds 50,000 yuan, the pending confirmation message is "Replenishment to be completed within 15 days, with expedited delivery service provided by the supplier"; if the purchase amount is less than 50,000 yuan, the pending confirmation message is "Replenishment to be completed within 15 days, following the regular delivery process." Similarly, for the performance element atomic item "Within one month (regular promotion preparation phase)," the server queries the purchase contract terms knowledge graph to find the corresponding second related clause instance and related second clause element constraints. For example, during the regular promotion preparation phase, if the promotion covers all stores in the supermarket, the purchase needs to be completed in two batches within one month, with the first batch arriving within 15 days; if it only covers some regional stores, the purchase can be completed in one go within one month. Based on these conditions, the server constructs the corresponding pending confirmation messages. If the regular promotional activities corresponding to this snack purchase cover all stores in the supermarket, the pending confirmation information is "purchased in two batches within one month, with the first batch arriving within 15 days"; if it only covers stores in certain areas, the pending confirmation information is "purchased in one go within one month". In this way, the server uses the constraints of the clause elements in the purchase contract terms knowledge graph to construct multiple pending confirmation information for each performance element atomic item, providing a more comprehensive and detailed reference for accurate reconciliation later.

[0137] In this embodiment of the invention, if the second associated clause instance corresponding to the atomic item of the performance element is obtained by querying the knowledge graph of the procurement contract terms, and the second associated clause instance corresponds to multiple second clause element constraints in the knowledge graph of the procurement contract terms, then the multiple pending confirmation information is constructed according to the multiple second clause element constraints. This can be implemented through the following example.

[0138] It is determined that querying the procurement contract terms knowledge graph yields a second associated clause instance corresponding to the atomic item of the performance element, and in the procurement contract terms knowledge graph, the second associated clause instance corresponds to multiple second clause element constraints;

[0139] The performance element atomic item and the fifth reconciliation prompt word corresponding to the performance element atomic item are loaded into the clause element constraint inference model to obtain at least one supplementary clause element constraint condition corresponding to the performance element atomic item.

[0140] The plurality of pending confirmation information is constructed based on the constraints of the plurality of second clause elements and the constraints of the at least one supplementary clause element.

[0141] In this embodiment of the invention, taking, for example, a server of a large chain supermarket group, the server is processing a reconciliation request regarding snack purchases with [supplier name] and has derived a "recent" fulfillment element atomic item, such as "within 15 days (emergency replenishment scenario)". First, the server determines that a query of the purchase contract terms knowledge graph has yielded a second associated clause instance corresponding to the fulfillment element atomic item "within 15 days (emergency replenishment scenario)". For example, a contract clause regarding emergency replenishment is found, stipulating that if the purchased snacks are perishable, delivery must be completed within 15 days, and cold chain transportation must be guaranteed; if they are ordinary, conventional transportation methods will be used. Here, "the type of snacks purchased (perishable or ordinary)" constitutes multiple second clause element constraints. Next, the server loads the fulfillment element atomic item "within 15 days (emergency replenishment scenario)" and its corresponding fifth reconciliation prompt into the clause element constraint inference model. The fifth reconciliation prompt may contain guiding information such as "Based on past emergency replenishment contracts, consider the impact of product characteristics on replenishment conditions". Through this model, the server obtains at least one supplementary clause constraint corresponding to the atomic item of the performance element. For example, based on past contract data and business logic, the model infers that if the weight of the purchased snacks exceeds a certain standard, additional loading and unloading equipment may be required; this is a supplementary clause constraint. Finally, based on the existing multiple second clause constraints (such as the category of purchased snacks) and the newly inferred at least one supplementary clause constraint (such as the need for additional loading and unloading equipment if the weight of purchased snacks exceeds a certain standard), the server constructs multiple pending confirmation messages. If the purchased snacks are perishable and exceed the standard weight, the pending confirmation message is "Replenishment to be completed within 15 days, ensuring cold chain transportation and providing additional loading and unloading equipment"; if the purchased snacks are ordinary and do not exceed the standard weight, the pending confirmation message is "Replenishment to be completed within 15 days, using conventional transportation methods, without the need for additional loading and unloading equipment." Similarly, for other performance element atomic items in the "recent" period, such as "within one month (regular promotion preparation phase)," the server operates according to this process. First, the corresponding second related clause instance and multiple second clause constraint conditions are determined; for example, the scale of the promotional activity will affect the procurement arrangements. Then, the atomic item of the performance element and the corresponding fifth reconciliation prompt are input into the clause element constraint inference model to obtain supplementary clause element constraints. For example, it is inferred that if the purchase volume reaches a certain value, the supplier must provide additional gifts. Finally, based on these conditions, pending confirmation information is constructed. If the promotional activity is large-scale and the purchase volume reaches the value, the pending confirmation information is "Purchase completed within one month, supplier provides additional gifts"; if the promotional activity is small-scale and the purchase volume does not reach the value, the pending confirmation information is "Purchase completed within one month, no additional gifts". Through such meticulous steps, the server provides comprehensive and accurate pending confirmation information for accurate reconciliation.

[0142] In this embodiment of the invention, after querying the knowledge graph of the procurement contract terms based on the atomic items of the performance elements, the method further includes:

[0143] If the knowledge graph of the procurement contract terms does not yield a second associated clause instance corresponding to the atomic item of the performance element, the atomic item of the performance element and the fourth reconciliation prompt word corresponding to the atomic item of the performance element are loaded into the contract element compliance verification model.

[0144] If the contract element compliance verification model identifies that the performance element atomic item is not a standard contract atomic item, a contract element update instruction is constructed, which is used to instruct the update of the performance element atomic item.

[0145] In this embodiment of the invention, an exemplary example is still taken: a large chain supermarket group's server processing a snack procurement reconciliation request with a supplier. When processing the second information item "recent," the server has already parsed out a performance element atomic item such as "within 15 days (special event preparation scenario)." Following the process, the server queries the procurement contract terms knowledge graph based on this performance element atomic item. However, this time, it does not obtain a second associated clause instance corresponding to "within 15 days (special event preparation scenario)" in the knowledge graph. Therefore, the server loads the performance element atomic item "within 15 days (special event preparation scenario)" and its corresponding fourth reconciliation prompt into the contract element compliance verification model. The fourth reconciliation prompt may contain content such as "Check whether the performance element atomic item complies with the supermarket procurement contract standards and specifications, paying attention to time limits, event scenario descriptions, etc.," guiding the model to perform targeted verification. The contract element compliance verification model begins working, performing a comprehensive analysis of "within 15 days (special event preparation scenario)." During the analysis, the model, based on the pre-defined standard contract atomic item specifications, identified that the description of "special event preparation scenario" was unclear and did not meet the requirements of standard contract atomic items, thus not belonging to the standard contract atomic item category. Based on this identification result, the server constructs a contract element update instruction. For example, the contract element update instruction might be expressed as: "The description of 'special event preparation scenario' in 'within 15 days (special event preparation scenario)' is vague and does not meet the requirements of standard contract atomic items. Please clarify the key information such as the activity type and purpose to update this performance element atomic item." This instruction clearly informs relevant personnel that the performance element atomic item needs to be modified to conform to the standard contract atomic item specifications, so that subsequent reconciliation and related business operations can be performed more accurately, ensuring the standardization and consistency of procurement contracts and avoiding potential risks and disputes caused by unclear contract elements. Through this rigorous process, the server can promptly detect and handle non-standard contract element issues that occur during the reconciliation process, ensuring the smooth progress of procurement operations and the accuracy of financial data.

[0146] In this embodiment of the invention, the fourth reconciliation prompt is obtained by the following process, which can be implemented through the following example.

[0147] Based on the contract terms specification of the standard contract atomic item, a fifth logic determination component is constructed. The fifth logic determination component is configured to determine whether the performance element atomic item is the standard contract atomic item based on the contract terms specification of the standard contract atomic item.

[0148] Based on the compliance verification rules of the standard contract atomic items, a compliance verification rule determination component is constructed;

[0149] The fourth reconciliation prompt is constructed based on the fifth logic determination component and the compliance verification rule determination component.

[0150] In this embodiment of the invention, an exemplary example is taken: a large chain supermarket group's server handles snack procurement reconciliation transactions. During the processing, for fulfillment element atomic items such as "within 15 days (special event preparation scenario)," the server begins to obtain the fourth reconciliation prompt. First, the server constructs a fifth logical judgment component based on the contract terms of the standard contract atomic item. For example, the standard contract atomic item stipulates that time-related descriptions must be precise to the specific activity type, such as "within 15 days (Spring Festival promotion activity preparation scenario)." The fifth logical judgment component then determines whether "within 15 days (special event preparation scenario)" conforms to the standard contract atomic item based on this specification. It analyzes whether the expression of "special event preparation scenario" is clear and whether it follows the requirements of the standard contract atomic item for activity scenario description. Next, the server constructs a compliance verification rule judgment component based on the clause compliance verification rules of the standard contract atomic item. Assuming that the clause compliance verification rules stipulate that the activity scenario description must be closely related to common supermarket promotions, procurement, and other business activities, and cannot contain ambiguous expressions, the compliance verification rule judgment component will further examine whether "within 15 days (special event preparation scenario)" is compliant based on these rules. It checks whether the "special event preparation scenario" is relevant to the supermarket business and whether there is any ambiguity. Finally, the server constructs the fourth reconciliation prompt based on the fifth logic judgment component and the compliance verification rule judgment component. If the fifth logic judgment component finds that the description of "special event preparation scenario" is unclear and does not conform to the standard contract atomic item, and the compliance verification rule judgment component also determines that it does not conform to the clause compliance verification rules, the fourth reconciliation prompt may be: "Your input 'within 15 days (special event preparation scenario)' does not conform to the standard contract atomic item specification. The standard requires that the description of the event scenario be accurate and closely related to the supermarket business. The description of 'special event preparation scenario' is vague and may lead to deviations in the understanding and execution of the contract. Please clarify the key information such as the event type." Through this construction process, the fourth reconciliation prompt can clearly inform relevant personnel of the problems with the performance element atomic item, guide them to make corrections according to the requirements of the standard contract atomic item, ensure the compliance of the contract elements, and lay the foundation for accurate reconciliation and business operations in the future.

[0151] In this embodiment of the invention, the step of optimizing the initial reconciliation request based on the required supplementary content for the first information item and the confirmation instructions for the plurality of pending confirmation information to obtain the target reconciliation request can be implemented through the following example.

[0152] Receive the required supplementary content for the first information item loaded by the current administrator and the confirmation instruction to select the target confirmation information from the plurality of pending confirmation information;

[0153] Based on the required supplementary information, supplement the first information item in the initial reconciliation request;

[0154] Based on the target confirmation information, the second information item in the initial reconciliation request is optimized, and the initial reconciliation request after supplementing the first information item and optimizing the second information item is taken as the target reconciliation request.

[0155] In this embodiment of the invention, taking a large chain supermarket group as an example, the server receives an initial reconciliation request: "Query reconciliation information with [supplier name] regarding recent large-scale purchases of snacks." After processing, the server finds that the first information item, such as "Purchase Contract Number," is missing. Simultaneously, for "recent," multiple pending confirmation information items are generated, such as "Within 15 days (urgent replenishment scenario, purchase amount exceeding 50,000, requiring expedited delivery)" and "Within one month (regular promotion preparation stage, promotion covering the entire supermarket, purchased in two batches)." At this time, the current administrator (the supermarket staff responsible for purchase reconciliation) performs relevant operations on the system operation interface. The administrator enters the required supplementary information, such as "SC202411001," at the location where the system prompts for the missing "Purchase Contract Number." Simultaneously, based on the actual situation of this purchase transaction, the administrator selects "Within 15 days (urgent replenishment scenario, purchase amount exceeding 50,000, requiring expedited delivery)" as the target confirmation information from the multiple pending confirmation information items and issues a confirmation instruction. The server receives the mandatory supplementary information "SC202411001" for the "Purchase Contract Number" loaded by the administrator, along with a confirmation instruction selecting "Within 15 days (urgent replenishment scenario, purchase amount exceeding 50,000, requiring expedited delivery)". Next, based on the mandatory supplementary information "SC202411001", the server supplements the first information item "Purchase Contract Number" in the initial reconciliation request. Then, based on the target confirmation information "Within 15 days (urgent replenishment scenario, purchase amount exceeding 50,000, requiring expedited delivery)", the server optimizes the second information item "Recent" in the initial reconciliation request. The original vague term "Recent" is now clearly defined as "Within 15 days (urgent replenishment scenario, purchase amount exceeding 50,000, requiring expedited delivery)". Finally, the server uses the initial reconciliation request with the supplemented "Purchase Contract Number" as "SC202411001" and the optimized "Recent" as "Within 15 days (urgent replenishment scenario, purchase amount exceeding 50,000, requiring expedited delivery)" as the target reconciliation request. The target reconciliation request now reads: "Query the reconciliation information for large-volume snack purchases within 15 days between contract number SC202411001 and [supplier name] (emergency replenishment scenario, purchase amount exceeding 50,000, requiring expedited delivery)." This optimized target reconciliation request is more accurate and detailed, allowing the server to perform precise queries in the target purchase entity's archive database, providing administrators with reconciliation results that better meet their needs.

[0156] In this embodiment of the invention, the step of optimizing the initial reconciliation request based on the required supplementary content for the first information item and the confirmation instructions for the plurality of pending confirmation information to obtain the target reconciliation request can be implemented through the following example.

[0157] Obtain the first contract classification identifier and the first transaction scenario characteristics of the initial reconciliation request;

[0158] Obtain the set of past reconciliation requests of the current administrator, where the current administrator is the administrator who initiated the initial reconciliation request;

[0159] For each past reconciliation request in the set of past reconciliation requests, obtain the second contract classification identifier and the second transaction scenario feature of the past reconciliation request;

[0160] If the second contract classification identifier of a past reconciliation request in the set of past reconciliation requests matches the first contract classification identifier, and the second transaction scenario feature of the past reconciliation request matches the first transaction scenario feature, then the first past information item and the second past information item in the past reconciliation request are obtained.

[0161] Based on the first past information item, the required supplementary content is constructed, and based on the second past information item, the confirmation instruction is constructed;

[0162] Based on the required supplementary information and the confirmation instruction, the initial reconciliation request is optimized to obtain the target reconciliation request.

[0163] In this embodiment of the invention, taking a large chain supermarket group as an example, the server receives an initial reconciliation request initiated by an administrator: "Query reconciliation information with [Supplier A] regarding a recent batch of daily necessities purchases." First, the server obtains the first contract classification identifier and the first transaction scenario feature of the initial reconciliation request. After analysis, the first contract classification identifier is determined to be "daily necessities purchase contract," and the first transaction scenario feature is identified as "regular bulk purchase, recent time range." Next, the server obtains the set of past reconciliation requests from the administrator who initiated the request. This administrator is responsible for the reconciliation of daily necessities purchases for the supermarket, and there are many related reconciliation request records in the past. Then, for each past reconciliation request in the set of past reconciliation requests, the server obtains its second contract classification identifier and second transaction scenario feature. For example, one past reconciliation request is "Query reconciliation information with [Supplier A] regarding a batch of shampoo purchases within one month," its second contract classification identifier is "daily necessities purchase contract," and its second transaction scenario feature is "regular bulk purchase, within one month." At this point, the server discovers that the second contract category identifier of this past reconciliation request matches the first contract category identifier "Daily Necessities Purchase Contract" of the current initial reconciliation request, and the second transaction scenario feature "Regular bulk purchase, within one month" has a high degree of matching with the first transaction scenario feature "Regular bulk purchase, recent time range" (here, "recent" and "within one month" are considered a match in this business scenario). Therefore, the server retrieves the first and second past information items from this past reconciliation request. Assume the first past information item is "Purchase Contract Number: RJ20241005," and the second past information item is "Within one month (Regular Purchase Scenario)." Next, the server constructs required supplementary content based on the first past information item "Purchase Contract Number: RJ20241005," and simultaneously constructs a confirmation instruction based on the second past information item "Within one month (Regular Purchase Scenario)." Finally, the server optimizes the initial reconciliation request based on the constructed required supplementary content and confirmation instruction. The "Purchase Contract Number: RJ20241005" was added to the missing first information item (Purchase Contract Number) in the initial reconciliation request, and the second information item, "Recent," was optimized to "Within one month (regular procurement scenario)." Thus, the optimized initial reconciliation request became the target reconciliation request: "Query the reconciliation information for a batch of daily necessities purchased from [Supplier A] within one month (regular procurement scenario) with contract number RJ20241005." In this way, the server leverages the administrator's past reconciliation experience to optimize the initial reconciliation request, making it more accurate. This allows for more effective subsequent queries in the target procurement entity's archive database, providing the administrator with more accurate reconciliation results.

[0164] This invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned intelligent reconciliation method for the purchasing entity. Figure 2 As shown, Figure 2 This is a structural block diagram of a computer device 100 provided in an embodiment of the present invention. The computer device 100 includes a memory 111, a processor 112, and a communication unit 113. To enable data transmission or interaction, the memory 111, processor 112, and communication unit 113 are electrically connected to each other directly or indirectly. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0165] For illustrative purposes, the foregoing description has been made with reference to specific embodiments. However, the foregoing illustrative discussions are not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed. Numerous modifications and variations are possible in accordance with the foregoing teachings. These embodiments were chosen and described in order to best illustrate the principles of the present disclosure and its practical application, thereby enabling those skilled in the art to best utilize the disclosure and to employ various embodiments with different modifications to suit a particular intended application.

Claims

1. An intelligent reconciliation method for procurement entities, characterized in that, include: Obtain the target procurement entity archive database and an initial reconciliation request for querying the target procurement entity archive database; The initial reconciliation request and the first reconciliation prompt word corresponding to the initial reconciliation request are loaded into the first information item requirement parsing model to obtain the first information item, wherein the first information item is a missing mandatory information item in the initial reconciliation request; The initial reconciliation request and the second reconciliation prompt corresponding to the initial reconciliation request are loaded into the second information item requirement parsing model to obtain the second information item, wherein the second information item is the ambiguous information item in the initial reconciliation request; Based on the second information item and the target procurement entity archive, determine multiple pending confirmation information items corresponding to the second information item; Based on the required supplementary content for the first information item and the confirmation instructions for the multiple pending confirmation information, the initial reconciliation request is optimized to obtain the target reconciliation request, and a query is performed in the target procurement entity archive database according to the target reconciliation request. The first reconciliation prompt is obtained through the following process: Obtain the first contract terms and first clause performance examples for each first information item; Based on the first contract terms and the first performance example, a second logic determination component is constructed. The second logic determination component is configured to determine whether the initial reconciliation request contains the first information item based on the first contract terms and the first performance example. Construct a clause matching prompt component and a clause missing prompt component. The clause matching prompt component is configured to provide a contract clause association prompt when the initial reconciliation request includes the first information item. The clause missing prompt component is configured to provide a contract element missing warning when the initial reconciliation request does not include the first information item. Based on the second logic determination component, the clause matching prompt component, and the clause missing prompt component, the first reconciliation prompt word is constructed; The construction of the clause matching prompt component and the clause missing prompt component includes: The contract terms parsing process, terms matching status code, and terms performance description are obtained. The contract terms parsing process is the parsing process of the initial reconciliation request containing the first information item. The terms matching status code indicates that the initial reconciliation request contains the first information item. The terms performance description is the text information of the initial reconciliation request containing the first information item. Based on the contract terms parsing process, the terms matching status codes, and the terms performance descriptions, construct the terms matching prompt component. Obtain the contract reconciliation anomaly analysis process, element missing identifier, and element missing reason explanation. The contract reconciliation anomaly analysis process is the analysis process in which the initial reconciliation request does not contain the first information item. The element missing identifier indicates that the initial reconciliation request does not contain the first information item, and the element missing reason explanation indicates that the first information item is missing in the initial reconciliation request. Based on the contract reconciliation anomaly analysis process, the missing element identifier, and the explanation of the missing element reasons, construct the missing clause prompt component; The second reconciliation prompt key is obtained through the following process: Obtain the second contract terms specifications and second contract performance examples for each second information item; Based on the second contract terms specification and the second terms performance example, a third logic determination component is constructed. The third logic determination component is configured to extract the second information item from the initial reconciliation request based on the second contract terms specification and the second terms performance example. Based on the priority rules for the reconciliation elements extracted from the second information item, a fourth logical judgment component is constructed; the priority rules for the reconciliation elements include the principle of mandatory parsing of amount elements, the principle of conflict handling of timeliness elements, and the principle of verification of signature validity. The second reconciliation prompt is constructed based on the third and fourth logic determination components.

2. The method according to claim 1, characterized in that, Before loading the initial reconciliation request and the first reconciliation prompt corresponding to the initial reconciliation request into the first information item demand parsing model to obtain the first information item, the method further includes: Based on the initial reconciliation request, query the target procurement entity's archive database to obtain the initial reconciliation results; The initial reconciliation request, the initial reconciliation result, and the third reconciliation prompt corresponding to the initial reconciliation request and the initial reconciliation result are loaded into the reconciliation complexity assessment model to obtain the inferred reconciliation complexity assessment result. If the reconciliation complexity assessment result indicates that the initial reconciliation request does not require complex reconciliation processing, the execution of the query based on the initial reconciliation request will be terminated.

3. The method according to claim 2, characterized in that, The step of querying the target procurement entity's archive database based on the initial reconciliation request to obtain the initial reconciliation result includes: Obtain the rules for extracting contract elements; According to the contract element extraction rules, extract contract elements from the initial reconciliation request; The coverage of the contract elements relative to the contract elements is determined based on the number of times the contract elements are matched in each related transaction instance in the target procurement entity's archive. Based on the number of related transaction instances containing the contract element in the target procurement entity's archive, the strength of the related transaction instance's clause constraint relative to the contract element is determined. Based on the coverage and binding strength of the terms, a set of write-off candidates is determined from each related transaction instance in the target procurement entity's archive, which serves as the initial reconciliation result.

4. The method according to claim 2, characterized in that, The third reconciliation prompt is obtained through the following process: Obtain the first business rule preset conditions used for determining complex reconciliation processing; Based on the preset conditions of the first business rule, a first logic determination component is constructed. The first logic determination component is configured to determine whether the initial reconciliation request requires complex reconciliation processing based on the preset conditions of the first business rule. Construct an execution strategy component and an archiving strategy component. The execution strategy component configures the handling when the initial reconciliation request requires complex reconciliation processing, and the archiving strategy component configures the handling when the initial reconciliation request does not require complex reconciliation processing. The third reconciliation prompt is constructed based on the first logic determination component, the execution strategy component, and the archiving strategy component.

5. The method according to claim 1, characterized in that, The target procurement entity archive contains a procurement contract terms knowledge graph, which contains multiple contract terms nodes. The step of determining multiple pending confirmation information corresponding to the second information item based on the second information item and the target procurement entity archive includes: For each of the second information items, query the first associated clause instance corresponding to the second information item in the plurality of contract clause nodes; If there are multiple instances of the first associated clause retrieved, add all instances of the first associated clause to the procurement term feature library; Referring to the aforementioned procurement terminology feature library, the second information item is subjected to contract element structured parsing to obtain performance element atomic items; Determine the multiple pending confirmation information corresponding to the atomic items of the performance element.

6. The method according to claim 5, characterized in that, The procurement contract terms knowledge graph also includes multiple clause element constraints corresponding to the contract terms nodes. The determination of the multiple pending confirmation information corresponding to the atomic item of the performance element includes: Based on the atomic items of the performance elements, query the knowledge graph of the procurement contract terms; If a second associated clause instance corresponding to the atomic item of the performance element is obtained by querying the knowledge graph of the procurement contract terms, and the second associated clause instance corresponds to multiple second clause element constraints in the knowledge graph of the procurement contract terms, then the multiple pending confirmation information is constructed based on the multiple second clause element constraints.

7. The method according to claim 6, characterized in that, If a query of the procurement contract terms knowledge graph yields a second associated clause instance corresponding to the atomic item of the performance element, and in the procurement contract terms knowledge graph, the second associated clause instance corresponds to multiple second clause element constraints, then based on the multiple second clause element constraints, the multiple pending confirmation information is constructed, including: It is determined that querying the procurement contract terms knowledge graph yields a second associated clause instance corresponding to the atomic item of the performance element, and in the procurement contract terms knowledge graph, the second associated clause instance corresponds to multiple second clause element constraints; The performance element atomic item and the fifth reconciliation prompt word corresponding to the performance element atomic item are loaded into the clause element constraint inference model to obtain at least one supplementary clause element constraint condition corresponding to the performance element atomic item. Based on the constraints of the multiple second clause elements and the constraints of the at least one supplementary clause element, the multiple pending confirmation information is constructed; After querying the knowledge graph of the procurement contract terms based on the atomic items of the performance elements, the method further includes: If the knowledge graph of the procurement contract terms does not yield a second associated clause instance corresponding to the atomic item of the performance element, the atomic item of the performance element and the fourth reconciliation prompt word corresponding to the atomic item of the performance element are loaded into the contract element compliance verification model. If the contract element compliance verification model identifies that the performance element atomic item is not a standard contract atomic item, a contract element update instruction is constructed, which is used to indicate the update of the performance element atomic item; The fourth reconciliation prompt key is obtained through the following process: Based on the contract terms specification of the standard contract atomic item, a fifth logic determination component is constructed. The fifth logic determination component is configured to determine whether the performance element atomic item is the standard contract atomic item based on the contract terms specification of the standard contract atomic item. Based on the compliance verification rules of the standard contract atomic items, a compliance verification rule determination component is constructed; The fourth reconciliation prompt is constructed based on the fifth logic determination component and the compliance verification rule determination component.

8. The method according to claim 1, characterized in that, The step of optimizing the initial reconciliation request based on the required supplementary content for the first information item and the confirmation instructions for the plurality of pending confirmation information to obtain the target reconciliation request includes: Receive the required supplementary content for the first information item loaded by the current administrator and the confirmation instruction to select the target confirmation information from the plurality of pending confirmation information; Based on the required supplementary information, supplement the first information item in the initial reconciliation request; Based on the target confirmation information, the second information item in the initial reconciliation request is optimized, and the initial reconciliation request after supplementing the first information item and optimizing the second information item is taken as the target reconciliation request.

9. The method according to claim 1, characterized in that, The step of optimizing the initial reconciliation request based on the required supplementary content for the first information item and the confirmation instructions for the plurality of pending confirmation information to obtain the target reconciliation request includes: Obtain the first contract classification identifier and the first transaction scenario characteristics of the initial reconciliation request; Obtain the set of past reconciliation requests of the current administrator, where the current administrator is the administrator who initiated the initial reconciliation request; For each past reconciliation request in the set of past reconciliation requests, obtain the second contract classification identifier and the second transaction scenario feature of the past reconciliation request; If the second contract classification identifier of a past reconciliation request in the set of past reconciliation requests matches the first contract classification identifier, and the second transaction scenario feature of the past reconciliation request matches the first transaction scenario feature, then the first past information item and the second past information item in the past reconciliation request are obtained. Based on the first past information item, the required supplementary content is constructed, and based on the second past information item, the confirmation instruction is constructed; Based on the required supplementary information and the confirmation instruction, the initial reconciliation request is optimized to obtain the target reconciliation request.

10. A server system, characterized in that, Includes a server, the server being used to perform the method according to any one of claims 1-9.

Images