A method for collecting and analyzing information feedback

Abstract

This application provides an information feedback collection and analysis method, including: counting the number of path nodes in the knowledge graph between the initial related entries and the original declaration field names after supplementation, identifying the number of cross-entry related entries, evaluating the relatedness level of each initial related entry according to a preset relatedness level table, and forming a list of leveled entries; classifying and splicing the original voucher names, approval process names, and original declaration field names of the target related entries according to the order of processing of each original voucher to its corresponding declaration entry and the approval process, to obtain a linked processing work order recording the order of processing of the original declaration field names and the target related entries; identifying the number of material items in the supplementary material list, determining the estimated processing cycle, and sending the supplementary material list and the estimated results to the target problem assignment node, and simultaneously feeding back to the enterprise's application end.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for collecting and analyzing information feedback. Background Technology

[0002] The efficient processing of feedback information regarding enterprise declaration results directly impacts the response speed of government services and the compliance costs for enterprises, holding immense strategic significance in the current context of digital government transformation. Existing technologies primarily employ keyword matching or semantic similarity calculation to establish a link between feedback text and declaration items, aiming to quickly identify the specific declaration field the feedback content refers to, thereby achieving initial information classification. However, in practical applications, it has been found that these conventional methods, when handling complex feedback scenarios involving multiple linked items, often only capture explicitly mentioned items, ignoring implicit business connections, leading to incomplete problem assignment and consequently affecting subsequent processing efficiency. In-depth investigation and analysis revealed that the root cause lies in the existing technological framework's focus on feedback location accuracy as a single optimization objective, failing to fully consider the inherent dependencies and logical connections between various fields in the enterprise declaration process. Specifically, to resolve the issue of ambiguous dispute points, conventional technical means typically attempt to precisely anchor the feedback content to a specific declaration field, such as questioning the "R&D expense deduction amount" in financial statements. While this high-precision positioning can clearly converge the focus of disputes and allow the system to quickly assign issues to the appropriate tax audit departments, this single-point-focused positioning method can severely obscure the business connections between the field and related items such as "Annual R&D Project List" and "R&D Personnel Hours Statistics." The underlying mechanism is that high-precision text anchoring is often limited to literal matching, lacking an understanding of deep business logic connections. This leads to the omission of related items that are not directly mentioned but actually require review from the assignment process. Improving the anchoring accuracy of feedback text often reduces the coverage of related items, resulting in information fragmentation; the two seem mutually exclusive. Therefore, how to improve the anchoring accuracy of feedback text for declaration items to accelerate the convergence of disputes, while dynamically identifying and including related items that are not directly mentioned but have business connections, avoiding information omissions due to high-precision positioning, and thus ensuring that issue assignment is both accurate and complete, has become a key technical challenge that urgently needs to be overcome in enterprise declaration feedback processing. Summary of the Invention

[0003] This invention provides a method for collecting and analyzing information feedback, the method comprising: Obtain the feedback text of the enterprise's application, extract the original application field names, reference positioning statements and disputed terms, and determine whether the reference positioning type is specific item anchoring or overall general pointing; Retrieve the enterprise's declaration ledger to establish a business knowledge graph. Query the initial related entries by the original declaration field name and the reference location description. If it is a general reference, then all declaration entries are listed as initial related entries. Count the number of path nodes between the initial related entries and the original declared field names in the knowledge graph, identify the number of cross-entry related entries, assess the relatedness level according to the preset relatedness level table, and form a list of leveled entries. Retrieve the level of involvement from the list of level items. If the disputed terms mention the item name corresponding to the general level of involvement, upgrade it to the close level of involvement. Select the close level of involvement items as the target involvement items and determine the name of the original voucher and the name of the approval process. The original voucher names, approval process names, and original declaration field names of the target-related items are categorized and concatenated according to the order of processing to obtain the linkage processing work order; Match the names of the approval stages of the linked processing work orders in the preset assignment rule library, determine the target issue assignment node, extract historical processing records, screen the processing files of disputes with similar terms to the focus of the dispute, and generate a supplementary materials list. Identify the number of material items in the supplementary materials list, determine the estimated processing cycle, and send the supplementary materials list to the target issue assignment node, while simultaneously feeding back to the enterprise's application end.

[0004] Furthermore, the feedback text obtained from the enterprise's application includes: The feedback text is obtained and divided into feedback statements according to periods, semicolons and line breaks. Each statement is matched against a pre-established table of application business terms. The terminology comparison table for declaration business records the standard names of fields, field aliases, declaration document names, and disputed keywords. The original declaration field names are extracted based on the matching results.

[0005] Furthermore, the step of retrieving enterprise declaration ledgers to establish a business knowledge graph includes: Obtain the enterprise's declaration ledger, establish declaration item nodes, original voucher nodes, and approval process nodes based on the item number, voucher number, and approval process name, and establish connections between the declaration item nodes and their corresponding original voucher nodes and approval process nodes to obtain a business knowledge graph.

[0006] Furthermore, the count of path nodes in the knowledge graph between the initial related entries and the original declared field names includes: Obtain the initial related entries. In the business knowledge graph, starting from the node of the original declaration field, find the connecting path to each initial related entry node. Take the connecting path with the fewest nodes as the target path. Count all the nodes contained in the target path to obtain the number of path nodes. Based on the target path, identify the declared entry nodes other than the starting point and the corresponding initial linked entry nodes, count them by deduplication according to the entry number, obtain the number of cross-entry linked entries, and combine them with the number of path nodes to form a linked count list; The level of involvement is determined by matching the list of involvement counts against the preset involvement level table.

[0007] Furthermore, the acquisition of the initial related entries, in the business knowledge graph, starting from the declaration entry node to which the original declaration field belongs, includes: Obtain the terms used to highlight the dispute, extract the names of other declaration items besides the original declaration field names according to the declaration business terminology comparison table, convert them into standard field names and remove duplicate names to obtain a list of items to be annotated; Based on the list of items to be annotated, the corresponding declaration item nodes are located and annotated in the business knowledge graph. Each annotated node is verified to see if it is connected to the same original voucher node or the same approval process node as the original declaration field, thus obtaining a list of associated items.

[0008] Furthermore, the level of involvement in the list of retrieved level entries includes: Obtain the initial related entries, starting from the declaration entry node to which the original declaration field belongs; If it is connected to the same original voucher node or approval process node as an initial related item, it is recorded as a level. If it is reached through at least one other declaration item node, it is recorded as an alternate level, thus obtaining a list of level positions. Based on the aforementioned hierarchical location list, first-level items are rated as closely related, and items in adjacent levels are rated as generally related.

[0009] Furthermore, the original voucher names, approval process names, and original declaration field names of the target-related items are categorized and concatenated according to the order of processing to obtain a linked processing work order, including: Obtain the target related entries, and based on the enterprise identifier, application batch identifier and entry number, read the corresponding application documents, original voucher names, application entry sequence numbers to which the vouchers belong, approval process names and processing order from the enterprise application ledger, and associate them with the original application field names to obtain the work order basic record set; Based on the aforementioned work order basic record set, the documents are categorized according to their respective application documents and arranged according to the application item number to which the document belongs; When the same application item corresponds to multiple approval stages, they are arranged according to the processing order of the approval stages to obtain the item processing sequence; According to the order of processing items, write in the original declaration field name, target related item name, original voucher name, approval process name and processing order in sequence.

[0010] Furthermore, the matching of the approval process name of the linked work order in the preset assignment rule base includes: Obtain the approval process name and application item in the linked processing work order, match them item by item with the approval process name, application item and acceptance node recorded in the preset assignment rule library, determine the target problem assignment node, and form a node matching record; Based on the node matching records, extract the historical processing records of the target issue assignment node, filter the processing files that simultaneously record the same field standard name and the same dispute focus words, and obtain a set of dispute files of the same type.

[0011] Furthermore, identifying the number of material items in the supplementary materials list includes: Obtain the supplementary materials list, remove duplicate material records according to the voucher name and the application item to which they belong, count the number of material items, and match the processing time limit corresponding to the interval to which the number of material items belongs in the preset processing cycle table to determine the estimated processing cycle result; The supplementary materials list and the estimated processing cycle are written into the linkage processing work order, distributed to the target problem assignment node according to the target problem assignment node identifier, and simultaneously fed back to the enterprise application end according to the enterprise identifier and the application batch identifier.

[0012] Furthermore, obtaining the business knowledge graph includes: Based on the business knowledge graph, locate the corresponding declaration item node according to the original declaration field name, determine the query starting point according to the item number, declaration document name and field name in the reference location description, and search for related declaration item nodes along the common original voucher nodes or approval process nodes to form a candidate related item list. For the candidate list of related items, if the reference positioning type is specific item anchoring, then the declared item in the list is used as the initial related item. If it is a general reference, then all the declaration entries of the declaration document to which the original declaration field belongs are added, and after removing duplicate entries, the initial related entries are obtained.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a method for collecting and analyzing information feedback, addressing issues in business scenarios such as the difficulty in accurately defining the scope of feedback information in enterprise application processes, incomplete identification of related items involved in disputes, and a lack of precise positioning in problem assignment and material supplementation. This invention constructs a business knowledge graph with application items, original vouchers, and approval stages as nodes, enabling intelligent parsing of original application fields and their references in feedback texts, distinguishing between specific item anchoring and general overall referencing types. Based on this, and combining other application items mentioned in the disputed terms, it analyzes shared voucher connections and connections within the same approval stage in the knowledge graph to automatically identify initial related items and count the number of path nodes and cross-item connections. Based on this, it assesses the level of involvement and upgrades the level of dispute-related items, thereby accurately determining the scope of target related items. Furthermore, it categorizes and splices these items according to their processing order to generate linked processing work orders, matches them with a preset assignment rule library to locate the target problem assignment node, and automatically generates a supplementary material list and processing cycle estimate by filtering similar historical dispute files. This invention achieves automated correlation analysis and precise problem positioning of enterprise application feedback information, effectively improving the efficiency and accuracy of application processing. Attached Figure Description

[0014] Figure 1 This is a flowchart of an information feedback collection and analysis method according to the present invention.

[0015] Figure 2 This is a schematic diagram of an information feedback collection and analysis method according to the present invention.

[0016] Figure 3 This is another schematic diagram of an information feedback collection and analysis method according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0018] like Figures 1-3 This embodiment of an information feedback collection and analysis method may specifically include: Step S101: Obtain the feedback text of the enterprise's declaration business, compare it sentence by sentence according to the declaration business terminology comparison table, extract the original declaration field names, reference positioning expressions and disputed terms, and determine the reference positioning type as specific item anchoring or overall general pointing according to whether the reference positioning expression has a specific item number.

[0019] The feedback text is obtained and divided into feedback statements based on periods, semicolons, and line breaks. Each statement is then matched against a pre-established table of application terminology. This table records the standard field names, field aliases, application document names, and key terms of contention. Based on the matching results, the original application field names are extracted to form a set of field reference statements. For this set of field reference statements, specific entry numbers, field titles, referencing terms, and key terms of contention are identified to form reference positioning statements and key terms of contention, respectively. If the reference positioning statement contains a specific entry number, the reference positioning type is determined to be specific entry anchoring; otherwise, it is determined to be a general overall referencing.

[0020] In one implementation, feedback text is obtained from the review management module of the enterprise's declaration end. This feedback text is an automatically generated feedback document by the review system after completing the review of the declaration materials, and is stored in the document management database of the declaration system in PDF or TXT format. The feedback text includes four types of information: review rejection opinions, correction notices, declaration result explanations, and field questioning content. Among them, review rejection opinions record specific non-compliance issues, correction notices list the materials that need to be supplemented, declaration result explanations provide a conclusive description of whether the review was passed or rejected, and field questioning content raises questions about the accuracy or completeness of data in specific fields of the declaration form. After obtaining the feedback text, the paragraph order, line break positions, punctuation marks, and original entry numbers are retained, while headers, footers, document number headers, and duplicate signature text are deleted to form the feedback text to be identified. The feedback text refers to the error message information generated by the customs declaration system during the document review process. It is usually output in the form of a text file or structured message and includes the declaration document number, error field, error type, and error description. The feedback statements are divided according to the periods and line break marks in the feedback text to be identified, and each feedback statement corresponds to an independent error message. If the item number is on the previous line and the field content is on the next line, they are merged into a single feedback statement based on the continuity of the text, preserving the correspondence between the item number and the field content. If a single feedback statement contains multiple declaration fields, they are separated by semicolons but are still processed as a whole and not split into multiple feedback statements. The declaration business terminology reference table is recorded according to the declaration document name. Each record records the standard field name, field alias, declaration document name, and disputed keywords. Field aliases include abbreviations, historical names, and synonyms. Disputed keywords include inconsistency, missing, duplicate, exceeding limits, unverifiable, inconsistent vouchers, and incorrect scope. Each field alias is associated with a unique standard field name. Feedback text refers to the declaration review opinion document returned by customs or tax authorities, which includes information such as the declaration document name, field name, error description, and rectification requirements.

[0021] In one possible implementation, the feedback statement is read sentence by sentence. First, the name of the application document is identified, and then the standard name and alias of the field are compared within the corresponding record range. If the feedback statement does not record the name of the application document, adjacent statements are searched bidirectionally forward and backward from the current statement as the base point. Priority is given to using the application document name closest to the current statement to limit the matching range. If it cannot be found forward, the nearest application document name is searched backward. When the title of the feedback text is inconsistent with the application document name in the adjacent statement, the application document name in the adjacent statement takes precedence. After matching the field alias, its associated standard name is read and determined as the original application field name. When the same feedback statement matches multiple field names, the field names and their text positions are retained respectively, forming a set of field reference statements. Further, the specific entry number, field title, referencing words, and disputed keywords are identified for the set of field reference statements. Specific entry numbers consist of a consecutive number marker followed by the field content; field names are the standard field names or aliases appearing in the feedback statement; referencing terms include "above," "this item," "this column," "this table," and the corresponding entry; the key terms of contention are based on the table of application terminology. The reference and positioning statements are formed according to the original order of various texts, and the key terms of contention and the field content they modify are combined into key terms of contention. The data structure includes field standard names, field aliases, dispute type codes, problem description text, and entry numbers. An entry number is only considered the corresponding specific entry number when it is in the same feedback statement as the original application field name, or when it is in two physically adjacent lines and the middle line does not contain other application field names. Physically adjacent lines refer to two consecutive lines of text arranged vertically in the document. The middle line not containing other application field names means that all lines between the two lines do not involve field names in the application form, and punctuation marks, serial numbers, or blank lines are allowed. If multiple numbers exist for the same feedback statement, a correspondence is established based on the field name immediately following the number. "Immediately following" means that the number and field name are separated only by a space, colon, or comma. Numbers that do not form a unique correspondence are not considered specific item numbers. Feedback text refers to the audit opinion document generated by the customs audit system for the enterprise's declaration documents, including the declaration document name, field names, error descriptions, and audit conclusions. Dispute focus terms are core disputed words extracted from the audit conclusion field of the feedback text, such as inconsistency, missing, duplication, exceeding limits, unverifiable, inconsistent documents, and incorrect definitions, used to identify the main problem types in this audit. Based on the above input, check whether the referenced location statement contains a specific item number corresponding to the original declaration field name. The item number format is "Item N," "Number N," etc., clearly pointing to a specific row or item in the declaration document.If the reference is present, the reference location type is determined to be specific item anchoring, and the corresponding item number, original declaration field name, and disputed terms are recorded; if it is not present, it is determined to be a general overall reference, and the declaration document name, original declaration field name, and disputed terms are recorded, thus obtaining the reference location result of the feedback text.

[0022] Step S102: Retrieve the enterprise's declaration ledger, establish a business knowledge graph with declaration items, original vouchers, and approval processes as nodes, and connect shared original vouchers or approval processes as links, and query the knowledge graph by the original declaration field name and the reference location description to obtain the initial related entries. If it is a general reference, then list all declaration items of the declaration documents to which the original declaration field name belongs as the initial related entries.

[0023] Obtain the enterprise's declaration ledger, and establish declaration item nodes, original voucher nodes, and approval process nodes based on item number, voucher number, and approval process name. Establish connections between declaration item nodes and their corresponding original voucher nodes and approval process nodes to obtain a business knowledge graph. Based on the business knowledge graph, locate the corresponding declaration item node according to the original declaration field name, and determine the query starting point based on the item number, declaration document name, and field name in the reference positioning description. Search for related declaration item nodes along the connected original voucher nodes or approval process nodes to form a candidate related item list. For the candidate related item list, if the reference positioning type is specific item anchoring, the declaration items in the list are used as the initial related items; if it is a general reference, all declaration items of the original declaration field's corresponding declaration document are added, and duplicate items are removed to obtain the initial related items.

[0024] In one implementation, the feedback text is a return notification text generated by the approval agency when a document fails to pass approval. It includes information such as the names of the application fields that need to be modified, the reason for the error, and the required supplementary materials. The enterprise application ledger corresponding to the feedback text is retrieved. This enterprise application ledger is automatically generated and stored by the application system when an enterprise submits an application. It stores application documents according to enterprise identifier, application item, application period, and application batch, and records the item number, field standard name, original voucher number, original voucher name, approval stage name, and the processing order of that approval stage. Records belonging to the same enterprise identifier, application item, application period, and application batch as the feedback text are extracted to form a target ledger record set. The business knowledge graph uses the application ledger data as its source and adopts a graph database storage structure, containing three types of nodes and two types of relationship edges. Node types include application item nodes, original voucher nodes, and approval stage nodes. Relationship edges include reference relationships between items and vouchers and process relationships between items and approval stages. Application entry nodes are identified by a combination of the application batch identifier and the entry number. Node attributes include entry name, amount, application date, associated application document identifier, and ledger record location. Original voucher nodes are identified by the voucher number. Node attributes include voucher type, issuing unit, validity period, and storage path. Approval process nodes are identified by a combination of the application item identifier and the approval process name. Node attributes include process name, approver role, time limit requirements, and preconditions. During construction, application records are extracted from the ledger system, and the entry information, associated voucher number, and approval process configuration of each record are parsed to create corresponding type nodes sequentially. The original voucher number recorded in each application entry node is read, and a reference relationship edge is established between the entry node and the corresponding voucher node. The edge attributes record the reference count and reference date. The approval process name of the application entry node is read, and a process relationship edge is established between the entry node and the corresponding approval process node. The edge attributes record the approval status and approval time. When multiple application entries reference the same original voucher, each entry node is connected to the same original voucher node through the reference relationship edge, forming a shared original voucher relationship. When multiple application items go through the same approval process, each item node is connected to the same approval process node through a relational edge, forming a relationship of belonging to the same approval process, thus obtaining a complete business knowledge graph.

[0025] In one possible implementation, when the original voucher is a summary voucher, the summary voucher node records a list of all detailed voucher numbers it covers. The summary voucher itself is not directly connected to the application entry node. Each detailed voucher node is connected to the application entry node that references that detailed voucher, forming a two-level connection structure from summary voucher to detailed voucher and from detailed voucher to application entry. An approval process name table is pre-established, containing the approval process names used by each business system and their corresponding standard names. For example, preliminary review, first review, and reconsideration are all mapped to the standard name first-level approval, while final review and second review are mapped to the standard name second-level approval. After the system reads the approval process name from the original voucher, it queries the name table to obtain the standard name and uses the standard name as the identifier for the approval process node. Approval processes with the same standard name are merged into a single node to avoid duplicate nodes for the same approval process.

[0026] Specifically, based on the original declared field name, records with the same standard field name are searched in the target ledger record set. The corresponding entry number is read, and the corresponding declared entry node is located in the business knowledge graph. If the same standard field name appears repeatedly in the same declared document, it is processed according to the type of reference positioning description. When the reference positioning type is specific entry anchoring, the query starting point is determined based on the declared document name and specific entry number in the reference positioning description, and the target field is searched by traversing downwards from that entry. When the reference positioning type is general pointing, all entries under the declared document are traversed, and the path with the fewest nodes is selected first, sorted in ascending order by entry number, and the first field node that appears is taken as the query result.

[0027] For example, when the reference positioning type is specific item anchoring, the specific item number, declaration document name, and field name are extracted from the reference positioning description. The node range is limited according to the declaration document name, and the corresponding declaration item node is checked against the specific item number and field name. The original voucher node and approval process node connected to the node are read, and other declaration item nodes connected to the original voucher node or approval process node are read. Together with the query starting point, they are collected into candidate related items. When the same declaration item is repeatedly obtained through shared original voucher relationship and same approval process relationship, duplicate records are removed according to the item number, and the original voucher name and approval process name are recorded as the source of the connection. Candidate related items are limited to the declaration batch corresponding to the target ledger record set. Items obtained through reference positioning and related items extracted through disputed terms must all meet this batch limitation. When the reference location type is a general reference, the relevant declaration document has been determined based on the original declaration field name and all the entry numbers of the document have been read during the initial related entry determination stage. In subsequent processing, the obtained initial related entries are directly used as candidate related entries for filtering, without the need to repeat the entry search operation.

[0028] Preferably, the candidate related entries are merged according to the application batch identifier and entry number, retaining the entry name, the corresponding application document, the name of the associated original voucher, the name of the associated approval process, and the source of the connection, to obtain the initial related entries. The initial related entries refer to the simplified dataset formed after merging, where each record corresponds to a unique combination of application batch identifier and entry number, and the record content includes key information fields involved in the application process. During the merging process, for multiple candidate related entries with the same application batch identifier and entry number, the system extracts their common entry name, the corresponding application document number, the list of associated original voucher names, the list of approval process names, and the source of connection identifier, merging them into a single initial related entry record.

[0029] For example, if there are 3 records in the candidate related entries that all correspond to the application entries with batch identifier B202401 and entry number C0012, then after merging, one initial related entry will be generated. The connected original voucher name field contains all voucher names that appear in these 3 records, and the connected approval process name field contains all approval process information.

[0030] Step S103: Count the number of path nodes in the knowledge graph between the initial related entries and the original declared field names after the addition, identify the number of cross-entry related entries, evaluate the relatedness level of each initial related entry according to the preset relatedness level table, and form a list of leveled entries.

[0031] The initial related entries after being supplemented are obtained. In the business knowledge graph, starting from the original declaration field's corresponding declaration entry node, the connecting paths leading to each initial related entry node are found. The connecting path with the fewest nodes is taken as the target path. All nodes contained in the target path are counted to obtain the path node count. Based on the target path, the declaration entry nodes other than the starting point and the corresponding initial related entry node are identified. The nodes are counted after deduplication according to the entry number to obtain the cross-entry related entry count, and this count is combined with the path node count to form a related entry count list. According to the correspondence between the path node count range, the cross-entry related entry count range, the related entry level, and the target level identifier in the preset related entry level table, the related entry count list is matched to determine the related entry level of each initial related entry. The evaluation result is written into the level entry list.

[0032] In one implementation, the initial related entries after being added are read, and each initial related entry node is located in the business knowledge graph based on the application batch identifier and entry number. At the same time, the application entry node to which it belongs is determined according to the original application field name, and this node is set as the starting point of the path retrieval, and the current initial related entry node is set as the ending point of the path retrieval.

[0033] Specifically, the process begins by reading directly connected original voucher nodes and approval stage nodes from the starting point of the path retrieval, followed by reading application entry nodes connected to the original voucher nodes or approval stage nodes. For newly read nodes, the source node and current level are recorded, and the process continues to read unvisited adjacent nodes until the path retrieval endpoint is reached. When multiple connected paths exist between the starting and ending points, these paths are compared in ascending order of the number of nodes, and the path with the fewest nodes is selected as the target path. If multiple connected paths have the same number of nodes, paths passing through shared original voucher nodes are prioritized, with paths having a higher priority for shared original voucher nodes. If multiple paths also have the same number of shared original voucher nodes, the target path is determined according to the processing order of the approval stages, prioritizing the path with the earliest processing time for the first approval stage node. If the processing times for the first approval stage are the same, the processing times for the second approval stage node are compared sequentially, and so on, until a unique target path is determined.

[0034] It should be noted that the number of path nodes includes the starting point, the ending point, and the original voucher nodes, approval process nodes, and declaration item nodes passed between them. A node appearing repeatedly in the same target path is counted only once. The number of path nodes is calculated according to the following rules: when the starting point and the ending point are different nodes, the number of path nodes equals the total number of all non-repeating nodes in the path; when the starting point and the ending point are the same node, the number of path nodes is recorded as 1. The initial linked entry refers to the declaration item to which the original declaration field belongs, determined in step S104 based on the declaration batch identifier and entry number of the original declaration field, which serves as the starting entry for tracing the linked relationship. When the declaration item node to which the original declaration field belongs is the same as the initial linked entry node, the number of path nodes is recorded as 1. Further, declaration item nodes are extracted from the target path. After excluding the starting point and the ending point, the declaration batch identifier and entry number of the remaining declaration item nodes are read. Records with the same declaration batch identifier and entry number are merged, and the number of the merged declaration item nodes is used as the number of cross-item linked entries. If the same intermediate declaration item appears repeatedly through both the original voucher node and the approval node, a single declaration item record is retained, and this record records both types of related sources. Therefore, the original voucher nodes and approval nodes traversed in the path are not counted in the cross-item related item count, and the current initial related item corresponding to the endpoint is also not counted repeatedly. In the R&D expense super-deduction declaration scenario, the R&D project item to which the original declaration field belongs is connected to the personnel working hours item through the project list voucher node; this personnel working hours item is one of the initial related items. The personnel working hours item is then connected to the expense collection item through the expense approval stage. This path includes a starting point, voucher node, intermediate declaration item node, approval node, and endpoint. The intermediate declaration item node, after deduplication, is counted in the cross-item related item count. This method is used to statistically analyze the complete path from the original declaration field to each endpoint item, determining the scope and quantity of cross-item related items.

[0035] Specifically, the entry number, path node count, and cross-entry linked entry count of each initial linked entry are written into the linked entry count list. A preset linked entry level table records the path node count range, the cross-entry linked entry count range, the corresponding linked entry level, and the target level identifier. In this embodiment, the preset linked entry level table is configured as follows: A path node count of no more than 3 and a cross-entry linked entry count of 0 corresponds to a close linked entry level; a path node count of 4 to 5, or a path node count of no more than 3 but a cross-entry linked entry count of 1 to 2, corresponds to a general linked entry level; a path node count exceeding 5, or a cross-entry linked entry count exceeding 2, corresponds to a weak linked entry level. Target level identifiers are configured for both close and general linked entry levels. A path node count of no more than 3 corresponds to the case where the starting point and the initial linked entry are directly connected via the same voucher node or approval process node; 4 to 5 corresponds to the case where the entry passes through an intermediate declaration entry. This aligns with the hierarchical position in S104, and the boundary values ​​of each interval are assigned to the interval with the smaller number of nodes. During matching, the system first searches for the range of path node counts, then searches for the range of cross-entry linked entries within that range to determine a unique linkage level. The preset linkage level table includes three categories: close linkage, general linkage, and weak linkage, each corresponding to a different linkage strength. Close linkage and general linkage levels are each configured with a target level identifier to mark linkage relationships requiring special attention. The system generates initial linkage entries based on the correlation analysis of the declared fields. Each initial linkage entry includes basic information such as entry number, target path, number of path nodes, and number of cross-entry linked entries. After reading the linkage level corresponding to the linkage count list, the system filters out initial linkage entries with a close or general linkage level, retains entries with target level identifiers, and forms target linkage entry records along with the entry number, target path, number of path nodes, number of cross-entry linked entries, and linkage level.

[0036] Extract the names of other declaration items mentioned in the controversial terms and mark them one by one on the knowledge graph nodes. Analyze whether there are any shared original documents or connections between the declaration item nodes where the original declaration fields are located and the marked nodes, determine the scope of declaration items, form an initial list of related items after supplementation, and indicate the source of each item's connection.

[0037] The disputed terms are obtained, and other declaration item names (excluding the original declaration field names) are extracted according to the declaration business terminology comparison table. These names are then converted into standard field names, and duplicate names are removed to obtain a list of items to be annotated. Based on this list, the corresponding declaration item nodes are located and annotated in the business knowledge graph. Each annotated node is verified to ensure it is connected to the same original voucher node or the same approval process node as the original declaration field's corresponding declaration item node, resulting in a list of related items. For this list of related items, the initial related items are merged and deduplicated according to the declaration batch identifier and item number. The shared original voucher or the same approval process source is recorded for each added item, resulting in a revised initial list of related items.

[0038] In one implementation, the disputed terms corresponding to the feedback text are obtained. These disputed terms include the original declared field names, the disputed statement, and the names of other declared items mentioned concurrently with the disputed matter. The text is segmented according to punctuation and field terminology positions in the feedback statement, retaining the declared item names that are in the same or consecutive statements as the original declared field names. A declaration business terminology lookup table is used to match the declared item names in the disputed terms one by one. The disputed terms refer to a set of professional terms extracted from the objection description paragraph of the feedback text, including field names, business terms, and their position information in the text. The declaration business terminology lookup table is established according to S101, recording the standard field name, field alias, declaration document name, and disputed terms; when used for item location, the lookup table also records the commonly used item numbers for each field. During the matching process, when a field alias is matched, its corresponding standard field name is read, while matching results identical to the original declared field name are automatically excluded, and the remaining standard field names form a list of other declared item names. This list is merged according to the standard field name, the corresponding declaration document, and the item number. When the same declared item appears repeatedly in the disputed terms as abbreviation, historical name and full name, a field standard name record is retained, and the text position corresponding to each term is saved in the record to obtain a list of items to be labeled.

[0039] In one possible implementation, the corresponding declaration entry node is located in the business knowledge graph based on the relevant declaration document and entry number in the list of items to be annotated. If the entry number is missing, it is located using both the field standard name and the relevant declaration document. After location, a dispute focus marker is written into the corresponding declaration entry node. The marker includes the dispute type code, trigger rule number, and marker timestamp, in the format of type code connected to rule number connected to timestamp, such as 02 connected to R1205 connected to 20240315143022. This is used for subsequent review processes to filter by dispute type and sort by time.

[0040] Specifically, starting from the declaration entry node to which the original declaration field belongs, the original voucher node and approval process node connected to that node are read respectively. For each marked node, the original voucher node and approval process node connected to it are read, and the node identifiers at both ends are compared. If both are connected to the same original voucher node, the marked node is recorded as a shared original voucher related entry; if both are connected to the same approval process node, the marked node is recorded as an approval process related entry. When both types of relationships are established simultaneously, two sources of connection are stored separately in the same entry record. A marked node can only be included in the related entry list if it is directly connected to the declaration entry node to which the original declaration field belongs through an original voucher node or an approval process node. Here, direct connection means that both declaration entry nodes have an edge connection relationship with the same original voucher node or the same approval process node, without passing through other declaration entry nodes as intermediate nodes. The difference between this identification rule and the initial related item search logic lies in their application scenarios. The initial related item search needs to discover potential connections to expand the scope of analysis, allowing for the derivation of relationships through multiple layers of business nodes. In contrast, related item identification is used to determine the direct related scope of the point of contention, requiring the existence of clear common vouchers or approval basis to avoid irrelevant items being included due to excessively long paths. For example, if both item A and item B reference an invoice with voucher number FP20240315001, then they are directly related. If item A references this invoice, and item C is indirectly related to this invoice through item B, then item A and item C do not constitute the related relationship referred to here.

[0041] Preferably, the list of related entries is merged with the initial list of related entries according to the enterprise identifier, application batch identifier, and entry number. For duplicate entries, only one record is retained, and its previously marked related sources are summarized and merged with the related sources identified in this instance to form a complete set of related sources. For new entries, the standard field name and the corresponding application document are written, and the original voucher name and / or approval process name are simultaneously written as its related source identifier based on the identification results, forming the initial list of related entries after supplementation.

[0042] Step S104: Retrieve the related levels of the graded items list and their hierarchical position relative to the node where the original declaration field is located. If the disputed terms mention the item name corresponding to the general related level, upgrade it to the closely related level. Select closely related level items as target related items and determine the original voucher name and approval process name of the target related items.

[0043] The initial related entries are obtained, starting from the node of the original declaration field. If an entry shares the same original voucher node or approval stage node with another initial related entry, it is considered a level. If it is reached through at least one other declaration entry node, it is considered an alternate level, resulting in a hierarchical position list. Based on the hierarchical position list, level-one entries are rated as closely related, and alternate-level entries are rated as generally related. If the disputed terminology mentions the name of an entry corresponding to the generally related level, it is upgraded to the closely related level, resulting in a level entry list. For the level entry list, closely related level entries are selected as target related entries. The original voucher node and approval stage node that form the related relationship are read to determine the original voucher name and approval stage name.

[0044] In one implementation, initial related entries are first obtained. These initial related entries refer to other declaration field entries that are directly or indirectly related to the original declaration field to be modified, and are derived through the influence scope analysis of the original declaration field in the aforementioned steps. The business knowledge graph is constructed according to S102, including three types of nodes: declaration entry nodes, original voucher nodes, and approval process nodes, as well as two types of edges: reference relationship edges between declaration entries and original vouchers, and process relationship edges between declaration entries and approval processes. Based on the three dimensions of enterprise identifier, declaration batch identifier, and entry number, the corresponding declaration entry node is located in the business knowledge graph. Simultaneously, the original declaration field name is read from the metadata of the original declaration field. This field name was obtained and stored when the modification request was received in the aforementioned steps. Then, the declaration document type and entry number to which the field belongs are read, and the corresponding declaration entry node is determined as the starting point for hierarchical position retrieval, so that the subsequent retrieval of affected related fields can be expanded outward along the graph edge relationships.

[0045] Specifically, the nodes in the business knowledge graph are stored and connected by edges. Edge types include edges linking application entries to original vouchers and edges linking application entries to approval processes. Initial linked entries refer to other application entry nodes that are associated with the retrieval starting point through original vouchers or approval processes. The hierarchy determination uses breadth-first traversal: starting from the retrieval starting point, it reads the set of original voucher nodes V and the set of approval process nodes A directly connected to it, then reads the set of application entry nodes C1 that are jointly connected to V or A, marking the nodes in C1 as level 1. For application entry nodes not appearing in C1, it continues to expand outwards from C1 as a new starting point. For each node reached through an application entry node via an original voucher or approval process, the level number is incremented by 1, recorded as level n, and the names of the transiting application entries are recorded. The search stops when the traversal reaches level 3 or the target entry is found. The upper limit of the traversal level is set to 3 levels: associations formed between application entries after more than three transits do not constitute direct voucher or approval linkages in practice; further expansion will only introduce irrelevant entries. For initial related items that can only be reached through paths at least three levels, they are treated as weakly related and are not included in the target related items. Initial related items refer to application items traced back from the original application item through shared original voucher relationships or belonging to the same approval process. When multiple paths exist for the same initial related item, the hierarchical position is determined according to the following rules: Prioritize the path with the fewest nodes passed through the application item; if the number of nodes is the same, select the path containing shared original voucher relationships, where shared original voucher relationships refer to two or more application items referencing the same original document or contract as supporting documentation; if still indistinguishable, select the path with the smaller approval completion timestamp value, i.e., the earlier processing time. Based on the determined hierarchical position, initial related items corresponding to one level are rated as closely related, and initial related items corresponding to two levels are rated as generally related. Each item record includes the item number, the corresponding application document, the hierarchical position, the related level, and the node number sequence of the path used, forming a list of graded items.

[0046] In one possible implementation, the names of declared items are first extracted from the text describing the points of contention, which originates from customs feedback or enterprise appeal records. The extracted declared item names are then converted into standard field names using a pre-defined field mapping table. This table stores the correspondence between common names, abbreviations, and aliases of declared items and standard field names; for example, goods names, commodity names, and product names are all mapped to the standard field name GOODSNAME. After conversion, the set of normalized standard field names in the terms of contention is checked against the list of graded items item by item. The checking dimensions include the standard field name, the type of declaration document to which it belongs, and the item number. An exact match rule is used during the checking process; that is, the standard field name in the terms of contention must be completely identical to the standard field name in the list of graded items to be considered a direct reference. If an item corresponding to a general level of involvement is directly mentioned in the terms of contention, the involvement level of that item is adjusted to a close involvement level, and the level identifier is updated from 1 to 2; items not directly mentioned retain their original involvement level. For each target-related entry, the actual sequence of nodes traversed is read from the shortest path record stored previously. The nodes in the path are traversed, and their types are identified: if it is an original voucher node, the original voucher name recorded in the node is extracted; if it is an approval process node, the approval process name is extracted. The path of the same entry may contain only original voucher nodes, only approval process nodes, or both; the corresponding names are retained according to the actual node types traversed.

[0047] Step S105: The original document name, approval process name, and original declaration field name of the target related items are categorized and concatenated according to the order of processing of each original document to its corresponding declaration item and the approval process, to obtain a linkage processing work order that records the order of processing of the original declaration field name and the target related items.

[0048] The target related entries are obtained. Based on the enterprise identifier, application batch identifier, and entry number, the corresponding application documents, original voucher names, application entry sequence numbers, approval process names, and processing order are read from the enterprise application ledger. These are then associated with the original application field names to obtain the work order basic record set. According to the work order basic record set, the entries are categorized by their corresponding application documents and arranged according to the application entry sequence number. When the same application entry corresponds to multiple approval processes, they are further arranged according to the processing order of the approval processes to obtain the entry processing sequence. According to the entry processing sequence, the original application field name, target related entry name, original voucher name, approval process name, and processing order are written sequentially. Records with the same enterprise identifier, application batch identifier, entry number, and approval process name are merged to obtain the linked processing work order.

[0049] In one implementation, when retrieving the target related entry, the corresponding record in the enterprise's declaration ledger is read simultaneously. The enterprise declaration ledger is stored in the customs business database and includes fields such as enterprise identifier, declaration batch identifier, entry number, associated declaration document, declaration time, commodity code, declared price, and quantity unit. It is automatically written by the enterprise in real time through the electronic port system. The system uses the enterprise identifier, declaration batch identifier, and entry number of the target related entry as a composite primary key to directly match the unique record in the enterprise's declaration ledger, establishing a one-to-one mapping relationship between the target related entry and the original declaration data. After a successful match, the ledger record is associated with the target related entry and stored. Subsequent analysis directly calls the already associated ledger data without needing to query again.

[0050] Specifically, the original voucher name, the application entry number to which the voucher belongs, the approval stage name, and the processing order of the approval stages are extracted from the entry records. The application entry number to which the voucher belongs is a fixed arrangement identifier for each entry within the application document, and the processing order of the approval stages is the sequential identifier of each approval stage within the same application matter. Further, based on the original application field name extracted from the feedback text, the application document and entry number to which the field belongs are read, and the original application field name is written into the record of each target related entry. Thus, each record simultaneously records the original application field name, the target related entry name, the original voucher name, the approval stage name, and the corresponding processing order, forming a basic record set for the work order.

[0051] In one possible implementation, the work order basic record set is categorized according to the corresponding declaration documents. The enterprise declaration ledger refers to a data table in the customs system that records the voucher information corresponding to each declaration item of an enterprise, including fields such as declaration item number, voucher number, voucher type, and voucher submission time. When multiple original vouchers correspond to the same target related item, if the declaration item numbers of each voucher are different, each original voucher record is retained separately and arranged in ascending order of the declaration item number. When multiple vouchers correspond to the same declaration item number, they are first arranged in ascending order of the voucher submission time recorded in the enterprise declaration ledger, and then arranged in ascending order of voucher number, retaining all voucher records without merging. Only when there are duplicate records with identical voucher numbers under the same declaration item number is a deduplication and merging operation performed, retaining the record with the earliest submission time.

[0052] Specifically, when the same application item goes through multiple approval stages, after the application item sequence number is completed, the records of each approval stage are arranged according to the processing order of the approval stages. Approval stages with the same processing order are arranged according to the entry time of the stage, with the stage entering earlier listed first, resulting in the item processing sequence. The stage entry time field is read from the enterprise application ledger. The enterprise application ledger is a data table that records the flow of application items through each approval stage, including fields such as item number, approval stage name, processing order, stage entry time, stage completion time, and approval status. It is automatically generated and updated in real time by the approval system when an item enters each stage.

[0053] For example, in the application for additional deduction of R&D expenses, the original field name was "Amount of Additional Deduction for R&D Expenses," and the target related items included the annual list of R&D projects and statistics on R&D personnel working hours. The two items are arranged according to their serial numbers in the application documents, and then each step is arranged according to the processing order of material verification, amount review, and approval confirmation, ensuring that the vouchers for the same item correspond to the approval records.

[0054] Preferably, the original declaration field name, the target related entry name, the corresponding declaration document, the original voucher name, the declaration entry number to which the voucher belongs, the approval stage name, and the processing order are written sequentially according to the processing sequence of the items. Records with the same enterprise identifier, declaration batch identifier, entry number, and approval stage name are merged, while different original voucher names are retained separately, resulting in a linked processing work order that records the processing order of the original declaration field name and the target related entry.

[0055] Step S106: Match the approval process name recorded in the linked processing work order with the preset assignment rule library to determine the target issue assignment node, extract the historical processing records of the target issue assignment node, screen the processing files of disputes with similar terms to the dispute focus, and generate a supplementary material list according to the list of vouchers that were previously required to be submitted in the file.

[0056] The approval process name and application item in the linked processing work order are obtained and matched item by item with the approval process name, application item, and acceptance node recorded in the preset assignment rule base to determine the target issue assignment node and form a node matching record. Based on the node matching record, the historical processing records of the target issue assignment node are extracted, and processing files that simultaneously record the same standard name and the same dispute focus keyword are filtered to obtain a set of similar dispute files. For the set of similar dispute files, the names of the vouchers that were previously required to be submitted and the corresponding application items are extracted. Records with the same voucher name and corresponding application item are merged to obtain a list of supplementary materials.

[0057] In one implementation, the approval process name, application item, enterprise identifier, and application batch identifier recorded in the linked processing work order are read. The approval process name is derived from the actual approval process that the target related item goes through during processing, and the application item is used to limit the scope of application of the assignment rules, forming a work order record to be matched.

[0058] Specifically, the pre-defined assignment rule base records the standard name, alias, acceptance node identifier, and scope of business for each approval step according to the application item. When an approval step name matches an alias, the corresponding standard name of the approval step is read. When the same approval step name corresponds to multiple acceptance nodes, the matching scope is limited based on the application item and the application document to which the target related item belongs. Further, the approval step name and application item in the work order record to be matched are checked item by item against the standard name, application item, and scope of business for each approval step in the pre-defined assignment rule base. Acceptance nodes where all three items match are determined as candidate issue assignment nodes. When multiple candidate issue assignment nodes exist, the acceptance node that undertakes the final approval step is selected as the target issue assignment node according to the processing order of the approval steps recorded in the linked processing work order.

[0059] In one possible implementation, based on the node identifier assigned to the target issue, historical processing records generated by that node under the same declaration item are retrieved. Each historical processing record must at least include the standard field name, the key term in dispute, the processed case file number, the name of the supplementary voucher, the declaration item to which the voucher belongs, and the processing conclusion. Records that do not record the processed case file number or lack disputed content are removed to obtain the set of historical records to be filtered.

[0060] Specifically, the standard field names and key dispute terms are retrieved from the current dispute focus terms and compared with each record in the historical data set to be screened. Case files with consistent standard field names and key dispute terms are identified as similar dispute files. The key dispute terms use the standard dispute expressions from the declaration business terminology comparison table. When the historical records contain synonyms, they are first converted to the corresponding standard key dispute terms before verification. When the current dispute focus terms simultaneously record discrepancies between the R&D expense deduction amount and the voucher, cases with the standard field name of R&D expense deduction amount and the key dispute term of voucher discrepancy are selected from the historical processing records. The dispute focus terms consist of two parts: the standard field name and the key dispute terms, obtained through keyword extraction and semantic analysis of taxpayer feedback text. The standard field name refers to the relevant declaration form fields, such as R&D expense deduction amount and taxable income, while the key dispute terms refer to the specific dispute type, such as voucher discrepancy, calculation error, and overdue declaration. The selected set of similar dispute files is used to generate processing suggestion templates and risk warning content. Case files that only record abnormal R&D expense amounts are not included in the same category of disputed case files because the key terms in dispute, after standardization, are incorrect in scope rather than inconsistent in documentation.

[0061] Preferably, the names of the documents and their corresponding declaration items that were previously requested to be submitted in similar disputed case files are read and merged according to the combination of document name and corresponding declaration item. Records with the same document name but different declaration items are retained separately; records with the same document name and corresponding declaration item are merged into one item, and the corresponding case file number is recorded to obtain a list of supplementary materials.

[0062] Step S107: Identify the number of material items in the supplementary material list, determine the estimated processing cycle, and send the supplementary material list and the estimated cycle to the target issue assignment node, and simultaneously provide feedback to the enterprise's application end.

[0063] Obtain the supplementary materials list, remove duplicate material records according to voucher name and corresponding application item, count the number of material items, and match the processing time limit corresponding to the interval of the number of material items in the preset processing cycle table to determine the estimated processing cycle result. Write the supplementary materials list and the estimated processing cycle result into the linkage processing work order, issue it to the target issue assignment node according to the target issue assignment node identifier, and simultaneously feed back to the enterprise application end according to the enterprise identifier and application batch identifier.

[0064] In one implementation, a supplementary materials list is obtained, and the voucher name, the corresponding application item, the enterprise identifier, and the application batch identifier are read from each material record. The combination of the voucher name and the corresponding application item is used as the identification basis for the material record. Records with the same combination are merged, while records with different combinations are retained, forming a deduplicated materials list.

[0065] Specifically, the number of records included in the deduplicated material list is counted to obtain the number of material items. A preset processing cycle table records the range of material item counts and the corresponding processing time limit for each range, with the processing time limit recorded in working days. In this embodiment, the preset processing cycle table is divided into three ranges based on the number of material items: 1 to 3 items correspond to 3 working days, 4 to 6 items correspond to 5 working days, and 7 items and above correspond to 8 working days. The boundary values ​​of the ranges are assigned to the ranges with smaller counts. The processing time limit is determined based on the actual time taken from the issuance to the completion of supplementary materials in the historical case files of the target issue assignment node. The upper quartile of the case file time within each range is taken to leave a margin for material verification and result feedback. The number of material items is checked against the starting and ending counts of each range, and the processing time limit corresponding to the range to which the number of material items belongs is read to determine the estimated processing cycle result. Further, the supplementary material list, the number of material items, the estimated processing cycle result, the enterprise identifier, and the application batch identifier are written into the linkage processing work order. The supplementary materials list in the linked processing work order is arranged according to the corresponding application item. When the same application item contains multiple vouchers, they are recorded separately according to the voucher name.

[0066] Preferably, based on the node identifier corresponding to the target issue, the linked processing work order is sent to the issue acceptance node; simultaneously, based on the enterprise identifier and application batch identifier, the specific application record under that batch in the enterprise's application terminal is located, and the supplementary material list and the estimated processing cycle result are written into the feedback record field of that application. The application record includes basic information such as the application content submitted by the enterprise, the application time, and the current processing status. The feedback record field stores feedback information such as the processing progress returned by the system, the required supplementary materials, and the estimated completion time, ensuring that the enterprise can view the processing dynamics of the matter in real time through the application terminal.

[0067] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the concept of this application. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for collecting and analyzing information feedback, characterized in that, The method includes: Obtain the feedback text of the enterprise's application, extract the original application field names, reference positioning statements and disputed terms, and determine whether the reference positioning type is specific item anchoring or overall general pointing; Retrieve the enterprise's declaration ledger to establish a business knowledge graph. Query the initial related entries by the original declaration field name and the reference location description. If it is a general reference, then all declaration entries are listed as initial related entries. Count the number of path nodes between the initial related entries and the original declared field names in the knowledge graph, identify the number of cross-entry related entries, assess the relatedness level according to the preset relatedness level table, and form a list of leveled entries. Retrieve the level of involvement from the list of level items. If the disputed terms mention the item name corresponding to the general level of involvement, upgrade it to the close level of involvement. Select the close level of involvement items as the target involvement items and determine the name of the original voucher and the name of the approval process. The original voucher names, approval process names, and original declaration field names of the target-related items are categorized and concatenated according to the order of processing to obtain the linkage processing work order; Match the names of the approval stages of the linked processing work orders in the preset assignment rule library, determine the target issue assignment node, extract historical processing records, screen the processing files of disputes with similar terms to the focus of the dispute, and generate a supplementary materials list. Identify the number of material items in the supplementary materials list, determine the estimated processing cycle, and send the supplementary materials list to the target issue assignment node, while simultaneously feeding back to the enterprise's application end.

2. The information feedback collection and analysis method according to claim 1, characterized in that, The feedback text obtained from the enterprise's application includes: The feedback text is obtained and divided into feedback statements according to periods, semicolons and line breaks. Each statement is matched against a pre-established table of application business terms. The terminology comparison table for declaration business records the standard names of fields, field aliases, declaration document names, and disputed keywords. The original declaration field names are extracted based on the matching results.

3. The information feedback collection and analysis method according to claim 1, characterized in that, The process of retrieving enterprise declaration ledgers to establish a business knowledge graph includes: Obtain the enterprise's declaration ledger, establish declaration item nodes, original voucher nodes, and approval process nodes based on the item number, voucher number, and approval process name, and establish connections between the declaration item nodes and their corresponding original voucher nodes and approval process nodes to obtain a business knowledge graph.

4. The information feedback collection and analysis method according to claim 1, characterized in that, The number of path nodes in the knowledge graph between the initial linked entries and the original declared field names includes: Obtain the initial related entries. In the business knowledge graph, starting from the node of the original declaration field, find the connecting path to each initial related entry node. Take the connecting path with the fewest nodes as the target path. Count all the nodes contained in the target path to obtain the number of path nodes. Based on the target path, identify the declared entry nodes other than the starting point and the corresponding initial linked entry nodes, count them by deduplication according to the entry number, obtain the number of cross-entry linked entries, and combine them with the number of path nodes to form a linked count list; The level of involvement is determined by matching the list of involvement counts against the preset involvement level table.

5. The information feedback collection and analysis method according to claim 4, characterized in that, The process of obtaining the initial related entries, in the business knowledge graph, starting from the node of the original declaration field's corresponding declaration entry, includes: Obtain the terms used to highlight the dispute, extract the names of other declaration items besides the original declaration field names according to the declaration business terminology comparison table, convert them into standard field names and remove duplicate names to obtain a list of items to be annotated; Based on the list of items to be annotated, the corresponding declaration item nodes are located and annotated in the business knowledge graph. Each annotated node is verified to see if it is connected to the same original voucher node or the same approval process node as the original declaration field, thus obtaining a list of associated items.

6. The information feedback collection and analysis method according to claim 1, characterized in that, The level of involvement in the list of retrieved level entries includes: Obtain the initial related entries, starting from the declaration entry node to which the original declaration field belongs; If it is connected to the same original voucher node or approval process node as an initial related item, it is recorded as a level. If it is reached through at least one other declaration item node, it is recorded as an alternate level, thus obtaining a list of level positions. Based on the aforementioned hierarchical location list, first-level items are rated as closely related, and items in adjacent levels are rated as generally related.

7. The information feedback collection and analysis method according to claim 1, characterized in that, The original document names, approval process names, and original declaration field names of the target-related items are categorized and concatenated according to the order of processing to obtain a linked processing work order, including: Obtain the target related entries, and based on the enterprise identifier, application batch identifier and entry number, read the corresponding application documents, original voucher names, application entry sequence numbers to which the vouchers belong, approval process names and processing order from the enterprise application ledger, and associate them with the original application field names to obtain the work order basic record set; Based on the aforementioned work order basic record set, the documents are categorized according to their respective application documents and arranged according to the application item number to which the document belongs; When the same application item corresponds to multiple approval stages, they are arranged according to the processing order of the approval stages to obtain the item processing sequence; According to the order of processing items, write in the original declaration field name, target related item name, original voucher name, approval process name and processing order in sequence.

8. The information feedback collection and analysis method according to claim 1, characterized in that, The matching of the work order approval process name according to the linkage process in the preset assignment rule library includes: Obtain the approval process name and application item in the linked processing work order, match them item by item with the approval process name, application item and acceptance node recorded in the preset assignment rule library, determine the target problem assignment node, and form a node matching record; Based on the node matching records, extract the historical processing records of the target issue assignment node, filter the processing files that simultaneously record the same field standard name and the same dispute focus words, and obtain a set of dispute files of the same type.

9. The information feedback collection and analysis method according to claim 1, characterized in that, The number of material items in the supplementary materials list includes: Obtain the supplementary materials list, remove duplicate material records according to the voucher name and the application item to which they belong, count the number of material items, and match the processing time limit corresponding to the interval to which the number of material items belongs in the preset processing cycle table to determine the estimated processing cycle result; The supplementary materials list and the estimated processing cycle are written into the linkage processing work order, distributed to the target problem assignment node according to the target problem assignment node identifier, and simultaneously fed back to the enterprise application end according to the enterprise identifier and the application batch identifier.

10. The information feedback collection and analysis method according to claim 3, characterized in that, The obtained business knowledge graph includes: Based on the business knowledge graph, locate the corresponding declaration item node according to the original declaration field name, determine the query starting point according to the item number, declaration document name and field name in the reference location description, and search for related declaration item nodes along the common original voucher nodes or approval process nodes to form a candidate related item list. For the candidate list of related items, if the reference positioning type is specific item anchoring, then the declared item in the list is used as the initial related item. If it is a general reference, then all the declaration entries of the declaration document to which the original declaration field belongs are added, and after removing duplicate entries, the initial related entries are obtained.

Images