College financial process automatic processing method based on RPA and natural language processing

Abstract

The application belongs to the technical field of financial automation processing, and provides a college financial process automatic processing method based on RPA and natural language processing, which comprises the following steps: collecting a reimbursement reason text and an invoice list and extracting invoice content information; extracting an expected destination set and an expected time interval from the reason text through natural language processing; taking the union of the expected destination and the city to which the applicant belongs as a geographic fence, determining an invoice geographic fence matching flag value through an indication function, and realizing preliminary screening of the invoice; then, performing secondary screening on the preliminary valid invoice according to the expected time interval, and generating a valid invoice set; sorting the valid invoice according to the event development logic according to the invoice type, constructing an ordered event space-time chain, quantifying the space-time chain rationality score and the space displacement feasibility score, and then obtaining the reimbursement request rationality score; performing compliance determination according to the rationality score and the number of valid invoices and completing process processing; and the application realizes automatic auditing of the space-time logic of the reimbursement invoice sequence.

Description

Technical Field

[0001] This invention belongs to the field of financial automation processing technology, and in particular relates to a method for automatically processing financial processes in universities based on RPA and natural language processing. Background Technology

[0002] Financial reimbursement review is a crucial aspect of financial management in universities. With increasingly frequent research and academic exchange activities, the number of reimbursement documents has surged, making it difficult for traditional manual review methods to balance efficiency and compliance. Therefore, it is necessary to introduce automation technology to improve review capabilities.

[0003] Currently, the financial reimbursement review in universities mainly relies on financial staff to verify the authenticity of invoices one by one and compare the reimbursement standards with the invoice dates. Some universities have introduced RPA to realize the image input of invoices and the verification of amounts, or used OCR to extract the structured information of invoices and match it with the fields of the financial system.

[0004] A few systems set simple compliance conditions based on rule engines, but the audit logic is limited to field-level comparison. The rule configuration is rigid and cannot adapt to the complex logic of correlation verification between multiple invoices. This results in the invoice geolocation verification relying on manual judgment, insufficient mining of semantic information of the reason text, and the lack of temporal and spatial coherence verification based on event logic between multiple invoices, making it difficult to effectively intercept false reimbursements. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an automated financial process processing method for universities based on RPA and natural language processing, thus solving the aforementioned problems.

[0006] To achieve the above objectives, this invention provides the following technical solution: an automated financial process processing method for universities based on RPA and natural language processing, which specifically includes:

[0007] Collect the text of the reason for the reimbursement request and the list of invoices in the financial process of universities, and extract the invoice content information; the invoice content information includes the invoice location, invoice time and invoice type;

[0008] Based on the text of the reason for the expense request, determine the expected set of destinations and the expected time range;

[0009] Based on the expected destination set of the reimbursement request, the geofence matching flag value of the invoice is determined according to the invoice content information.

[0010] The set of valid invoices is determined based on the geofence matching flag value and the expected time range.

[0011] Based on the set of valid invoices, the reasonableness score of the reimbursement request is determined according to the content information of the valid invoices;

[0012] The university's financial processes are processed based on the reasonableness score of reimbursement requests and the set of valid invoices.

[0013] Based on the above technical solutions, the present invention also provides the following optional technical solutions:

[0014] Further technical solution: The determination of the invoice geofence matching flag value is expressed by the formula as follows: ;

[0015] In the above formula, This refers to the geofence matching flag value for the l-th invoice. This refers to the standard city name text mapped from the original text of the invoice location for the l-th invoice. This refers to the expected set of destinations. This refers to the standardized name of the city where the applicant for the reimbursement request resides. For indicator functions, Refers to indicator functions The determination criteria.

[0016] Further technical solution: The step of determining the set of valid invoices based on the geofence matching flag value of the invoices specifically includes:

[0017] Invoices are initially screened based on the geofence matching flag values; the specific method for this initial screening is as follows:

[0018] when conditions At the time of establishment, the invoice geofence matching flag value If the value is 1, it is marked as a valid invoice and saved to the initial set of valid invoices;

[0019] when conditions If not, the invoice geofence matching flag value. If the value is 0, the invoice is marked as invalid.

[0020] Based on the expected time frame, a second screening process is performed on the invoices in the initial set of valid invoices. Specifically:

[0021] If the invoice's issuance time is outside the expected timeframe, the invoice will be retained; otherwise, the invoice will be removed.

[0022] Further technical solution: The determination of the reasonableness score of the reimbursement request specifically includes:

[0023] Create a corresponding event node for each invoice;

[0024] Based on the information in the valid invoices, an ordered event spatiotemporal chain is created; wherein, the ordered event spatiotemporal chain includes the occurrence time and location of each event node;

[0025] Collect the time intervals between adjacent event nodes in the ordered event spatiotemporal chain, and determine the spatiotemporal chain rationality score based on the time intervals between adjacent event nodes;

[0026] Based on the location of the event node, determine the feasibility score of spatial displacement;

[0027] The reasonableness score of the reimbursement request is determined based on the spatiotemporal chain rationality score and the spatial displacement feasibility score.

[0028] Further technical solution: The step of creating an ordered event spatiotemporal chain based on the content information of a valid invoice specifically includes:

[0029] Based on the invoice type of valid invoices and the basic logic of event type development, event nodes are sorted to determine the development sequence of event nodes, thereby creating an ordered event spatiotemporal chain.

[0030] Further technical solution: The determination of the spatiotemporal chain rationality score is expressed by the formula: ;

[0031] In the above formula, This refers to the spatiotemporal chain rationality score. This refers to the time interval between the (m+1)th event node and the mth event node in an ordered event spatiotemporal chain, where M refers to the number of valid invoices. For indicator functions, For indicator functions The criteria for determination;

[0032] Where, when M is less than or equal to 1, let The value of is 1.

[0033] Further technical solution: the indicator function The specific determination method is as follows:

[0034] like If the condition is met, it means that the time interval between the (m+1)th event node and the mth event node in the ordered event spatiotemporal chain is negative, indicating an error in the occurrence time of the event node. The value is 1;

[0035] like This condition is not true, meaning the time interval between the (m+1)th event node and the mth event node in the ordered event spatiotemporal chain is negative, and there is no error in the occurrence time of the event nodes. The value is 0.

[0036] Further technical solution: The determination of spatial displacement feasibility score specifically includes:

[0037] Based on the location of adjacent event nodes, determine whether there has been a change in location between adjacent time nodes;

[0038] If a location change occurs, determine the shortest reasonable time threshold for the transfer between the locations of adjacent event nodes based on their occurrence locations, and mark the adjacent event nodes as a location change event node group;

[0039] Through the formula: ;

[0040] In the above formula, This refers to the feasibility score for spatial displacement. This refers to the time interval between adjacent event nodes in the Lth location change event node group. This refers to the minimum reasonable time threshold for the transfer of the Lth location change event node group. For indicator functions, For indicator functions The determination criterion is that K refers to the number of location change event node groups;

[0041] Among them, when At the time of its establishment, The value of is 1, when When not valid, The value of is 0.

[0042] Further technical solution: The reasonableness score of the reimbursement request is determined based on the spatiotemporal chain reasonableness score and the spatial displacement feasibility score, expressed by the formula: ;

[0043] In the above formula, This refers to the rating of the reasonableness of the reimbursement request. This refers to the feasibility score for spatial displacement. This refers to the spatiotemporal chain rationality score. All are weighting coefficients, and .

[0044] Further technical solution: The process of processing university financial procedures based on the reasonableness score of reimbursement requests and the set of valid invoices specifically includes:

[0045] Set a reasonableness score threshold and set conditions based on the reasonableness score of the reimbursement request and the set of valid invoices;

[0046] When the reasonableness score of a reimbursement request is greater than or equal to the reasonableness score threshold, and the number of valid invoices in the valid invoice set is equal to the total number of invoices, the reimbursement request is deemed compliant, the reimbursement request is entered, and the approval is awaited.

[0047] If the reasonableness score of a reimbursement request is less than the reasonableness score threshold, or if the number of valid invoices in the valid invoice set is less than the total number of invoices, the reimbursement request is deemed suspicious and is returned.

[0048] This invention provides an automated financial process processing method for universities based on RPA and natural language processing, which has the following advantages compared with existing technologies:

[0049] This invention automatically extracts semantic information from the event text to determine the expected destination, combines invoice geofencing matching and time interval verification to filter valid invoices, and constructs an ordered event spatiotemporal chain for multi-dimensional rationality scoring. This achieves automated spatiotemporal logic review of reimbursement invoice sequences, effectively solving the problems of reliance on manual verification of invoice geographical location, insufficient mining of event text, and lack of spatiotemporal spatial verification of multiple invoices. It has the advantages of automated verification of invoice geographical location, in-depth mining of event text semantic information, and realization of spatiotemporal and spatial coherence verification among multiple invoices, thereby effectively identifying suspicious reimbursement requests and improving the efficiency and compliance of financial audits in universities. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating the automatic processing method for university financial processes based on RPA and natural language processing provided by the present invention.

[0051] Figure 2 This is a schematic diagram of the S50 process provided by the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0053] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0054] Please see Figure 1 The present invention provides an embodiment of an automated financial process processing method for universities based on RPA and natural language processing, comprising the following steps:

[0055] S10. Collect the text of the reason for the reimbursement request and the list of invoices in the university's financial process, and extract the invoice content information; among which, the invoice content information includes the invoice location, invoice time, and invoice type (departure route, accommodation, return trip, meeting, etc.).

[0056] S20. Based on the text of the reason for the expense request, determine the expected set of destinations and the expected time range;

[0057] S30. Based on the expected destination set of the reimbursement request, determine the geofence matching flag value of the invoice according to the invoice content information;

[0058] A geofence refers to a virtual geographical boundary that can be set around a real-world geographical area. When entering or leaving this geographical area, the system can trigger an action or notification. In this embodiment, the concept of a geofence is used to verify the matching relationship between the invoice's invoice location and the expected destination of the reimbursement request.

[0059] S40. Determine the set of valid invoices based on the invoice geofence matching flag value and the expected time interval;

[0060] S50. Based on the set of valid invoices, determine the reasonableness score of the reimbursement request according to the content information of the valid invoices;

[0061] S60. Process university financial procedures based on the reasonableness score of reimbursement requests and the set of valid invoices; specifically including:

[0062] Set a reasonableness score threshold and set conditions based on the reasonableness score of the reimbursement request and the set of valid invoices;

[0063] When the reasonableness score of a reimbursement request is greater than or equal to the reasonableness score threshold, and the number of valid invoices in the valid invoice set is equal to the total number of invoices, the reimbursement request is deemed compliant, the reimbursement request is entered, and the approval is awaited.

[0064] If the reasonableness score of a reimbursement request is less than the reasonableness score threshold, or if the number of valid invoices in the valid invoice set is less than the total number of invoices, the reimbursement request is deemed suspicious and is returned.

[0065] Specifically, in S10, the text of the reason for the reimbursement request and the list of invoices from the university's financial process are collected, and the invoice content information is extracted. The reason for the request can be entered by the applicant or obtained by scanning paper reimbursement documents and performing OCR recognition. The invoice list can be obtained by uploading electronic invoice files by the applicant or by scanning paper invoices and performing OCR recognition. The invoice content information includes the invoice location, invoice date, and invoice type. The invoice location can be the city name directly read from the invoice, the invoice date can be the date and time read from the invoice, and the invoice type can be classified according to the consumption content or category of the invoice, for example, it can be marked as departure route, accommodation, return trip, or meeting, etc.

[0066] In step S20, based on the reason text, the expected destination set and expected time interval for the reimbursement request are determined. The reason text typically contains information such as the purpose of the reimbursement request and the itinerary. For example, location information from the reason text can be extracted as the expected destination, and date information can be extracted as the expected time interval. The expected destination set can include locations, and the expected time interval can be a start date and an end date.

[0067] In step S30, based on the expected destination set of the reimbursement request, the geofence matching flag value of the invoice is determined according to the invoice content information. For each invoice in the list, it can be checked whether the invoice's issuing location matches any location in the expected destination set, or whether it matches the city where the reimbursement applicant resides. If they match, the geofence matching flag value of the invoice is set to a preset value, such as 1; if they do not match, it is set to another preset value, such as 0.

[0068] In step S40, the set of valid invoices is determined based on the invoice geofence matching flag value and the expected time interval. First, invoices can be filtered based on their geofence matching flag value; for example, only invoices with a geofence matching flag value of 1 are retained. Then, for the filtered invoices, the invoice issuance time of each invoice is checked to see if it falls within the expected time interval. If the issuance time is within the expected time interval, the invoice is retained; otherwise, it is removed. After these two filtering steps, the set of valid invoices is obtained.

[0069] In S50, based on the set of valid invoices, a reasonableness score for the reimbursement request is determined according to the information contained in the valid invoices. This score can be determined from multiple aspects. For example, it can check whether the invoice dates of valid invoices are in reverse order, or whether there is any unreasonable spatial displacement between the invoice locations of adjacent invoices. Based on these checks, a comprehensive reasonableness score is given; this score can be a numerical value or a rating.

[0070] In S60, the university's financial processes are handled based on the reasonableness score of the reimbursement request and the set of valid invoices. Specifically, a reasonableness score threshold can be set, such as 60 points. Simultaneously, conditions are set based on the reasonableness score and the set of valid invoices. When the reasonableness score of the reimbursement request is greater than or equal to the reasonableness score threshold, and the number of valid invoices in the set of valid invoices equals the total number of invoices, the reimbursement request is deemed compliant and entered into the financial system, awaiting approval. When the reasonableness score of the reimbursement request is less than the reasonableness score threshold, or the number of valid invoices in the set of valid invoices is less than the total number of invoices, the reimbursement request is deemed questionable and returned to the applicant for supplementary explanation or modification.

[0071] In summary, this invention automatically extracts semantic information from the event text to determine the expected destination, combines invoice geofencing matching and time interval verification to filter valid invoices, and constructs an ordered event spatiotemporal chain for multi-dimensional rationality scoring. This achieves automated spatiotemporal logic review of reimbursement invoice sequences, effectively solving the problems of reliance on manual verification of invoice geographical location, insufficient mining of event text, and lack of spatiotemporal spatial verification of multiple invoices. It has the advantages of automated verification of invoice geographical location, in-depth mining of semantic information from event text, and verification of spatiotemporal and spatial coherence among multiple invoices, thereby effectively identifying suspicious reimbursement requests and improving the efficiency and compliance of financial audits in universities.

[0072] Preferably, the present invention further proposes the method of determining the geofence matching flag value for invoices, which is expressed by the formula: ;

[0073] In the above formula, This refers to the geofence matching flag value for the l-th invoice. This refers to the standard city name text mapped from the original text of the invoice location for the l-th invoice. This refers to the expected set of destinations. This refers to the standardized name of the city where the applicant for the reimbursement request resides. For indicator functions, Refers to indicator functions The criteria for determination;

[0074] The geofence matching flag value for the l-th invoice serves as a binary indicator to quantify whether the invoice's location matches the expected geographical range of the reimbursement request.

[0075] The technical feature of mapping the original invoice location text of the l-th invoice to the standard city name text aims to standardize the original invoice location information extracted from the invoice, thereby eliminating inconsistencies in location names caused by different invoice formats, input habits, or regional differences. For example, the original text might be "Beijing," "Beijing Municipality," "Capital Beijing," or some place names within Beijing; after mapping, all would be unified as "Beijing Municipality." Implementation methods could include: constructing a standardized city database containing city names and their aliases and abbreviations, and mapping the original location text to the standard city names in the database using fuzzy matching, regular expression matching, or natural language processing techniques; or, utilizing Geographic Information System (GIS) services to parse the original location text into geographic coordinates, and then querying the corresponding standard city names.

[0076] The expected destination set, derived from the analysis of the reason text of the reimbursement request, represents a list of cities that the reimbursement applicant expects to travel to or pass through based on the reason text. This set is a core reference for assessing the geographical relevance of invoices.

[0077] The standardized name of the city where the applicant for reimbursement resides is specified. This technical feature takes into account the reasonableness of the applicant's departure point. Even if the invoice issuance location is not in the expected destination set, it should still be considered geographically reasonable if it is located in the applicant's city. This standardized name can be obtained through system presets, extraction from user profile information, or mapping based on the applicant's IP address.

[0078] Indicator functions are mathematically defined as returning 1 if a given condition is true, and 0 otherwise. In this context, it transforms the result of a logical judgment into a binary numerical value, allowing the geographic matching results to be quantified.

[0079] This refers to the condition for the indicator function, the core of which is to determine whether the standardized invoice location of the l-th invoice belongs to the union of the expected destination set and the applicant's city. In other words, the condition is met as long as the invoice location is one of the expected destinations in the reimbursement reason, or the applicant's departure city.

[0080] Specifically, after collecting the reason text and invoice list for the reimbursement request, and extracting the invoice content information (including the invoice issuance location), the system determines the expected destination set for the reimbursement request based on the reason text. Simultaneously, it obtains the standardized name of the city where the applicant resides. For each invoice, its original invoice issuance location text is mapped to the standardized city name text. Subsequently, the system determines whether the standardized invoice issuance location is included in the union of the expected destination set and the applicant's city. If it is included, it indicates that the invoice issuance location and the reimbursement reason are geographically consistent, the indicator function's condition is met, and the invoice geofence matching flag value is set to 1. Conversely, if it is not included, it indicates a geographical mismatch, the indicator function's condition is not met, and the invoice geofence matching flag value is set to 0.

[0081] This invention's formula-based determination mechanism transforms the potentially ambiguous process of "determining the geofence matching flag value for invoices" into a clear and repeatable logical operation. It ensures that the geofence matching result of each invoice is evaluated according to a unified standard, avoiding the subjectivity and inconsistency that can arise from manual judgment. By precisely matching the invoice's issuing location with the expected destination set and the applicant's city, this scheme can effectively identify invoices whose geographic scope does not match the reimbursement reason. This lays a solid foundation for subsequent screening of valid invoices and scoring the reasonableness of reimbursement requests, significantly improving the accuracy and reliability of automated financial process handling.

[0082] Preferably, the present invention further proposes determining the set of valid invoices based on invoice geofence matching flag values, specifically including:

[0083] Invoices are initially screened based on the geofence matching flag values; the specific method for this initial screening is as follows:

[0084] when conditions At the time of establishment, the invoice geofence matching flag value If the value is 1, it is marked as a valid invoice and saved to the initial set of valid invoices;

[0085] when conditions If not, the invoice geofence matching flag value. If the value is 0, the invoice is marked as invalid.

[0086] Based on the expected time frame, a second screening process is performed on the invoices in the initial set of valid invoices. Specifically:

[0087] If the invoice's issuance time is outside the expected time range, the invoice will be retained; if the invoice's issuance time is outside the expected time range, the invoice will be removed.

[0088] The process involves initial screening of invoices based on the value of the geofence matching flag. This aims to determine the geographical compliance of invoices by comparing the invoice's issuing location with the expected destination of the reimbursement request or the applicant's city. This process quickly eliminates invoices that clearly do not meet the geographical requirements, thereby reducing the complexity of subsequent processing.

[0089] Based on the expected time frame, a secondary screening of invoices in the initial set of valid invoices is performed. This aims to further determine the time compliance of the initially screened invoices based on whether their invoice issuance time falls within the expected timeframe of the reimbursement request. Its purpose is to ensure that the final valid invoices are not only geographically reasonable but also time-consistent with the reimbursement reason, avoiding invoices with inconsistent or irrelevant times. One implementation method is that after obtaining the initial set of valid invoices, the system extracts the invoice issuance time for each invoice in the set. Then, it compares this invoice issuance time with the expected time frame of the reimbursement request. If the invoice issuance time falls between the start and end times of the expected time frame (including the boundaries), the invoice is considered time-compliant and retained. Otherwise, the invoice is removed from the initial set of valid invoices.

[0090] This invention employs a two-stage screening mechanism: first, an initial screening based on geographical location, followed by a review based on time. This ensures that the final set of valid invoices not only matches the reimbursement reason spatially but also has temporal validity. This progressive screening approach allows subsequent calculations of the reasonableness score for reimbursement requests to be based on more accurate and reliable invoice data, thereby significantly improving the accuracy and efficiency of the entire automated financial process.

[0091] For preferred options, please refer to [link / reference]. Figure 2 The present invention further proposes the aforementioned scoring method for determining the reasonableness of reimbursement requests, specifically including:

[0092] S51. Create a corresponding event node for each invoice;

[0093] S52. Based on the information in the valid invoice, create an ordered event spatiotemporal chain; wherein, the ordered event spatiotemporal chain includes the occurrence time and location of each event node;

[0094] The creation of the ordered event spatiotemporal chain specifically includes:

[0095] Based on the invoice type of valid invoices and the basic logic of event type development, the event nodes are sorted to determine the development sequence of event nodes, thereby creating an ordered event spatiotemporal chain;

[0096] S53. Collect the time interval between adjacent event nodes in the ordered event spatiotemporal chain, and determine the spatiotemporal chain rationality score based on the time interval between adjacent event nodes.

[0097] S54. Determine the feasibility score of spatial displacement based on the location of the event node;

[0098] S55. Determine the reasonableness score of the reimbursement request based on the spatiotemporal chain rationality score and the spatial displacement feasibility score.

[0099] S51 aims to abstract each initially screened valid invoice into a basic unit for subsequent analysis. By encapsulating key invoice information, such as invoice location, invoice time, and invoice type, into a unified data structure or object, the system can process and analyze this discrete invoice data in a standardized manner. This lays the data foundation for subsequent event sequence construction and spatiotemporal rationality analysis. As one possible implementation, the system can generate a unique identifier for each valid invoice and store its associated invoice location, invoice time, invoice type, and other information in a structured data record, which serves as an event node.

[0100] S52 aims to arrange a series of independent event nodes (i.e., invoices) according to their inherent logic or chronological order, thereby forming a sequence that reflects the complete development trajectory of the reimbursement process. By constructing an ordered spatiotemporal chain of events, the system can integrate originally isolated invoice information into an event flow with spatiotemporal continuity, which is crucial for subsequent overall rationality analysis.

[0101] Specifically, events can be sorted in ascending order based on their occurrence time (i.e., the invoice issuance time) to form a strictly chronological sequence. Alternatively, by combining invoice type (e.g., departure, accommodation, meeting, return) and issuance time, and using pre-defined business logic rules, an event progression sequence that conforms to the actual travel or activity flow can be constructed. For example, departure could precede accommodation, followed by the meeting, and finally the return trip.

[0102] S53 aims to evaluate the logical rationality of the entire event chain in the time dimension by analyzing the time difference between two consecutive event nodes in an ordered event spatiotemporal chain. Its function is to detect whether there are logical errors in the order of event occurrence (e.g., a later event occurs earlier than a previous event) or whether the time interval is abnormal (e.g., the time interval between two closely related events is too long or too short). One implementation is to directly calculate the difference between the occurrence time of the (m+1)th event node and the occurrence time of the mth event node in the ordered event spatiotemporal chain to obtain the time interval. Then, a rationality judgment rule for the time interval can be set. For example, if the time interval is negative, it indicates an incorrect time sequence, resulting in a lower rationality score; if the time interval is too long or too short, points can also be deducted according to preset business rules.

[0103] S54 aims to assess the physical feasibility of location transfers between adjacent event nodes in an ordered spatiotemporal chain of events. Its core objective is to determine whether a reasonable displacement can be completed within a given time interval from one event location to the next. This helps identify geographically illogical relocation behaviors, such as moving from a distant city to another in a very short time. In practice, Geographic Information Systems (GIS) or map service APIs can be used to calculate the shortest travel time or distance between adjacent event locations. Then, the calculated shortest travel time is compared with the actual event time interval. If the actual time interval is less than the theoretical shortest travel time, the spatial displacement is considered infeasible, thus lowering the spatial displacement feasibility score. Alternatively, the theoretical transfer time can be estimated based on preset transportation types (e.g., airplanes, high-speed trains, cars) and average speeds, combined with the distance between the two locations, and used as a basis for judgment.

[0104] S55 aims to comprehensively consider both temporal rationality and spatial feasibility to derive a comprehensive and quantifiable overall rationality index for expense reimbursement requests. By integrating two independent assessment results, the system can more accurately reflect the authenticity and compliance of expense reimbursement requests. One implementation method is to use a weighted average approach, linearly combining the spatiotemporal chain rationality score and the spatial displacement feasibility score according to preset weight coefficients to obtain the final expense reimbursement request rationality score. For example, different importance weights can be assigned to temporal rationality or spatial feasibility based on business needs.

[0105] Specifically, this invention first abstracts and transforms each valid invoice into an independent event node. This node contains key spatiotemporal information such as the time and location of the invoice's occurrence, thus standardizing discrete invoice data into analyzable units. Subsequently, these event nodes are organized into an ordered spatiotemporal chain. This chain, by sorting the event nodes, constructs the time trajectory of the reimbursement process. The establishment of this ordered chain allows the system to grasp the continuity and logic of the entire reimbursement process from a macroscopic perspective. Based on this, the system performs a dual-dimensional rationality assessment of the ordered spatiotemporal chain. On one hand, by collecting the time intervals between adjacent event nodes, the system can determine a spatiotemporal chain rationality score. This score is mainly used to detect logical errors in the order of event occurrence, such as whether there are time reversals, ensuring that the reimbursement events are smooth and logical on the timeline. On the other hand, the system determines a spatial displacement feasibility score based on the locations of adjacent event nodes. This score aims to verify whether moving from one location to another within the time interval between adjacent events is physically feasible, thereby effectively identifying reimbursement behaviors that are geographically contradictory or impossible. Ultimately, by combining the spatiotemporal chain rationality score and the spatial displacement feasibility score, the system can obtain a comprehensive reimbursement request rationality score.

[0106] This comprehensive evaluation mechanism not only considers the validity of a single invoice, but also analyzes the internal logic and physical feasibility of the entire reimbursement event sequence in greater depth, significantly improving the accuracy and reliability of judging the reasonableness of reimbursement requests and effectively making up for the shortcomings of relying solely on invoice validity judgment.

[0107] Preferably, the present invention further proposes the determination of the spatiotemporal chain rationality score, expressed by the formula: ;

[0108] In the above formula, This refers to the spatiotemporal chain rationality score. This refers to the time interval between the (m+1)th event node and the mth event node in an ordered event spatiotemporal chain, where M refers to the number of valid invoices. For indicator functions, For indicator functions The determination condition; where, when M is less than or equal to 1, let The value of is 1;

[0109] The indicator function The specific determination method is as follows:

[0110] like If the condition is met, it means that the time interval between the (m+1)th event node and the mth event node in the ordered event spatiotemporal chain is negative, indicating an error in the occurrence time of the event node. The value is 1;

[0111] like This condition is not true, meaning the time interval between the (m+1)th event node and the mth event node in the ordered event spatiotemporal chain is negative, and there is no error in the occurrence time of the event nodes. The value is 0.

[0112] The spatiotemporal chain rationality score is an indicator used to quantitatively evaluate the rationality of the time sequence of events in an ordered spatiotemporal chain. Its core idea is that if the time sequence of events is reversed, that is, the later event occurs earlier than the earlier event, then the time sequence is considered to be unreasonable. The score usually ranges from 0 to 1, where 1 indicates that the time sequence is completely reasonable and 0 indicates that the time sequence is completely unreasonable.

[0113] The time interval between the (m+1)th event node and the mth event node in an ordered event spatiotemporal chain. This time interval can be obtained by directly calculating the difference between the timestamps of two adjacent event nodes. For example, if the timestamps of the event nodes are stored in minutes or hours, the time interval can be obtained by direct subtraction. Alternatively, if the timestamps are datetime objects, the difference can be calculated using the datetime processing functions provided by the programming language.

[0114] M refers to the number of valid invoices, that is, the total number of valid invoices used to construct an ordered event spatiotemporal chain. This number can be obtained by counting the number of invoices that are ultimately recognized as valid after initial and secondary screening.

[0115] The function of an indicator is to output a binary value (usually 0 or 1) based on whether a specific condition is met. When the condition is true, the indicator function takes the value 1; when the condition is false, the indicator function takes the value 0.

[0116] The indicator function's condition is used to determine whether the time interval between adjacent event nodes is negative. When A negative number indicates that the (m+1)th event node's time is earlier than the mth event node's time, meaning there's a logical error in the order of events. This condition can be determined through direct comparison. It is implemented using a value of 0.

[0117] When M is less than or equal to 1, let The value of is 1, which is a boundary condition handling mechanism: when the number of valid invoices M is 0 or 1, it is impossible to form at least two adjacent event nodes, so the time interval cannot be calculated. In this case, the time order of the spatiotemporal chain is assumed to be completely reasonable, thus the spatiotemporal chain reasonableness score is directly set to 1.

[0118] Specifically, this invention first creates an ordered spatiotemporal chain of events based on the content information of valid invoices. This chain includes the occurrence time and location of each event node. Based on this, the scheme calculates the time interval between adjacent event nodes in the ordered spatiotemporal chain. If a time interval is found to be negative, it indicates that the time of the (m+1)th event node is earlier than the time of the mth event node, which is logically unreasonable. An indicator function plays a role here, marking each case of reversed time as 1, thereby quantifying the irrationality of the time order. Finally, by subtracting the proportion of unreasonable time intervals from 1, a spatiotemporal chain rationality score is obtained.

[0119] This calculation method provides an objective and repeatable mechanism to evaluate the rationality of the time chain, effectively solving the ambiguity and subjectivity problems that may arise when determining the rationality score of the spatiotemporal chain. When the number of valid invoices M is less than or equal to 1, since a meaningful time interval cannot be formed, the system directly sets the spatiotemporal chain rationality score to 1, ensuring the robustness of the algorithm.

[0120] Preferably, the present invention further proposes the aforementioned determination of spatial displacement feasibility score, specifically including:

[0121] Based on the location of adjacent event nodes, determine whether there has been a change in location between adjacent time nodes;

[0122] If a location change occurs, determine the shortest reasonable time threshold for the transfer between the locations of adjacent event nodes based on their occurrence locations, and mark the adjacent event nodes as a location change event node group;

[0123] Through the formula: ;

[0124] In the above formula, This refers to the feasibility score for spatial displacement. This refers to the time interval between adjacent event nodes in the Lth location change event node group. This refers to the minimum reasonable time threshold for the transfer of the Lth location change event node group. For indicator functions, For indicator functions The determination criterion is that K refers to the number of location change event node groups;

[0125] Among them, when At the time of its establishment, The value of is 1, when When not valid, The value of is 0.

[0126] One step involves determining whether there has been a location change between adjacent event nodes based on their occurrence locations. This step aims to identify whether there is geographical movement between consecutive events in an ordered spatiotemporal chain of events. This can be achieved by comparing the standardized location information associated with adjacent event nodes. For example, the standardized city names or geographic coordinates of two event nodes can be compared to see if they are consistent; if they are inconsistent, a location change is determined. Alternatively, a Geographic Information System (GIS) can be used to calculate the straight-line distance or actual travel distance between two event locations; if this distance exceeds a preset minimum movement threshold, a location change is considered to have occurred.

[0127] If a location change occurs, a minimum reasonable time threshold for the transfer between adjacent event nodes is determined based on their respective locations. This threshold represents the shortest time required to move from one location to another under normal circumstances. This threshold can be determined in various ways. For example, online map service APIs (such as Amap, Baidu Maps, and Google Maps) can be used to query the shortest travel time between two locations using different modes of transportation (such as airplanes, high-speed trains, and cars), and the shortest one can be selected as the minimum reasonable time threshold for the transfer. Alternatively, it can be estimated based on preset traffic rules and average speeds, combined with the geographical distance between the two locations. For example, for long-distance travel across cities, an empirical minimum time based on flight or high-speed rail can be set.

[0128] The step of marking adjacent event nodes as a location change event node group aims to categorize and manage all adjacent event node pairs that have undergone location changes. This can be achieved by storing each pair of adjacent event nodes for a location change, along with their associated actual time interval and the calculated minimum reasonable transfer time threshold, in a data structure. For example, a list or array could be created where each element contains the identifier of a pair of event nodes, their actual time interval, and the calculated minimum reasonable transfer time threshold; alternatively, a separate record object could be created for each location change event node group for subsequent unified processing and scoring calculations.

[0129] The indicator function is determined as follows: if the time interval between adjacent event nodes in the Lth location change event node group is less than its corresponding minimum reasonable transfer time threshold, it indicates that the location change is temporally infeasible or highly unreasonable, and the indicator function is set to 1. Conversely, if the time interval is greater than or equal to the threshold, it indicates that the location change is temporally feasible, and the indicator function is set to 0. Finally, by summing all indicator function results and normalizing them, a spatial displacement feasibility score is obtained, which reflects the overall reasonableness of all location changes in the reimbursement request.

[0130] Specifically, the system first identifies all adjacent event node pairs involving location changes in the ordered spatiotemporal chain of events, and calculates the corresponding shortest reasonable transfer time threshold for each location change event node group. Then, by comparing the actual time interval with this threshold, it determines whether each location change is physically feasible. If the actual time is insufficient to complete a reasonable location transfer, the location change is marked as abnormal and included in the spatial displacement feasibility score. Ultimately, the spatial displacement feasibility score comprehensively reflects the reasonableness of all location changes in the reimbursement request. This score, working in conjunction with the spatiotemporal chain reasonableness score, can more comprehensively and accurately assess the overall reasonableness of the reimbursement request, thereby effectively identifying abnormal reimbursement behaviors that appear reasonable in time but contradict each other spatially.

[0131] Preferably, the present invention further proposes to determine the rationality score of the reimbursement request based on the spatiotemporal chain rationality score and the spatial displacement feasibility score, expressed by the formula: ;

[0132] In the above formula, This refers to the rating of the reasonableness of the reimbursement request. This refers to the feasibility score for spatial displacement. This refers to the spatiotemporal chain rationality score. All are weighting coefficients, and ;

[0133] Among them, the weighting coefficients are important parameters used to adjust the proportion of spatial displacement feasibility scores and spatiotemporal chain rationality scores in the final reimbursement request rationality score. These weighting coefficients can be set according to actual business needs, risk preferences, or historical data analysis results. For example, in some scenarios with high requirements for travel trajectory, they can be assigned weights. Higher values; while in scenarios sensitive to time sequence, they can be assigned... A higher value. In addition, the sum of the weighting coefficients is set to 1, which ensures that the weighted average of the two scores keeps the final reimbursement request rationality score Q within a similar numerical range to the spatial displacement feasibility score and the spatiotemporal chain rationality score, making it easier to understand and compare.

[0134] This application's solution integrates the rationality assessment results obtained from two independent dimensions, time and space, by introducing a weighted summation formula. The core of this formula lies in its ability to allow the system to flexibly adjust the contribution of the spatiotemporal chain rationality score and the spatial displacement feasibility score to the final reimbursement request rationality score Q, based on the actual business scenario and risk control focus. For example, for reimbursements involving long-distance business trips, the rationality of spatial displacement may be more critical, and in this case, the contribution can be appropriately increased. The weighting of expenses is relatively high; however, for expense reimbursements related to daily commutes or short meetings, the order of events may be more important, in which case increasing the weighting can be beneficial. The weighting and fusion mechanism ensures that the final reimbursement request rationality score not only comprehensively reflects the spatiotemporal consistency of the reimbursement behavior, but also reflects the relative importance of different evaluation dimensions in specific contexts through the adjustment of weight coefficients.

[0135] In this way, the system can generate a more refined and instructive comprehensive score, thereby providing a quantitative and configurable decision-making basis for subsequent financial process processing (such as compliance judgment, data entry or return), significantly improving the intelligence level and accuracy of reimbursement review.

[0136] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for automatically processing university financial processes based on RPA and natural language processing, characterized in that: The method specifically includes: Collect the text of the reason for the reimbursement request and the list of invoices in the financial process of universities, and extract the invoice content information; the invoice content information includes the invoice location, invoice time and invoice type; Based on the text of the reason for the expense request, determine the expected set of destinations and the expected time range; Based on the expected destination set of the reimbursement request, the geofence matching flag value of the invoice is determined according to the invoice content information. The set of valid invoices is determined based on the geofence matching flag value and the expected time range. Based on the set of valid invoices, the reasonableness score of the reimbursement request is determined according to the content information of the valid invoices; The university's financial processes are processed based on the reasonableness score of reimbursement requests and the set of valid invoices.

2. The method according to claim 1, characterized in that, The determination of the invoice geofence matching flag value is expressed by the following formula: ; In the above formula, This refers to the geofence matching flag value for the l-th invoice. This refers to the standard city name text mapped from the original text of the invoice location for the l-th invoice. This refers to the expected set of destinations. This refers to the standardized name of the city where the applicant for the reimbursement request resides. For indicator functions, Refers to indicator functions The determination criteria.

3. The method according to claim 2, characterized in that, The step of determining the set of valid invoices based on the geofence matching flag value of the invoices specifically includes: Invoices are initially screened based on the geofence matching flag values; the specific method for this initial screening is as follows: when conditions At the time of establishment, the invoice geofence matching flag value If the value is 1, it is marked as a valid invoice and saved to the initial set of valid invoices; when conditions If not, the invoice geofence matching flag value. If the value is 0, the invoice is marked as invalid. Based on the expected time frame, a second screening process is performed on the invoices in the initial set of valid invoices. Specifically: If the invoice's issuance time is outside the expected timeframe, the invoice will be retained; otherwise, the invoice will be removed.

4. The method according to claim 1, characterized in that, The determination of the reasonableness score of the reimbursement request specifically includes: Create a corresponding event node for each invoice; Based on the information in the valid invoices, an ordered event spatiotemporal chain is created; wherein, the ordered event spatiotemporal chain includes the occurrence time and location of each event node; Collect the time intervals between adjacent event nodes in the ordered event spatiotemporal chain, and determine the spatiotemporal chain rationality score based on the time intervals between adjacent event nodes; Based on the location of the event node, determine the feasibility score of spatial displacement; The rationality score of the reimbursement request is determined based on the spatiotemporal chain rationality score and the spatial displacement feasibility score.

5. The method according to claim 4, characterized in that, The step of creating an ordered event spatiotemporal chain based on the content information of valid invoices specifically includes: Based on the invoice type of valid invoices and the basic logic of event type development, event nodes are sorted to determine the development sequence of event nodes, thereby creating an ordered event spatiotemporal chain.

6. The method according to claim 5, characterized in that, The determination of the rationality score of the spatiotemporal chain is expressed by the following formula: ; In the above formula, This refers to the spatiotemporal chain rationality score. This refers to the time interval between the (m+1)th event node and the mth event node in an ordered event spatiotemporal chain, where M refers to the number of valid invoices. For indicator functions, For indicator functions The criteria for determination; Where, when M is less than or equal to 1, let The value is 1.

7. The method according to claim 6, characterized in that, The indicator function The specific determination method is as follows: like If the condition is met, it means that the time interval between the (m+1)th event node and the mth event node in the ordered event spatiotemporal chain is negative, indicating an error in the occurrence time of the event node. The value is 1; like This condition is not true, meaning the time interval between the (m+1)th event node and the mth event node in the ordered event spatiotemporal chain is negative, and there is no error in the occurrence time of the event nodes. The value is 0.

8. The method according to claim 4, characterized in that, The determination of the feasibility score for spatial displacement specifically includes: Based on the location of adjacent event nodes, determine whether there has been a change in location between adjacent time nodes; If a location change occurs, determine the shortest reasonable time threshold for the transfer between the locations of adjacent event nodes based on their occurrence locations, and mark the adjacent event nodes as a location change event node group; Through the formula: ; In the above formula, This refers to the feasibility score for spatial displacement. This refers to the time interval between adjacent event nodes in the Lth location change event node group. This refers to the minimum reasonable time threshold for the transfer of the Lth location change event node group. For indicator functions, For indicator functions The determination criterion is that K refers to the number of location change event node groups; Among them, when At the time of its establishment, The value of is 1, when When not valid, The value of is 0.

9. The method according to claim 4, characterized in that, The reimbursement request rationality score is determined based on the spatiotemporal chain rationality score and the spatial displacement feasibility score, expressed by the following formula: ; In the above formula, This refers to the rating of the reasonableness of the reimbursement request. This refers to the feasibility score for spatial displacement. This refers to the spatiotemporal chain rationality score. All are weighting coefficients, and .

10. The method according to claim 1, characterized in that, The process of handling university financial procedures based on the reasonableness score of reimbursement requests and the set of valid invoices specifically includes: Set a reasonableness score threshold and set conditions based on the reasonableness score of the reimbursement request and the set of valid invoices; When the reasonableness score of a reimbursement request is greater than or equal to the reasonableness score threshold, and the number of valid invoices in the valid invoice set is equal to the total number of invoices, the reimbursement request is deemed compliant, the reimbursement request is entered, and the approval is awaited. If the reasonableness score of a reimbursement request is less than the reasonableness score threshold, or if the number of valid invoices in the valid invoice set is less than the total number of invoices, the reimbursement request is deemed suspicious and is rejected.

Images