A security check query method and device based on natural language processing and a medium

Abstract

The application discloses a security check query method and device based on natural language processing and a medium, relates to the technical field of natural language processing, and comprises the following steps: analyzing a security check query statement, performing term constraint mapping, entity sorting and intention merging, obtaining query semantic elements, performing semantic constraint reorganization on the query semantic elements, generating semantic structure intermediate representation, mapping the semantic structure intermediate representation to semantic relations in a security check database, forming semantic corresponding relations, performing cross comparison, forming link starting relations, determining inter-table association relations in the security check database according to the link starting relations, arranging main business association links, locating limited directions in target positioning data along the main business association links, forming link limitation relations, and completing query organization to generate query result data. The application improves the stability and structural consistency of security check query semantic expression by performing expression path division processing on limited semantic relations.

Description

Technical Field

[0001] This invention relates to the field of natural language processing technology, and in particular to a security check query method, device and medium based on natural language processing. Background Technology

[0002] With the continuous development of natural language processing, semantic parsing, and intelligent query technologies for structured databases, business query methods are gradually shifting from traditional field retrieval and rule matching to semantically driven query organization. In security inspection information processing scenarios, natural language queries not only involve the semantic understanding of the query statement itself, but also the collaborative correspondence between security inspection business semantics, database semantic relationships, and the organization of query results. The application of artificial intelligence-related technologies in security inspection informatization construction is deepening, and research on security inspection query statement parsing, query semantic modeling, database semantic mapping, and query output organization has become an important development direction for intelligent security inspection information processing.

[0003] Existing technologies in security check query scenarios often focus on direct matching of natural language statements to database fields. When faced with complex semantic constraints between security check business terms, entity limitation information, and query intent, they are prone to problems such as scattered query semantics and discontinuous query organization links, resulting in a lack of consistent correspondence between query results and original query semantics. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a security check query method based on natural language processing to solve the problems of scattered query semantics and discontinuous query organization links.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a security check query method based on natural language processing, which includes, The security check query statement is parsed using a natural language processing algorithm, and term constraint mapping, entity consolidation, and intent merging are performed to obtain the query semantic elements. The query semantic elements are then reorganized with semantic constraints to generate an intermediate semantic structure representation. The security check query statement is a natural language statement that expresses the security check business query requirements. The semantic structure intermediate representation is mapped to the semantic relationship in the security inspection database to form a semantic correspondence. Cross-comparison is performed on the semantic correspondence to obtain a multi-evidence correspondence. A supporting closed chain is generated based on the multi-evidence correspondence, and conflict elimination and master data location are performed on the supporting closed chain to form the query target location data. Perform association direction placement on the target location data to form a link start relationship. Based on the link start relationship, trace the inter-table relationship in the security inspection database, arrange and form the main business association link. Locate the limited direction in the target location data along the main business association link to form a link limit relationship. Complete the query organization based on the main business association link and the link limit relationship, and generate query result data. The query results data are checked against the semantic structure intermediate representation to form a security inspection business result that is consistent with the business. The record content, result hierarchy and display order in the security inspection business result are reorganized and sorted to output the security inspection query output data.

[0007] As a preferred embodiment of the security check query method based on natural language processing described in this invention, the security check query statement is parsed by performing word segmentation, syntactic relation parsing, and semantic recognition on the security check query statement using a natural language processing algorithm. This process identifies business terms, name information, document information, flight information, time information, location information, and query meaning in the security check query statement, and organizes them into query semantic elements according to the corresponding semantic relations. The terminology constraint mapping maps business terms in security check query statements to semantic relationships in the security check database, forming terminology constraint relationships. The entity consolidation refers to the unified identification and standardized organization of name information, document information, flight information, time information, and location information in security check query statements; The intent merging refers to converging the query meanings in security check query statements into a unified query semantic direction; The query semantic elements include business term elements, entity elements, and intent elements.

[0008] As a preferred embodiment of the security check query method based on natural language processing described in this invention, the generation of the semantic structure intermediate representation specifically includes: Write the business term elements, entity elements, and intent elements in the query semantic elements into a unified semantic level to form semantic constraint relationships; Based on semantic constraints, restrictive correspondences and target orientations are established between business terminology elements, entity elements, and intent elements to form semantic associations. According to semantic relationships, business term elements, entity elements, and intent elements are assigned to corresponding semantic positions, and the mutually exclusive semantics, subordinate semantics, and supporting semantics in the semantic positions are regularized to form restricted semantic relationships. Coexisting semantics in restricted semantic relations are grouped into the same expression path, while non-coexisting semantics in restricted semantic relations are separated into parallel expression paths, forming a semantic constraint body; Arrange the main expression relation, the limiting expression relation, and the target expression relation according to the semantic constraint body to form the intermediate representation of the semantic structure.

[0009] As a preferred embodiment of the security inspection query method based on natural language processing described in this invention, the unified semantic hierarchy is a hierarchical semantic structure that categorizes business terminology elements, entity elements, and intent elements according to unified semantic rules.

[0010] As a preferred embodiment of the security check query method based on natural language processing described in this invention, wherein: the formation of query target location data specifically includes, Cross-comparison is performed on semantic correspondences to screen out mutually corroborating semantic correspondences and isolate semantic correspondences that deviate from the intended direction, thus obtaining multi-evidence correspondences; Based on the mutual support status between multiple evidence correspondences, the multiple evidence correspondences are linked together for verification. The mutually supportive multiple evidence correspondences are linked together into a support closed chain. The linking verification is based on the semantic support direction, the receiving object and the order of the multiple evidence correspondences to check whether adjacent multiple evidence correspondences can be continuously connected and point to the same query target. Based on the supporting closed chain, the semantic correspondence is conflict-rejected, and the conflict-rejected semantic correspondence is merged into the main data positioning relationship. The master data location relationship is used to perform location adjustment and location writing to form the target location data for the query.

[0011] As a preferred embodiment of the security check query method based on natural language processing described in this invention, the step of connecting mutually supporting multiple evidence correspondences into a supporting closed chain specifically involves... Based on the mutual support between the multiple evidence correspondences, the multiple evidence correspondences are linked for verification to form a linked verification result; Based on the results of the joint verification, the corresponding relationships of multiple pieces of evidence that support each other are retained and linked in sequence according to the order of succession, forming a supporting sequence. The interrupted and deviated relationships in the supporting sequence are separated and registered, and the correspondence of multiple pieces of evidence in continuous succession is retained to form a closed candidate sequence; The multiple evidence correspondences in the closed candidate sequence are connected in the order of their succession to form a supporting closed chain.

[0012] As a preferred embodiment of the security check query method based on natural language processing described in this invention, the generation of query result data specifically includes: The query target location data is incorporated into the business table association direction, field association direction, and result carrying direction of the security inspection database, and the oriented placement is performed on the incorporated query target location data to form the link start relationship; Based on the link initiation relationship, the inter-table relationship in the security inspection database is traced, and the inter-table relationship that takes over the master data location relationship is continued as the business relationship link; Based on the sequential order of acceptance in the business association link, the business association link is corrected for sequential connection, and the inter-table relationship pointed to by the semantics of continuously accepted queries is retained in the business association link to form the main business association link. Locate and query the limited pointer in the target location data along the main business association link, and configure the limited pointer to the corresponding association relationship in the main business association link to form a link limitation relationship; The query acceptance order is arranged according to the sequential acceptance order in the main business related links. The association relationship in the main business related links and the limiting relationship in the link limiting relationship are attached to the corresponding query position in the query acceptance order. The attached query positions are then fixed according to the sequential acceptance order to form query result data.

[0013] As a preferred embodiment of the security check query method based on natural language processing described in this invention, the semantic relationship in the security check database is the correspondence between business term elements, entity elements, and intent elements in the business tables, field items, enumeration items, and association relationships of the security check database. The corresponding verification process involves comparing the query result data with the query semantic pointers in the intermediate representation of the semantic structure item by item to confirm the corresponding consistency.

[0014] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the security check query method based on natural language processing as described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the security check query method based on natural language processing as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By performing expression path segmentation processing on restricted semantic relationships, a unified distinction and placement relationship and succession arrangement mechanism are established between coexisting and incompatible semantics, enabling business terminology elements, entity elements, and intent elements to form a consistent semantic constraint structure. This processing method ensures that query items, limiting information, and result pointers maintain coordinated correspondence within the same semantic link, thereby improving the stability and structural consistency of security check query semantic expression. Simultaneously, by establishing distinguishing relationships between expression paths at the semantic constraint layer, the parsing results of AI-based security check query statements possess stronger controllability and interpretability, effectively avoiding the problems of mixing, conflict, and deviation among multiple limiting information, and providing structured support for subsequent semantic mapping, master data location, and query organization, ensuring accuracy and consistency between query results and the original query intent. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a security check query method based on natural language processing.

[0019] Figure 2 A flowchart for generating intermediate representations of semantic structures.

[0020] Figure 3 A flowchart for generating data to locate the query target.

[0021] Figure 4 A flowchart for generating query result data and output data. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a security check query method based on natural language processing, including the following steps: S1. Parse the security check query statement using natural language processing algorithms, and perform term constraint mapping, entity consolidation, and intent merging to obtain the query semantic elements. Reorganize the query semantic elements with semantic constraints to generate an intermediate semantic structure representation. The security check query statement is a natural language statement that expresses the security check business query requirements.

[0026] S1.1 Parse the security check query statement and perform term constraint mapping, entity consolidation, and intent merging to obtain the query semantic elements, specifically: Parsing security check queries involves using natural language processing algorithms to segment, parse syntactic relationships, and identify semantics. This process identifies business terms, names, identification information, flight information, time information, location information, and the meaning of the query, and organizes them into semantic elements based on their corresponding semantic relationships. The security check query statement is input into a natural language processing algorithm. Based on the punctuation pauses, conjunctions, business term boundaries, name information boundaries, document information boundaries, flight information boundaries, time information boundaries, and location information boundaries in the security check query statement, the algorithm performs sentence segmentation and word segmentation to form a word segment sequence.

[0027] Within the segment sequence, business terminology is identified for segments that can independently represent security check business matters, such as blacklist verification, baggage opening records, review records, release results, video viewing, and handling records. Name information is identified for segments that continuously represent passenger names, personnel titles, and personnel identification names. Document information is identified for segments that continuously represent ID card numbers, document types, and the last four digits of the ID card. Flight information is identified for segments that continuously represent flight numbers, flight numbers, and flight-related identification content. Time information is identified for segments that continuously represent dates, times, time intervals, and relative time references. Location information is identified for segments that continuously represent airport areas, security checkpoints, check-in areas, waiting areas, and handling locations.

[0028] After completing the segment recognition, the syntactic relationships of action words, object words, qualifier words and result words in the segment sequence are parsed to determine the modification relationship, qualifying relationship, pointing relationship and succession relationship between business terms, name information, certificate information, flight information, time information, location information and query meaning.

[0029] Based on the syntactic relation parsing results, semantic recognition is performed on the meanings of queries such as viewing, checking, retrieving, and statistics. Business terms, name information, document information, flight information, time information, and location information are then merged and organized with the query meanings according to their corresponding semantic relations. Business terms are organized into business term elements, name information, document information, flight information, time information, and location information are organized into entity elements, and query meanings are organized into intent elements, ultimately forming the query semantic elements.

[0030] The training process for natural language processing algorithms is as follows: Natural language processing algorithms include a word vector encoding layer, a context semantic encoding layer, a syntactic relation annotation layer, an entity recognition layer, an intent determination layer, and a semantic processing output layer. The input content is the character sequence and word segment sequence corresponding to the security check query statement, and the output content is the business term label sequence, name information label sequence, document information label sequence, flight information label sequence, time information label sequence, location information label sequence, syntactic relation label sequence, and query meaning label.

[0031] Training samples were compiled from historical security check query statements, security check business Q&A records, and security check business retrieval instructions. The boundaries of business terms, name information, document information, flight information, time information, location information, syntactic relationship types, and query meaning types in the training samples were manually annotated to form a training sample set, a verification sample set, and a test sample set.

[0032] The training sample set is input into the word vector encoding layer, mapping characters and word segments into semantic vectors. These semantic vectors are then input into the context semantic encoding layer to form a context semantic representation containing information about the context. This context semantic representation is then fed into the syntactic relation annotation layer, entity recognition layer, and intent determination layer to obtain syntactic relation labeling results, entity labeling results, and query meaning labeling results. The semantic processing output layer then generates query semantic element output results based on these results. During the training phase, a joint annotation method is used, allowing the syntactic relation annotation layer, entity recognition layer, and intent determination layer to share the parameters of the context semantic encoding layer and to determine the semantic elements based on the syntactic relation annotation layer. The joint error function is formed by the label error, entity label error, and query meaning label error, and is used to iteratively adjust the parameters of the natural language processing algorithm. The parameters of the natural language processing algorithm include word vector dimension, number of context semantic encoding layers, hidden dimension, number of syntactic relation label types, number of entity label types, number of query meaning label types, batch size, learning rate, and training epochs. For example, the word vector dimension can be 128 dimensions, the number of context semantic encoding layers can be two layers, the hidden dimension can be 256 dimensions, the batch size can be 32, the learning rate can be 0.05%, and the training epochs are determined based on the change in joint label accuracy on the validation sample set.

[0033] During the optimization phase, a gradient update method is used to control the convergence of the joint error function. The optimizer is the Adam optimizer, and the momentum and regularization parameters are set based on the convergence stability on the validation sample set. Training ends when the syntactic relation annotation results, entity recognition results, and query meaning determination results on the validation sample set remain unchanged in consecutive training rounds, or when the average decrease of the joint error function in consecutive training rounds is lower than a preset stopping threshold. The trained natural language processing algorithm is then obtained. The trained natural language processing algorithm is used to perform word segmentation, syntactic relation parsing, and semantic recognition on security check queries, and outputs the query semantic elements. The stopping threshold is set based on the distribution of the decrease of the joint error function on the validation sample set in consecutive training rounds. For example, a value of 0.001 corresponds to the joint error function on the validation sample set entering a stable plateau range, and the syntactic relation annotation results, entity recognition results, and query meaning determination results on the test sample set no longer improve synchronously, indicating an increasing risk of overfitting.

[0034] It should be noted that terminology constraint mapping maps business terms in security check queries to semantic relationships in the security check database, forming terminology constraint relationships; entity consolidation unifies and standardizes name information, document information, flight information, time information, and location information in security check queries; intent merging converges the query meaning in security check queries into a unified query semantic direction; query semantic elements include business terminology elements, entity elements, and intent elements.

[0035] S1.2. The business terminology elements, entity elements, and intent elements in the query semantic elements are written into a unified semantic hierarchy to form semantic constraints. The unified semantic hierarchy is a hierarchical semantic structure that organizes business terminology elements, entity elements, and intent elements according to unified semantic rules. Specifically, The business terminology elements, entity elements, and intent elements in the query semantic elements are written in accordance with the semantic pointers in the same security check query statement. First, the query items are determined based on the business terminology elements. Then, the name information, document information, flight information, time information, and location information that can limit the query items are written into the corresponding semantic positions. Subsequently, the unified query semantic pointers formed by viewing, verifying, retrieving, and statistically analyzing are written into the result pointer position. During the writing process, the correspondence between business terminology elements and entity elements that can jointly limit the same query item is registered as a limiting relationship.

[0036] The correspondence between entity elements and intent elements that can both point to the same query result is registered as a pointing relationship. The correspondence between business term elements, entity elements and intent elements that are mutually exclusive, sequential, or scope-limited is registered as mutual exclusion constraint, sequential constraint and scope constraint, respectively. Business term elements, entity elements and intent elements that cannot form a clear correspondence retain their original writing position and are marked as relationships to be verified later. In this way, a correspondence structure with constraint meaning is formed between business term elements, entity elements and intent elements, and finally a semantic constraint relationship is formed.

[0037] It should be noted that mutual exclusion occurs when two or more business term elements, entity elements, or intent elements in the same security check query statement cannot be simultaneously true under the same query semantics. For example, when "blacklist verification" and "release result" point to opposite query meanings in the same result position, they constitute mutual exclusion.

[0038] The sequential connection occurs when business terminology, entity elements, and intent elements are sequentially connected in the same security check query statement according to the semantic development order of query item first, limiting information in the middle, and result pointing to last. For example, in the query "Query Zhang San's bag opening record yesterday", "bag opening record - Zhang San - yesterday - query" constitutes a sequential connection.

[0039] Scope narrowing occurs when entity elements or business term elements continuously narrow the scope of application of the originally broad query semantics. For example, "last three days", "T2 terminal", and "review record" narrow the scope of query time period, query location, and query item, respectively.

[0040] S1.3. Based on semantic constraints, establish restrictive correspondences and target-oriented relationships among business terminology elements, entity elements, and intent elements to form semantic associations. Specifically, Based on the semantic location and constraint registration already determined in the semantic constraint relationship, the correspondence between business term elements and entity elements that can jointly define the same query item is written as a restricted correspondence relationship. Among them, business term elements are used to clarify the business meaning of the query item, and name information, document information, flight information, time information, and location information are used to converge the query object, converge the query time period, converge the query location, and converge the query scope, respectively. Business term elements and entity elements that can simultaneously point to the same query item are kept in the same correspondence link, while business term elements and entity elements that cannot jointly define the same query item are registered separately and the conflict position is marked.

[0041] The correspondence between intent elements, business terminology elements, and entity elements that can jointly determine the destination of query results is written as a target pointing relationship. Among them, viewing the corresponding record presents the pointing, verifying the corresponding status confirms the pointing, retrieving the corresponding details returns the pointing, and statistically summarizing the corresponding results points the pointing. When the intent element can form a consistent query destination with the business terminology element and entity element, it is registered as a valid pointing. When the intent element and the business terminology element and entity element have inconsistent query destinations, the original semantic position is retained and it is registered as a relationship pending further verification. The restriction correspondence and the target pointing relationship are concatenated and organized according to the semantic sequence in the same security check query statement to obtain the semantic association relationship between business terminology elements, entity elements, and intent elements.

[0042] S1.4. Based on semantic relationships, classify business terminology elements, entity elements, and intent elements into their corresponding semantic positions, and then regularize the mutually exclusive, subordinate, and supporting semantics within these positions to form restricted semantic relationships. Specifically, Based on the registered restriction correspondence and target pointing relationship in the semantic association, business term elements such as blacklist verification, package opening record, review record, release result, video viewing, and disposal record are respectively assigned to the semantic position of the corresponding query item. Entity elements such as name information, document information, flight information, time information, and location information are respectively assigned to the corresponding restricted semantic position. Intent elements formed by viewing, verifying, retrieving, and statistically analyzing are assigned to the corresponding result pointing semantic position.

[0043] After the data is collected, business term elements, entity elements, and intent elements whose meanings cannot be simultaneously established within the same semantic location are registered as mutually exclusive semantics. Business term elements, entity elements, and intent elements directly corresponding to the query item are retained according to the priority direction of the semantic association. Other business term elements, entity elements, and intent elements that cannot be simultaneously established are removed from the current semantic location and registered as conflicting locations. Business term elements, entity elements, and intent elements within the same semantic location used to supplement another semantic location are registered as subordinate semantics, and subordinate semantics are arranged adjacent to the corresponding primary semantic location to ensure that name information, document information, flight information, time information, and location information are one-to-one with the corresponding query item. Business term elements, entity elements, and intent elements within the same semantic location used to prove the validity of the query item, the validity of the limitation direction, or the validity of the result are registered as supporting semantics, and supporting semantics are written after the corresponding primary semantic location to form a supporting order.

[0044] After the separation of mutually exclusive semantics, the attachment of subordinate semantics, and the continuation of supporting semantics, clear boundaries and continuation boundaries are retained between semantic positions, forming a restricted semantic relationship.

[0045] It should be noted that clearly defined boundaries are used to delineate which part of the business terminology elements, entity elements, and intent elements have a limiting effect on which query item, query object, query time period, query location, or result.

[0046] The boundary of connection is used to delineate the positional boundaries between the preceding and following semantic positions according to the semantic expansion order of the query item, limiting information, and result.

[0047] S1.5. Group coexisting semantics within restricted semantic relations into the same expression path, and separate incompatible semantics within restricted semantic relations into parallel expression paths, forming a semantic constraint body. Specifically, Business terminology elements, entity elements, and intent elements that point to the same query item, have consistent limiting directions, and consistent result destinations within restricted semantic relationships are concatenated and grouped into the same expression path according to semantic sequence. Specifically, business terminology elements such as blacklist verification, package opening records, review records, release results, video viewing, and handling records, along with name information, document information, flight information, time information, location information, and intent elements matching the query item (viewing, verification, retrieval, and statistics) that can jointly limit the same query item, are written into the same expression path, maintaining the order of query item first, limiting semantics in the middle, and result destination last. For business terminology elements, entity elements, and intent elements that cannot simultaneously meet the same query item, have opposite limiting directions, different result destinations, or correspond to different query items within restricted semantic relationships, they are no longer concatenated in the same expression path. Instead, they are split into different expression paths and arranged side-by-side, retaining their respective query items, limiting semantics, and result destinations within each expression path. This ensures that the semantics within each expression path are independent and that the semantics between different expression paths do not mix. Thus, the same expression path and parallel expression paths are collectively grouped into a semantic constraint body.

[0048] S1.6. Arrange the main expression relation, the limiting expression relation, and the target expression relation according to the semantic constraint body to form the intermediate representation of the semantic structure, specifically, Arrange the expression paths in the semantic constraint body according to the order in which the query semantics are established. Arrange the business term elements corresponding to blacklist verification, package opening record, review record, release result, video viewing, and disposal record as the main expression relationship. Then, arrange the entity elements in name information, document information, flight information, time information, and location information that are used to limit the query object, query time period, query location, and query scope after the main expression relationship to form the restriction expression relationship. Arrange the intent elements formed by viewing, verification, retrieval, and statistics after the restriction expression relationship to form the target expression relationship. Finally, gather the main expression relationship, restriction expression relationship, and target expression relationship arranged in each expression path according to the original order to form the intermediate representation of the semantic structure.

[0049] S2. Map the intermediate representation of the semantic structure to the semantic relationship in the security inspection database to form a semantic correspondence. Perform cross-comparison on the semantic correspondence to obtain a multi-evidence correspondence. Generate a supporting closed chain based on the multi-evidence correspondence. Perform conflict elimination and master data location on the supporting closed chain to form the query target location data.

[0050] S2.1, The semantic relationships in the security inspection database are the correspondences formed between business terminology elements, entity elements, and intent elements in the business tables, field items, enumeration items, and associations of the security inspection database. Specifically, The business terminology elements, entity elements, and intent elements in the semantic structure intermediate representation are mapped to the business tables, field items, enumeration items, and association relationships in the security inspection database according to the main expression relationship, restrictive expression relationship, and target expression relationship in each expression path. The business terminology elements are used to determine the business table that carries the query items and the corresponding field items and enumeration items in the table. The entity elements are used to determine the field items that limit the query object, query time period, query location, query scope, and the association relationships between field items. The intent elements are used to determine the result pointing field items corresponding to record presentation, status verification, detail retrieval, and summary statistics.

[0051] During the correspondence process, business terminology elements, entity elements, and intent elements that can form a consistent semantic endpoint in business tables, field items, enumeration items, and relationships are defined as valid correspondence items. For business terminology elements, entity elements, and intent elements that can only correspond to some field items but cannot close to the complete query semantics, the existing correspondence items are retained and marked for subsequent comparison and verification. For business terminology elements, entity elements, and intent elements with multiple corresponding endpoints, the multiple corresponding endpoints and their respective limiting directions are listed in parallel, so that each business terminology element, entity element, and intent element has a clear corresponding position in the security inspection database, ultimately forming a semantic correspondence relationship.

[0052] S2.2. Perform cross-comparison on semantic correspondences, screen out mutually corroborating semantic correspondences and isolate semantic correspondences that deviate from the intended meaning, to obtain multi-evidence correspondences, specifically as follows: The semantic correspondence is compared item by item according to the query item, query object, query time period, query location, query scope and result direction in the same expression path. When the business table, field item and enumeration item corresponding to the business term element and the field item and association relationship corresponding to the entity element can all point to the same query item, and the result direction corresponding to the intent element is consistent with the result corresponding to the business term element and the entity element, the corresponding semantic correspondence is defined as a mutually confirming semantic correspondence.

[0053] When the direction of the query item corresponding to the business term element is inconsistent with that of the entity element, the objects of the entity elements are inconsistent, the result of the intent element is inconsistent with the query item, or multiple non-coherent corresponding points appear at the same semantic location, the corresponding semantic correspondence is defined as a semantic correspondence that deviates from the intended direction and is separated from the current correspondence link for registration. Mutually corroborating semantic correspondences are merged and organized according to the same query item, so that the corresponding points of business term elements, entity elements and intent elements in the security inspection database form mutually supportive corresponding combinations, and finally obtain multi-evidence correspondence.

[0054] S2.3. Based on the mutual support among the multiple evidence correspondences, the multiple evidence correspondences are linked together for verification, and the mutually supporting multiple evidence correspondences are connected into a supporting closed chain, specifically as follows: S2.3.1. Based on the mutual support among the multiple evidence correspondences, perform joint verification on the multiple evidence correspondences to form a joint verification result, specifically as follows: By comparing the query item correspondence, the limited direction correspondence, and the result direction correspondence in the multi-evidence correspondence, the continuity, semantic consistency, and limited direction consistency of adjacent multi-evidence correspondences are checked one by one. Multi-evidence correspondences in which the query item correspondence can be connected, the limited direction correspondence can converge to the same query semantics, and the result direction correspondence can be kept consistent are registered as connectable relationships. Multi-evidence correspondences in which the query item correspondence is interrupted, the limited direction correspondence deviates, or the result direction correspondence is inconsistent are registered as non-connectable relationships. Connectable and non-connectable relationships are written into the connection status marker to form the connection verification result.

[0055] Among them, the contiguous verification is a process that verifies whether adjacent multi-evidence correspondences can be continuously connected and point to the same query target based on the semantic support direction, the receiving object and the order of the multiple evidence correspondences.

[0056] S2.3.2. Based on the joint verification results, the corresponding relationships of multiple pieces of evidence that support each other are retained and linked in sequence according to the order of their succession, forming a supporting sequence, specifically as follows: Based on the joint verification results, the multi-evidence correspondences marked as connectable are sequentially retained according to the order of their semantic reference in the security inspection database. The query item correspondence in the previous multi-evidence correspondence is sequentially linked with the limitation reference and result reference in the subsequent multi-evidence correspondence. Multi-evidence correspondences with continuous positions, consistent semantic references, and consistent limitation directions are written into the same continuation link. Multi-evidence correspondences with discontinuous positions but still maintaining the same semantic reference are retained in the joint verification results and the position mark is added. In this way, the mutually supporting multi-evidence correspondences are organized into a continuously expandable support sequence.

[0057] S2.3.3. Separate and register the interrupted and deviated relationships in the supporting sequence, and retain the correspondence of multiple pieces of evidence in continuous succession to form a closed candidate sequence. Specifically, Each segment of the adjacent multi-evidence correspondence in the supporting sequence is examined. Cases where the preceding multi-evidence correspondence cannot continue to connect with the following multi-evidence correspondence are registered as interrupted relationships. Cases where the preceding and following multi-evidence correspondences maintain a superficial connection but the correspondence of query items, the limited pointing correspondence, and the result pointing correspondence are deviated are registered as deviation relationships. Interrupted relationships and deviation relationships are separated from the supporting sequence and written into the anomaly registration position. For multi-evidence correspondences that do not have interrupted relationships or deviation relationships, they are retained in the supporting sequence according to the original preceding and following sequence. The continuously retained multi-evidence correspondences are organized into a complete connection chain to form a closed candidate sequence.

[0058] S2.3.4 Connect the multiple evidence correspondences in the closed candidate sequence in the order of succession to form a supporting closed chain. The order of succession is the arrangement of the multiple evidence correspondences under the same query semantics according to the query item correspondence, the limitation direction correspondence, and the result direction correspondence.

[0059] S2.4. Based on the supporting closed chain, perform conflict elimination on the semantic correspondence, and merge the conflict-eliminated semantic correspondence into the main data positioning relationship. Specifically, Based on the consecutively established query item correspondences, limited pointing correspondences, and result pointing correspondences in the supporting closed chain, the semantic correspondences are compared item by item. Semantic correspondences that are inconsistent with any of the query item correspondences, limited pointing correspondences, and result pointing correspondences in the supporting closed chain are registered as conflicting relationships and separated from the current correspondence link. For semantic correspondences that are consistent with all of the query item correspondences, limited pointing correspondences, and result pointing correspondences in the supporting closed chain, they are retained in the current correspondence link. The retained semantic correspondences are then merged and organized according to the same query object, the same query item, and the same result pointing to form master data positioning relationships.

[0060] S2.5. Perform location consolidation and location writing on the master data location relationship to form the query target location data, specifically, The query object correspondence, query item correspondence, limited pointing correspondence, and result pointing correspondence in the master data positioning relationship are organized according to the same query semantic pointing. The master data positioning relationships pointing to the same business table, the same field item, the same enumeration item, and the same association relationship are merged and written into the same positioning group. The limited positions corresponding to name information, document information, flight information, time information, and location information are specified to the corresponding field item and association position in each positioning group.

[0061] In cases where multiple location groups point to the same query object, the location group with more complete definition and more consistent result is retained, while the location groups with missing definition or offset result are separated and registered. Separation registration involves removing the location group with missing definition or offset result from its current location and writing it separately into the exception registration location to distinguish it from the retained location group. When writing the location data, the retained location group is written with the query object landing point, query item landing point, definition landing point, and result landing point specified in sequence, and the connection relationship between each landing point is also written together, forming query target location data that can directly indicate the query object, query item, definition location, and result destination.

[0062] S3. Perform association direction placement on the target location data to form a link start relationship. Based on the link start relationship, determine the inter-table relationship in the security inspection database, arrange to form the main business association link, locate the limited direction in the target location data along the main business association link, form the link limitation relationship, and complete the query organization based on the main business association link and the link limitation relationship to generate query result data.

[0063] S3.1. The query target location data is incorporated into the business table association direction, field association direction, and result carrying direction of the security inspection database. The incorporated query target location data is then oriented and placed to form the initial link relationship. Specifically, The query object landing point, query item landing point, limited pointing landing point, and result pointing landing point in the target location data are respectively mapped to the business table association direction, field association direction, and result carrying direction in the security inspection database. Among them, the query object landing point and query item landing point are written into the connection direction between the business tables that carry the location relationship of the master data, forming the business table association direction. The limited pointing landing point is written into the field connection direction inside and outside the business table for query object, query item, time information, location information, flight information, and document information, forming the field association direction. The result pointing landing point is written into the result landing direction corresponding to record presentation, status verification, detail retrieval, and summary statistics, forming the result carrying direction.

[0064] After the data is grouped, the business table association direction, field association direction, and result carrying direction are oriented and grouped according to the same query semantic direction. Directional relationships that can be continuously connected around the same query object, the same query item, and the same result direction are arranged into the same carrying sequence, while directional relationships that cannot be continuously connected around the same query semantic direction are separated and registered. This forms the link starting relationship that indicates the starting direction of the business table, the field connection direction, and the result carrying direction.

[0065] S3.2. Based on the link initiation relationship, trace the inter-table relationships in the security inspection database, and connect the inter-table relationships that inherit the master data location relationship into the business relationship link, specifically, Based on the business table association direction, field association direction, and result carrying direction already specified in the link start relationship, the inter-table association relationship connected to the query object landing point, query item landing point, limit pointing landing point, and result pointing landing point is determined segment by segment along the existing association fields between the business tables in the security inspection database.

[0066] The inter-table relationships that can take over the master data location relationship are sequentially connected in the connection direction of the link start relationship. The inter-table relationship between the business table corresponding to the query object and the business table corresponding to the query item is written in the first segment position. The inter-table relationship between the business table where the specified field is located and the business table corresponding to the query item is written in the middle segment position. The connection position of the result pointing to the corresponding business table and the aforementioned inter-table relationship is written in the last segment position.

[0067] When the relationship between tables is consistent with the business table association direction, field association direction, and result carrying direction in the link's initial relationship, it is retained in the current connection link. When the relationship between tables is inconsistent with the business table association direction, field association direction, and result carrying direction in the link's initial relationship, it is separated from the current connection link and registered. After continuous connection, a business association link is formed.

[0068] S3.3. Based on the sequential order of business association links, perform sequential correction on the business association links, and retain the inter-table relationships pointed to by consecutively received queries in the business association links to form the main business association links. Specifically, Based on the order of table relationships in the business association chain, the business table connection position, field connection position, and result carrying position between adjacent table relationships are checked segment by segment. The table relationships that can be continuously connected around the same query semantics are retained in the current business association chain.

[0069] Separate and register the inter-table relationships that are disconnected from business tables, offset in field connections, offset in result carrying, or shifted in query semantics from the current business relationship chain. For the retained inter-table relationships, rearrange them according to the following order: from the business table corresponding to the query object to the business table corresponding to the query item, from the business table corresponding to the query item to the business table corresponding to the limited pointer, and from the business table corresponding to the business table corresponding to the limited pointer to the business table corresponding to the result. This ensures that the inter-table relationships remain continuous under the same query semantics, forming the main business relationship chain.

[0070] S3.4. Locate and query the specified pointer in the target location data along the main business association link, and configure the specified pointer to the corresponding association relationship in the main business association link to form a link restriction relationship. Specifically, Along the table relationships and field connections in the main business linkage, the name information, document information, flight information, time information, and location information in the target data of the query are matched one by one. The name information is written into the corresponding relationship between the business table corresponding to the query object and the business table corresponding to the query item. The document information, flight information, time information, and location information are written into the corresponding relationship between the business table corresponding to the specified information and the business table corresponding to the query item.

[0071] When the specified pointer is consistent with the business table destination, field connection position, and query semantics in the main business association link, it is retained in the current configuration position. When the specified pointer is inconsistent with the business table destination, field connection position, and query semantics in the main business association link, it is separated from the current configuration position and registered. The retained specified pointer and its corresponding association relationship are written one by one to form the link restriction relationship. Among them, the specified pointer is the corresponding landing point of the name information specified pointer, document information specified pointer, flight information specified pointer, time information specified pointer, and location information specified pointer in the main business association link.

[0072] It should be noted that the name information limitation refers to the location of the name information in the main business association link corresponding to the identification field of the query object, and is used to narrow the scope of personnel query. The personnel query scope is set based on the number of matching records of the name information in the security inspection database. For example, when it corresponds to one personnel record, it is set as a unique personnel query scope, and when it corresponds to multiple records with the same or similar names, it is set as a multi-person candidate query scope. This can first distinguish between uniquely identifiable personnel and candidates that still need further screening, thereby reducing irrelevant personnel records from entering the subsequent query link.

[0073] The document information limitation refers to the location of the document information in the main business association link corresponding to the document recognition field and is used to narrow the scope of document query. The document query scope is set based on the number of corresponding records of the document information in the security inspection database. For example, when it corresponds to one document record, it is a unique document query scope. When it corresponds to multiple document records or can only correspond to some document fields, it is a multi-document candidate query scope, which is the matching result of document number, document last digit, or document type in the security inspection database.

[0074] Flight information limitation refers to the location of flight information in the main business association link corresponding to the flight identifier field, which is used to narrow the scope of flight query. The flight query scope is set based on the flight landing point of the flight information in the security check database. For example, when the flight number and flight date both correspond to a flight, it is a single flight query scope. When the flight number can correspond to multiple flight records of different dates or different time periods, it is a cross-date flight query scope. When there is only a flight number but no flight date or flight time period, which makes it impossible to directly land on a specific flight, it is a flight query scope to be supplemented. It is the joint landing point result of flight number, flight date and flight time period in the security check database.

[0075] The time information limitation refers to the location of the time field corresponding to the time information in the main business linkage, and is used to narrow down the query time period range. The query time period range is set based on the time coverage of the time information in the security inspection database. For example, when the time information corresponds to a clear start and end time, it is a closed time period query range; when the time information only corresponds to a single date or a single time interval, it is a single-point extended time period query range; when the time information only expresses relative time such as "recent", "today", "last week" and needs to be converted before being stored in the database, it is a converted time period query range. It is the completeness of the start and end time of the time information in the security inspection database, the time granularity, and the time conversion result.

[0076] Location information limitation refers to the location information corresponding to the location field position in the main business association link and is used to narrow down the query location range. The query location range is based on the spatial coverage level of the location information in the security check database. For example, when the location information directly corresponds to a single security check channel or a single processing point, it is the point query range; when the location information corresponds to a certain security check area, waiting area, or check-in area, it is the area query range; when the location information corresponds to a certain terminal or a combination of multiple adjacent areas, it is the region query range. It is the spatial hierarchy landing point and coverage boundary of the location information in the security check database.

[0077] It should also be noted that the security check database is a business data repository used to store passenger security check records, security check object identification information, flight association information, time and location information, and security check handling results.

[0078] S3.5. Arrange the query acceptance order according to the sequential acceptance order in the main business association chain. Attach the association relationships and link constraint relationships in the main business association chain to the corresponding query positions in the query acceptance order, and solidify the attached query positions according to the sequential acceptance order to form the query result data. Specifically, According to the order of table relationships in the main business linkage, the positions corresponding to the query object, the query item, the limit pointer, and the result pointer are arranged sequentially. The position corresponding to the query object is placed in the order of the first segment of the query, the position corresponding to the query item is placed in the order of the second segment of the query, the limit pointer is written between the corresponding position of the query item and the field connection position, and the result pointer is placed in the order of the last segment of the query.

[0079] The name information constraint pointer, document information constraint pointer, flight information constraint pointer, time information constraint pointer, and location information constraint pointer in the link constraint relationship are written to the query acceptance order position that is consistent with the connection position of their respective fields. When the constraint pointer is consistent with the query acceptance order position, the writing result is retained; when the constraint pointer is inconsistent with the query acceptance order position, they are separately registered, thus forming a query acceptance order that unfolds continuously according to the main business association link. Among them, the constraint pointer is the corresponding writing pointer of the name information constraint pointer, document information constraint pointer, flight information constraint pointer, time information constraint pointer, and location information constraint pointer in the query acceptance order.

[0080] The table relationships, field connections, and result carrying relationships in the main business association chain are respectively attached to the corresponding positions of the query object, query item, limited pointer, and result pointer in the query acceptance order. The name information limited pointer, document information limited pointer, flight information limited pointer, time information limited pointer, and location information limited pointer in the link limited relationship are written into their respective corresponding query positions. After the attachment is completed, the adjacent query positions in the query acceptance order are checked sequentially. Query positions with continuous table relationships, continuous field connections, continuous limited pointers, and continuous result carrying relationships are kept in the same acceptance chain. Query positions with interrupted table relationships, offset field connections, offset limited pointers, and offset result carrying relationships are separated and registered. The retained query positions are fixedly written according to the original acceptance order to form the query result data.

[0081] S4. Verify the correspondence between the query result data and the semantic structure intermediate representation to form a security inspection business result that is consistent with the business. Then, reorganize and organize the record content, result hierarchy and display order in the security inspection business result, and output the security inspection query output data.

[0082] S4.1 Correspondence verification is the process of comparing the query result data with the query semantic pointers in the intermediate semantic structure item by item to confirm the correspondence consistency. Specifically, The query results data, including the query object corresponding to the result, the query item corresponding to the result, the limited pointer corresponding to the result, and the result pointer corresponding to the result, are compared item by item with the main expression relation, the limited expression relation, and the target expression relation in the semantic structure intermediate representation. The comparison content includes whether the query object is consistent, whether the query item is consistent, whether the name information limitation is consistent, whether the document information limitation is consistent, whether the flight information limitation is consistent, whether the time information limitation is consistent, whether the location information limitation is consistent, and whether the result pointer is consistent.

[0083] When the corresponding result in the query result data is consistent with the corresponding semantic position in the intermediate representation of the semantic structure, the corresponding result is written to the consistency relation position. When the corresponding result in the query result data is inconsistent with, missing, or offset from the corresponding semantic position in the intermediate representation of the semantic structure, the corresponding result is written to the difference relation position and separated from the current consistency link.

[0084] The query results corresponding to the objects, items, limited pointers, and results pointers written to the consistent relation positions are merged according to the original order of the query semantic pointers. Query results that can completely cover the main expression relation, the limited expression relation, and the target expression relation are retained as business consistent results, while query results that cannot completely cover the main expression relation, the limited expression relation, and the target expression relation are excluded from the business consistent results, ultimately forming a business consistent security inspection business result.

[0085] S4.2. Group the results corresponding to the query object, query item, limited pointer, and result pointer in the security check business results according to the same query semantic pointer. Write the record content belonging to the same query object and the same query item into the same result group. Attach the limited results corresponding to name information, document information, flight information, time information, and location information to the corresponding record content position. Write the result pointers corresponding to record presentation, status verification, detail retrieval, and summary statistics into the result position of each result group.

[0086] Records with complete content, consistent result hierarchy, and display order consistent with the query semantics are retained in the current result group. Records with missing content, misaligned result hierarchy, or offset display order are separated and registered. The retained result groups are reorganized in the order of query object first, query item in the middle, limited result following, and result pointer last. This order is based on the sequence of the main expression relationship, limited expression relationship, and target expression relationship in the semantic structure. The results corresponding to the query object, the results corresponding to the query item, the limited result, and the result pointer are arranged in sequence to form the security check query output data.

[0087] This embodiment also provides a computer device applicable to the security check query method based on natural language processing, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the security check query method based on natural language processing as proposed in the above embodiment. The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0088] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the security check query method based on natural language processing as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0089] In summary, this invention achieves this by: performing expression path segmentation on restricted semantic relationships, establishing a unified distinction and arrangement mechanism between coexisting and incompatible semantics, thus forming a consistent semantic constraint structure for business terminology elements, entity elements, and intent elements. This approach ensures that query items, limiting information, and result pointers maintain coordination and correspondence within the same semantic link, thereby improving the stability and structural consistency of security check query semantic expression. Simultaneously, by establishing distinguishing relationships between expression paths at the semantic constraint layer, the parsing results of AI-based security check query statements possess stronger controllability and interpretability, effectively avoiding the mixing, conflict, and offset problems between multiple limiting information, and providing structured support for subsequent semantic mapping, master data location, and query organization, ensuring accuracy and consistency between query results and the original query intent.

[0090] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A security check query method based on natural language processing, characterized in that: include, The security check query statement is parsed using a natural language processing algorithm, and term constraint mapping, entity consolidation, and intent merging are performed to obtain the query semantic elements. The query semantic elements are then reorganized with semantic constraints to generate an intermediate semantic structure representation. The security check query statement is a natural language statement that expresses the security check business query requirements. The intermediate representation of the semantic structure is mapped to the semantic relationships in the security inspection database to form a semantic correspondence; Cross-comparison is performed on semantic correspondences to screen out mutually corroborating semantic correspondences and isolate semantic correspondences that deviate from the intended direction, thus obtaining multi-evidence correspondences; Based on the mutual support status between multiple evidence correspondences, the multiple evidence correspondences are linked together for verification. The mutually supportive multiple evidence correspondences are linked together into a support closed chain. The linking verification is based on the semantic support direction, the receiving object and the order of the multiple evidence correspondences to check whether adjacent multiple evidence correspondences can be continuously connected and point to the same query target. Based on the supporting closed chain, the semantic correspondence is conflict-rejected, and the conflict-rejected semantic correspondence is merged into the main data positioning relationship. Perform location consolidation and location writing on the master data location relationship to form the target location data for the query; The query target location data is incorporated into the business table association direction, field association direction, and result carrying direction of the security inspection database, and the oriented placement is performed on the incorporated query target location data to form the link start relationship; Based on the link initiation relationship, the inter-table relationship in the security inspection database is traced, and the inter-table relationship that takes over the master data location relationship is continued as the business relationship link; Based on the sequential order of acceptance in the business association link, the business association link is corrected for sequential connection, and the inter-table relationship pointed to by the semantics of continuously accepted queries is retained in the business association link to form the main business association link. Locate and query the limited pointer in the target location data along the main business association link, and configure the limited pointer to the corresponding association relationship in the main business association link to form a link limitation relationship; The query acceptance order is arranged according to the sequential acceptance order in the main business related links. The association relationship in the main business related links and the limiting relationship in the link limiting relationship are attached to the corresponding query position in the query acceptance order. The attached query positions are then fixed according to the sequential acceptance order to form query result data. The query results data are checked against the semantic structure intermediate representation to form a security inspection business result that is consistent with the business. The record content, result hierarchy and display order in the security inspection business result are reorganized and sorted to output the security inspection query output data.

2. The security check query method based on natural language processing as described in claim 1, characterized in that: The security check query statement is parsed by performing word segmentation, syntactic relation parsing, and semantic recognition on the security check query statement using natural language processing algorithms. This process identifies business terms, name information, document information, flight information, time information, location information, and query meaning in the security check query statement, and organizes them into query semantic elements according to the corresponding semantic relations. The terminology constraint mapping maps business terms in security check query statements to semantic relationships in the security check database, forming terminology constraint relationships. The entity consolidation refers to the unified identification and standardized organization of name information, document information, flight information, time information, and location information in security check query statements; The intent merging refers to converging the query meanings in security check query statements into a unified query semantic direction; The query semantic elements include business term elements, entity elements, and intent elements.

3. The security check query method based on natural language processing as described in claim 1, characterized in that: The intermediate representation of the generated semantic structure is specifically as follows: Write the business term elements, entity elements, and intent elements in the query semantic elements into a unified semantic level to form semantic constraint relationships; Based on semantic constraints, establish restrictive correspondences and target orientations between business terminology elements, entity elements, and intent elements to form semantic associations. According to semantic relationships, business term elements, entity elements, and intent elements are assigned to corresponding semantic positions, and the mutually exclusive semantics, subordinate semantics, and supporting semantics in the semantic positions are regularized to form restricted semantic relationships. Coexisting semantics in restricted semantic relations are grouped into the same expression path, while non-coexisting semantics in restricted semantic relations are separated into parallel expression paths, forming a semantic constraint body; Arrange the main expression relation, the limiting expression relation, and the target expression relation according to the semantic constraint body to form the intermediate representation of the semantic structure; Arrange the expression paths in the semantic constraint body according to the order in which the query semantics are established. Arrange the business term elements corresponding to blacklist verification, package opening record, review record, release result, video viewing, and disposal record as the main expression relationship. Then, arrange the entity elements in name information, document information, flight information, time information, and location information that are used to limit the query object, query time period, query location, and query scope after the main expression relationship to form the restriction expression relationship. Arrange the intent elements formed by viewing, verification, retrieval, and statistics after the restriction expression relationship to form the target expression relationship. Finally, gather the main expression relationship, restriction expression relationship, and target expression relationship arranged in each expression path according to the original order to form the intermediate representation of the semantic structure.

4. The security check query method based on natural language processing as described in claim 3, characterized in that: The unified semantic hierarchy is a hierarchical semantic structure that organizes business terminology elements, entity elements, and intent elements according to unified semantic rules.

5. The security check query method based on natural language processing as described in claim 1, characterized in that: The process of connecting multiple mutually supporting pieces of evidence into a supporting closed chain specifically involves... Based on the mutual support between the multiple evidence correspondences, the multiple evidence correspondences are linked for verification to form a linked verification result; Based on the results of the joint verification, the corresponding relationships of multiple pieces of evidence that support each other are retained and linked in sequence according to the order of succession, forming a supporting sequence. The interrupted and deviated relationships in the supporting sequence are separated and registered, and the correspondence of multiple pieces of evidence in continuous succession is retained to form a closed candidate sequence; The multiple evidence correspondences in the closed candidate sequence are connected in the order of their succession to form a supporting closed chain.

6. The security check query method based on natural language processing as described in claim 1, characterized in that: The semantic relationships in the security inspection database are the correspondences formed between business term elements, entity elements, and intent elements in the business tables, field items, enumeration items, and associations of the security inspection database. The corresponding verification process involves comparing the query result data with the query semantic pointers in the intermediate representation of the semantic structure item by item to confirm the corresponding consistency.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the security check query method based on natural language processing as described in any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the security check query method based on natural language processing as described in any one of claims 1 to 6.

Images