Intelligent semantic analysis method and system based on MDX framework

Abstract

The application relates to the technical field of data analysis, and particularly relates to an intelligent semantic analysis method and system based on an MDX framework. The method comprises the following steps: performing associated search in an MDX database according to a query request text of a user to obtain a plurality of key objects; searching a plurality of candidate object groups with context sharing relationships in an object space corresponding to the plurality of key objects; analyzing a content matching degree between each candidate object group and the query request text in the plurality of candidate object groups to obtain a demand matching value of each candidate object group; and determining a candidate object group with the highest demand matching value and passing a non-empty check as a target object group in the plurality of candidate object groups, and generating a demand analysis result according to the target object group. The application can improve the processing effect of natural language of a user in the MDX application field.

Description

Technical Field

[0001] This invention relates to the technical field of data analysis, specifically to an intelligent semantic parsing method and system based on the MDX framework. Background Technology

[0002] The Multidimensional Expression (MDX) query language allows users to aggregate and analyze metric data (such as sales and inventory) from different dimensions (such as time, product, and region), which can greatly facilitate users' data mining work.

[0003] Currently, existing technologies mostly use Natural Language Processing (NLP) to convert users' natural language into standard MDX query language to lower the barrier for users to use MDX language. However, in engineering applications, it has been found that natural language in MDX databases can develop multiple interpretations, and existing technologies lack effective identification of multiple interpretations. This results in users' natural language being converted into multiple possible MDX query languages. Iterating through and executing multiple MDX query languages ​​not only causes a huge waste of computing resources, but also makes users wait too long for the results to be output. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent semantic parsing method and system based on the MDX framework, which solves the technical problem of poor processing effect of users' natural language in the field of MDX application.

[0005] In a first aspect, one embodiment of the present invention provides an intelligent semantic parsing method based on the MDX framework, the method comprising:

[0006] Based on the user's query request text, a correlation search is performed in the MDX database to obtain multiple key objects. The key objects are metadata objects in the MDX database that are related to the query request text, and the metadata objects correspond to at least one metric group ID.

[0007] In the object space corresponding to the multiple key objects, search for multiple candidate object groups with context sharing relationship, wherein all key objects included in the candidate object group have at least one common metric group ID;

[0008] Among the multiple candidate object groups, the content matching degree between each candidate object group and the query request text is analyzed to obtain the demand matching value of each candidate object group;

[0009] Among the multiple candidate object groups, the candidate object group that passes the non-empty check and has the highest requirement matching value is determined as the target object group, and the requirement parsing result is generated based on the target object group.

[0010] In some embodiments, the step of performing a correlation search in the MDX database based on the user's query request text to obtain multiple key objects includes:

[0011] The query request text is segmented into words to obtain multiple entity words;

[0012] In the multiple entity segmentations, obtain multiple metadata objects associated with each entity segmentation in the MDX database;

[0013] In the multiple entity segmentations, the degree of matching between each entity segmentation and the object name of each associated metadata object is analyzed to obtain multiple name matching degrees for each entity segmentation;

[0014] Based on the name matching degree of each entity segment, all metadata objects associated with multiple entity segments are filtered to obtain multiple key objects, wherein the name matching degree of the key objects is greater than or equal to the matching degree threshold.

[0015] In some embodiments, the step of filtering all metadata objects associated with multiple entity segmentations based on multiple name matching degrees for each entity segmentation to obtain multiple key objects includes:

[0016] Based on the matching degree threshold and multiple name matching degrees of the target entity segment, multiple metadata objects associated with the target entity segment are filtered to determine at least one target metadata object associated with the target entity segment, wherein the target entity segment is any entity segment among the multiple entity segments, and the name matching degree corresponding to the target metadata object is greater than or equal to the matching degree threshold.

[0017] If the number of at least one target metadata object associated with the target entity segmentation is less than or equal to the number threshold, all at least one target metadata object associated with the target entity segmentation will be identified as the key object.

[0018] If the number of at least one target metadata object associated with the target entity segmentation is greater than the number threshold, then among the at least one target metadata object associated with the target entity segmentation, the K target metadata objects with the highest name matching degree are all determined as the key objects, where K is the number threshold.

[0019] In some embodiments, the step of searching for multiple candidate object groups with context-sharing relationships in the object space corresponding to the multiple key objects includes:

[0020] An object graph is constructed using the multiple key objects as graph nodes, wherein the connecting edges in the object graph are used to indicate that: two corresponding key objects are associated with different entity segments, and two corresponding key objects have a common metric value group ID;

[0021] Perform a maximally connected subgraph search on the object graph to obtain multiple initial subgraphs;

[0022] The intersection of the multiple initial subgraphs is verified to obtain the multiple candidate object groups.

[0023] In some embodiments, the step of analyzing the content matching degree between each candidate object group and the query request text in the plurality of candidate object groups to obtain the demand matching value of each candidate object group includes:

[0024] Analyze the name matching degree between the target candidate object group and the query request text to obtain the name matching factor of the target candidate object group, wherein the target candidate object group is any one of the multiple candidate object groups;

[0025] Analyze the entity coverage between the target candidate object group and the query request text to obtain the entity coverage factor of the target candidate object group;

[0026] The demand matching value of the target candidate object group is determined based on the name matching factor and entity coverage factor of the target candidate object group.

[0027] In some embodiments, the name matching factor of the target candidate object group is the sum of the name matching degrees of all the target candidate object groups.

[0028] In some embodiments, the step of analyzing the entity coverage between the target candidate object group and the query request text to obtain the entity coverage factor of the target candidate object group includes:

[0029] Calculate the ratio of the number of all entity segments corresponding to the target candidate object group to the number of all entity segments included in the query request text to obtain the entity coverage ratio of the target candidate object group;

[0030] The entity coverage ratio of the target candidate object group is amplified to obtain the entity coverage factor of the target candidate object group.

[0031] In some embodiments, the name matching factor and the corresponding demand matching value are positively correlated, and the entity coverage factor and the corresponding demand matching value are positively correlated.

[0032] In some embodiments, the step of determining the candidate object group with the highest requirement matching value that passes the non-empty check as the target object group among the plurality of candidate object groups includes:

[0033] Based on the demand matching value of each candidate object group, the multiple candidate object groups are sorted in descending order to obtain the object group sequence.

[0034] Starting from the first sequence element of the object group sequence, the traversal is terminated if the currently traversed sequence element passes the non-empty check, and the currently traversed sequence element is determined as the target object group.

[0035] Secondly, another embodiment of the present invention also provides an intelligent semantic parsing system based on the MDX framework, the system comprising:

[0036] The related search module is used to perform related searches in the MDX database based on the user's query request text to obtain multiple key objects. The key objects are metadata objects in the MDX database that are related to the query request text, and the metadata objects correspond to at least one metric group ID.

[0037] The candidate search module is used to search for multiple candidate object groups with context-sharing relationships in the object space corresponding to the multiple key objects, wherein all key objects included in the candidate object group have at least one common metric group ID.

[0038] The matching analysis module is used to analyze the content matching degree between each candidate object group and the query request text in the plurality of candidate object groups, and to obtain the required matching value for each candidate object group.

[0039] The result determination module is used to determine the candidate object group with the highest requirement matching value that passes the non-empty check as the target object group among the multiple candidate object groups, and generate requirement parsing results based on the target object group.

[0040] Thirdly, in another embodiment of the present invention, an electronic device is provided, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method described in the first aspect.

[0041] Fourthly, in another embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0042] The present invention has the following beneficial effects:

[0043] This invention first performs a correlation search in the MDX database based on the user's query request text to initially filter out several metadata objects that are related to the query request text. Then, among the filtered metadata objects, it searches for multiple object combinations that can be used to respond to the query request text using a common metric group ID as the search condition. It further analyzes the content matching degree between each object combination and the query request text. Finally, it provides the user with the requirement parsing result corresponding to the object combination that passes the non-empty check and has the highest content matching degree. This avoids the computational resource overhead and time overhead caused by executing multiple MDX query languages, and improves the natural language processing effect for users in the field of MDX applications. Attached Figure Description

[0044] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0045] Figure 1 This is a flowchart illustrating an intelligent semantic parsing method based on the MDX framework provided in an embodiment of the present invention.

[0046] Figure 2 This is a schematic diagram of the structure of an intelligent semantic parsing system based on the MDX framework provided in an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0048] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent semantic parsing method and system based on the MDX framework proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0049] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0050] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent semantic parsing method and system based on the MDX framework provided by this invention.

[0051] In one embodiment, the present invention provides an intelligent semantic parsing method based on the MDX framework, such as... Figure 1 As shown, the method includes:

[0052] Step S1: Perform a correlation search in the MDX database based on the user's query request text to obtain multiple key objects.

[0053] The key object is a metadata object in the MDX database that is associated with the query request text, and the metadata object corresponds to at least one metric group ID.

[0054] The query request text mentioned above can be understood as natural language text input by the user carrying their data analysis needs, such as: I want to see the sales data for 2023.

[0055] The aforementioned MDX database is a database built on the MDX framework. The database stores the schema definition (i.e., metadata object) of the target cube that can be read through the standard interface of the OLAP engine.

[0056] The aforementioned target multidimensional dataset can be understood as a collection of multidimensional data involved in a certain application scenario in the field of data analysis. For example, when the target multidimensional dataset corresponds to a supermarket scenario, the multiple data dimensions involved can be time dimension (year, quarter, month, day), product dimension (product category, product subcategory), region dimension (region, province, city) and measurement dimension (sales revenue, inventory).

[0057] Specifically, each metadata object includes at least an object name (a path string used to uniquely identify the object in the OLAP engine, whose name consists of the dimension name of the corresponding data dimension, the level name of the specific level under the corresponding data dimension (if a level exists), and the member name of the specific member under the corresponding level, for example: [TIME].[YEAR].

[2023] ) and a corresponding type identifier (divided into MEASURE and DIMENSION).

[0058] This invention further constructs a mapping relationship between members and metric groups to clarify the scope of all fact tables that each specific member can legally participate in the calculation. When the type identifier corresponding to the member is MEASURE, its corresponding metric group ID is the ID of a single metric group that includes its physical affiliation. When the type identifier corresponding to the member is DIMENSION, its corresponding metric group ID is the ID of all metric groups associated with its dimension.

[0059] For example, for a metadata object with the object name [TIME].[YEAR].

[2023] , its corresponding type identifier is DIMENSION (dimension), and the multiple measure value group IDs it corresponds to can indicate a sales fact group, a budget fact group, and an inventory fact group; while for a metadata object with the object name [measure].[turnover], its corresponding type is MEASURE (measure), and the measure value group it corresponds to only indicates a sales fact group.

[0060] Specifically, the steps of performing an associated search in the MDX database based on the user's query request text to obtain multiple key objects include:

[0061] Perform word segmentation on the query request text to obtain multiple entity word segments;

[0062] Among the multiple entity word segments, obtain multiple metadata objects associated with each entity word segment in the MDX database;

[0063] Among the multiple entity word segments, analyze the matching degree between each entity word segment and the object name of each metadata object associated with it to obtain multiple name matching degrees for each entity word segment;

[0064] Based on the multiple name matching degrees of each entity word segment, screen all the metadata objects associated with the multiple entity word segments to obtain multiple key objects, where the name matching degree corresponding to the key object is greater than or equal to the matching degree threshold.

[0065] In the present invention, a natural language processing word segmenter (such as Jieba tool, Stanford NLP tool, and TextBlob tool, etc.) is used to perform word segmentation on the query request text, and during the word segmentation process, a stop word filtering operation is performed to remove auxiliary words and function words without actual business meanings such as "of", "is", and "I want to see", etc.

[0066] By traversing the multiple entity word segments to obtain multiple metadata objects associated with each entity word segment in the MDX database, where in the aforementioned MDX database, the currently traversed entity word segment can be used as a retrieval keyword for fuzzy matching retrieval to obtain multiple metadata objects associated with the currently traversed entity word segment.

[0067] It should be noted that due to the existence of cases of identical members or similar names, an entity word segment may match multiple metadata objects in the MDX database. Therefore, the multiple metadata objects associated with each entity word segment obtained here are only used as its potential interpretation set and need to be further screened and processed later before they can be used.

[0068] The aforementioned name matching degree is used to indicate the degree of matching between the object name of the corresponding metadata object and the corresponding entity segment. In this invention, normalized edit distance similarity is used as the name matching degree between the two. The larger the value, the higher the degree of matching between the two, which means that the corresponding metadata object is more likely to indicate the user's true intent.

[0069] Furthermore, based on the name matching degrees of each entity segment, the steps of filtering all metadata objects associated with multiple entity segmentations to obtain multiple key objects include:

[0070] Based on the matching degree threshold and multiple name matching degrees of the target entity segment, multiple metadata objects associated with the target entity segment are filtered to determine at least one target metadata object associated with the target entity segment, wherein the target entity segment is any entity segment among the multiple entity segments, and the name matching degree corresponding to the target metadata object is greater than or equal to the matching degree threshold.

[0071] If the number of at least one target metadata object associated with the target entity segmentation is less than or equal to the number threshold, all at least one target metadata object associated with the target entity segmentation will be identified as the key object.

[0072] If the number of at least one target metadata object associated with the target entity segmentation is greater than the number threshold, then among the at least one target metadata object associated with the target entity segmentation, the K target metadata objects with the highest name matching degree are all determined as the key objects, where K is the number threshold.

[0073] Based on the above settings, the number of multiple key objects can be quickly reduced by combining the matching degree threshold and the number threshold. While ensuring the accuracy and comprehensiveness of the data in subsequent analysis operations, the scale of the data involved in subsequent analysis operations is controllable, thereby improving the output efficiency and accuracy of the final results.

[0074] The process of identifying the top K target metadata objects with the highest name matching degree as the key objects is as follows: First, sort at least one target metadata object associated with the target entity segmentation in descending order to obtain the object sequence corresponding to the target entity segmentation. Then, identify the top K target metadata objects in the object sequence corresponding to the target entity segmentation as key objects.

[0075] For example, the matching degree threshold can be set to 0.6 based on experience, and the number threshold can be set to 5 based on experience.

[0076] Step S2: In the object space corresponding to the multiple key objects, search for multiple candidate object groups with context sharing relationship.

[0077] Among them, all key objects included in the candidate object group have at least one common metric group ID.

[0078] The object space corresponding to multiple key objects can be understood as a search space formed by randomly combining any number of different key objects among the multiple key objects, which is used to indicate all possible combinations of multiple key objects.

[0079] As mentioned earlier, although the number of key objects was reduced based on the matching degree threshold and the number threshold, the number of object groups generated by randomly combining the remaining key objects was still large, and most of them had structural conflicts. Limiting the common metric group ID can effectively filter out this large number of unreasonable object groups, thereby further reducing the number of object groups to be carefully screened in the future, thus improving the output efficiency of the final result and reducing the computing resources occupied during the final result output.

[0080] Specifically, the step of searching for multiple candidate object groups with context-sharing relationships in the object space corresponding to the multiple key objects includes:

[0081] An object graph is constructed using the multiple key objects as graph nodes, wherein the connecting edges in the object graph are used to indicate that: two corresponding key objects are associated with different entity segments, and two corresponding key objects have a common metric value group ID;

[0082] Perform a maximally connected subgraph search on the object graph to obtain multiple initial subgraphs;

[0083] The intersection of the multiple initial subgraphs is verified to obtain the multiple candidate object groups.

[0084] The above settings, through the construction of an object graph, enable topological processing of the object space corresponding to multiple key objects, simplify the search difficulty of multiple candidate object groups, and filter out significantly unreasonable object combinations, thus completing the initial screening of the combination methods of multiple candidate object groups.

[0085] The aforementioned significantly unreasonable object combination methods specifically refer to: combining different key objects that are associated with the same entity segmentation, and combining different key objects that do not have a common metric group ID.

[0086] In this invention, the Bron-Kerbosch algorithm is specifically used to complete the above-mentioned maximal connected subgraph search operation. In practical applications, other connected subgraph search algorithms can also be used to complete the above operation, and this invention does not limit this to any particular algorithm.

[0087] The intersection verification process described above is as follows: check whether the multiple key objects included in the corresponding initial subgraph have a common metric group ID. If they do, it is determined that the multiple key objects included in the corresponding initial subgraph constitute a candidate object group. If they do not, it is determined that the corresponding initial subgraph cannot be used to construct a candidate object group.

[0088] Step S3: Analyze the content matching degree between each candidate object group and the query request text in the plurality of candidate object groups to obtain the demand matching value of each candidate object group.

[0089] In the aforementioned steps, after filtering the metadata objects associated with the entity and the combination methods of the metadata objects, the number of candidate object groups is effectively controlled. However, the remaining candidate object groups may have two or more queries that are not empty. If they are output directly without distinction, it will not only affect the user experience, but also cause unnecessary computing and network resource overhead. Based on this, the above settings analyze the content matching degree between each candidate object group and the query request text to support the output of unique query results that have a high probability of hitting the user's needs. This can effectively overcome the above problems and improve the processing effect of natural language for users.

[0090] Specifically, the step of analyzing the content matching degree between each candidate object group and the query request text to obtain the demand matching value for each candidate object group includes:

[0091] Analyze the name matching degree between the target candidate object group and the query request text to obtain the name matching factor of the target candidate object group, wherein the target candidate object group is any one of the multiple candidate object groups;

[0092] Analyze the entity coverage between the target candidate object group and the query request text to obtain the entity coverage factor of the target candidate object group;

[0093] The demand matching value of the target candidate object group is determined based on the name matching factor and entity coverage factor of the target candidate object group.

[0094] For any candidate object group among multiple candidate object groups, the higher the degree of matching between the object name of the key object included in it and the corresponding entity segmentation, the higher the degree of name matching between the candidate object group and the query request text, that is, the higher the probability that the candidate object group hits the user's real query needs.

[0095] Correspondingly, for any candidate object group among multiple candidate object groups, the more entity segments it has, the higher the probability that the candidate object group hits the user's actual query needs.

[0096] The above settings comprehensively determine the content matching differences between different candidate object groups and the query request text from two aspects: name matching degree and entity hit rate, so as to ensure the accuracy of the calculated demand matching value.

[0097] The name matching factor and the corresponding demand matching value are positively correlated, and the entity coverage factor and the corresponding demand matching value are also positively correlated.

[0098] Specifically, the name matching factor of the target candidate object group is the sum of the name matching degrees of all the target candidate object groups.

[0099] The step of analyzing the entity coverage between the target candidate object group and the query request text to obtain the entity coverage factor of the target candidate object group includes:

[0100] Calculate the ratio of the number of all entity segments corresponding to the target candidate object group to the number of all entity segments included in the query request text to obtain the entity coverage ratio of the target candidate object group;

[0101] The entity coverage ratio of the target candidate object group is amplified to obtain the entity coverage factor of the target candidate object group.

[0102] The reason for amplifying the entity coverage ratio of the candidate object group in the above settings (specifically, this invention uses an exponential amplification method for adjustment) is to enhance the numerical influence of the entity coverage factor on the demand matching value, amplify the numerical advantage of the candidate object group with high entity coverage, and increase the probability that the candidate object group with high entity coverage will be recommended for output.

[0103] For example, a certain candidate object group Demand matching value It can be represented as:

[0104]

[0105] in, Represents candidate object group The first in Name matching degree of key objects Represents candidate object group The total number of corresponding entity segments, where N represents the total number of entity segments obtained after the query request text has undergone segmentation processing. Represented by natural constant An exponential function with base 0.

[0106] Step S4: Among the multiple candidate object groups, the candidate object group that passes the non-empty check and has the highest requirement matching value is determined as the target object group, and a requirement parsing result is generated based on the target object group.

[0107] Specifically, step S4 includes:

[0108] Based on the demand matching value of each candidate object group, the multiple candidate object groups are sorted in descending order to obtain the object group sequence.

[0109] Starting from the first sequence element of the object group sequence, the traversal is terminated if the currently traversed sequence element passes the non-empty check, and the currently traversed sequence element is determined as the target object group.

[0110] Based on the above sorting and the measure of terminating traversal if the first query is not empty, the cost of obtaining the query results of the candidate object group with the highest matching value and non-empty query results can be reduced. This can avoid unnecessary resource overhead while ensuring the accuracy of the final output target query results.

[0111] It should be noted that if all candidate object groups in the object group sequence fail the non-empty check, a prompt message indicating that the query result corresponding to the query text is empty will be directly output.

[0112] It should be added that, since the candidate object group only includes various metadata objects and corresponding entity segmentation (such as sales revenue, 2022, etc.), and lacks the complete context information (such as specific currencies) required to perform specific checks, this invention will combine system default information to complete it in actual implementation to ensure that each candidate object group can be checked smoothly.

[0113] Specifically, before performing a non-null check based on a candidate object group, a non-null check original table is pre-determined, which includes all dimensions involved in the non-null check and the default value of each dimension (obtained through system default information). Then, the non-null check original table is updated using several entity tokens corresponding to the corresponding candidate object group to replace the default values ​​of the corresponding dimensions in the non-null check original table. Then, the non-null check operation is performed based on the updated non-null check original table (using a bitmap index or pre-aggregated table header to manage data distribution; the operation of checking whether there is data at a multi-dimensional coordinate point only involves index scanning and does not involve actual data block reading and deserialization, and its overhead is much lower than performing a complete data query).

[0114] In this invention, the process of generating requirement analysis results based on the target object group is as follows:

[0115] Based on the object name and type identifier of each key object in the target object group, the preset MDX query template is populated. Key objects with the type identifier MEASURE (as the data column of the query) are populated into the ON COLUMNS clause of the MDX query template; key objects with the type identifier DIMENSION (as the row header of the query) are populated into the ON ROWS clause of the MDX query template; and metadata objects supplemented based on system default information during the non-null check (as the global slicing condition of the query) are populated into the WHERE clause of the MDX query template, thus obtaining the requirement parsing result. In an example, when the query request text is "I want to see the sales data for 2023", the corresponding output requirement parsing result can be represented as:

[0116] SELECT

[0117] { [Measures].[Sales Amount], ...} ON COLUMNS,

[0118] { [Time].

[2023] , [Product].[Category A], ...} ON ROWS

[0119] FROM [Sales Cube]

[0120] WHERE ( [Currency].[USD], ... ).

[0121] In summary, this invention first performs a correlation search in the MDX database based on the user's query request text to initially filter out several metadata objects that are related to the query request text. Then, among the filtered metadata objects, it searches for multiple object combinations that can be used to respond to the query request text using a common metric group ID as the search condition. Furthermore, it analyzes the content matching degree between each object combination and the query request text. Finally, it provides the user with the requirement parsing result corresponding to the object combination that passes the non-empty check and has the highest content matching degree. This can avoid the computational resource overhead and time overhead caused by executing multiple MDX query languages, and improve the natural language processing effect for users in the field of MDX applications.

[0122] In some embodiments, the present invention also provides an intelligent semantic parsing system based on the MDX framework, such as... Figure 2 As shown, the system 200 includes:

[0123] The related search module 201 is used to perform related searches in the MDX database based on the user's query request text to obtain multiple key objects, wherein the key objects are metadata objects in the MDX database that are related to the query request text, and the metadata objects correspond to at least one metric group ID.

[0124] The candidate search module 202 is used to search for multiple candidate object groups with context sharing relationships in the object space corresponding to the multiple key objects, wherein all key objects included in the candidate object group have at least one common metric group ID.

[0125] Matching analysis module 203 is used to analyze the content matching degree between each candidate object group and the query request text in the plurality of candidate object groups, and obtain the required matching value of each candidate object group;

[0126] The result determination module 204 is used to determine the candidate object group with the highest requirement matching value that passes the non-empty check as the target object group among the plurality of candidate object groups, and generate a requirement parsing result based on the target object group.

[0127] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the intelligent semantic parsing system based on the MDX framework and the intelligent semantic parsing method based on the MDX framework provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0128] This invention also provides an electronic device. Please refer to [link to relevant documentation]. Figure 3 The electronic device may include a processor 301, a memory 302, and a program 3021 stored in the memory 302 and capable of running on the processor 301.

[0129] When program 3021 is executed by processor 301, it can achieve the following: Figure 1 Any steps in the corresponding method embodiments and the achievement of the same beneficial effects will not be repeated here.

[0130] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by hardware related to program instructions, and the program can be stored in a readable medium.

[0131] This invention also provides a readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described functions. Figure 1 Any step in the corresponding method embodiment can achieve the same technical effect, and will not be repeated here to avoid repetition.

[0132] The computer-readable storage medium of this invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0133] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0134] The program code contained on the storage medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0135] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or terminal. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0136] This invention also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned steps to implement the intelligent semantic parsing method based on the MDX framework provided in the above embodiments.

[0137] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0138] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. An intelligent semantic parsing method based on the MDX framework, characterized in that, The method includes: Based on the user's query request text, a correlation search is performed in the MDX database to obtain multiple key objects. The key objects are metadata objects in the MDX database that are related to the query request text, and the metadata objects correspond to at least one metric group ID. In the object space corresponding to the multiple key objects, search for multiple candidate object groups with context sharing relationship, wherein all key objects included in the candidate object group have at least one common metric group ID; Among the multiple candidate object groups, the content matching degree between each candidate object group and the query request text is analyzed to obtain the demand matching value of each candidate object group; Among the multiple candidate object groups, the candidate object group that passes the non-empty check and has the highest requirement matching value is determined as the target object group, and the requirement parsing result is generated based on the target object group; The step of searching for multiple candidate object groups with context-sharing relationships in the object space corresponding to the multiple key objects includes: An object graph is constructed using the multiple key objects as graph nodes, wherein the connecting edges in the object graph are used to indicate that: two corresponding key objects are associated with different entity segments, and two corresponding key objects have a common metric value group ID; Perform a maximally connected subgraph search on the object graph to obtain multiple initial subgraphs; The intersection of the multiple initial subgraphs is verified to obtain the multiple candidate object groups; The step of analyzing the content matching degree between each candidate object group and the query request text in the plurality of candidate object groups to obtain the demand matching value for each candidate object group includes: Analyze the name matching degree between the target candidate object group and the query request text to obtain the name matching factor of the target candidate object group, wherein the target candidate object group is any one of the multiple candidate object groups; Analyze the entity coverage between the target candidate object group and the query request text to obtain the entity coverage factor of the target candidate object group; The demand matching value of the target candidate object group is determined based on the name matching factor and entity coverage factor of the target candidate object group; The name matching factor of the target candidate object group is the sum of the name matching degrees of all the target candidate object groups.

2. The intelligent semantic parsing method based on the MDX framework according to claim 1, characterized in that, The steps involved in performing a correlation search in the MDX database based on the user's query request text to obtain multiple key objects include: The query request text is segmented into words to obtain multiple entity words; In the multiple entity segmentations, obtain multiple metadata objects associated with each entity segmentation in the MDX database; In the multiple entity segmentations, the degree of matching between each entity segmentation and the object name of each associated metadata object is analyzed to obtain multiple name matching degrees for each entity segmentation; Based on the name matching degree of each entity segment, all metadata objects associated with multiple entity segments are filtered to obtain multiple key objects, wherein the name matching degree of the key objects is greater than or equal to the matching degree threshold.

3. The intelligent semantic parsing method based on the MDX framework according to claim 2, characterized in that, The steps for filtering all metadata objects associated with multiple entity segmentations based on multiple name matching degrees for each entity segmentation to obtain multiple key objects include: Based on the matching degree threshold and multiple name matching degrees of the target entity segment, multiple metadata objects associated with the target entity segment are filtered to determine at least one target metadata object associated with the target entity segment, wherein the target entity segment is any entity segment among the multiple entity segments, and the name matching degree corresponding to the target metadata object is greater than or equal to the matching degree threshold. If the number of at least one target metadata object associated with the target entity segmentation is less than or equal to the number threshold, all at least one target metadata object associated with the target entity segmentation will be identified as the key object. If the number of at least one target metadata object associated with the target entity segmentation is greater than the number threshold, then among the at least one target metadata object associated with the target entity segmentation, the K target metadata objects with the highest name matching degree are all determined as the key objects, where K is the number threshold.

4. The intelligent semantic parsing method based on the MDX framework according to claim 1, characterized in that, The step of analyzing the entity coverage between the target candidate object group and the query request text to obtain the entity coverage factor of the target candidate object group includes: Calculate the ratio of the number of all entity segments corresponding to the target candidate object group to the number of all entity segments included in the query request text to obtain the entity coverage ratio of the target candidate object group; The entity coverage ratio of the target candidate object group is amplified to obtain the entity coverage factor of the target candidate object group.

5. The intelligent semantic parsing method based on the MDX framework according to claim 1, characterized in that, The name matching factor and the corresponding demand matching value are positively correlated, and the entity coverage factor and the corresponding demand matching value are also positively correlated.

6. The intelligent semantic parsing method based on the MDX framework according to claim 1, characterized in that, The step of determining the candidate object group with the highest requirement matching value that passes the non-empty check as the target object group among the plurality of candidate object groups includes: Based on the demand matching value of each candidate object group, the multiple candidate object groups are sorted in descending order to obtain the object group sequence. Starting from the first sequence element of the object group sequence, the traversal is terminated if the currently traversed sequence element passes the non-empty check, and the currently traversed sequence element is determined as the target object group.

7. An intelligent semantic parsing system based on the MDX framework, characterized in that, The system includes: The related search module is used to perform related searches in the MDX database based on the user's query request text to obtain multiple key objects. The key objects are metadata objects in the MDX database that are related to the query request text, and the metadata objects correspond to at least one metric group ID. The candidate search module is used to search for multiple candidate object groups with context-sharing relationships in the object space corresponding to the multiple key objects, wherein all key objects included in the candidate object group have at least one common metric group ID. The matching analysis module is used to analyze the content matching degree between each candidate object group and the query request text in the plurality of candidate object groups, and to obtain the required matching value for each candidate object group. The result determination module is used to determine the candidate object group with the highest requirement matching value that passes the non-empty check as the target object group among the multiple candidate object groups, and generate requirement parsing results based on the target object group. The step of searching for multiple candidate object groups with context-sharing relationships in the object space corresponding to the multiple key objects includes: An object graph is constructed using the multiple key objects as graph nodes, wherein the connecting edges in the object graph are used to indicate that: two corresponding key objects are associated with different entity segments, and two corresponding key objects have a common metric value group ID; Perform a maximally connected subgraph search on the object graph to obtain multiple initial subgraphs; The intersection of the multiple initial subgraphs is verified to obtain the multiple candidate object groups; The step of analyzing the content matching degree between each candidate object group and the query request text in the plurality of candidate object groups to obtain the demand matching value for each candidate object group includes: Analyze the name matching degree between the target candidate object group and the query request text to obtain the name matching factor of the target candidate object group, wherein the target candidate object group is any one of the multiple candidate object groups; Analyze the entity coverage between the target candidate object group and the query request text to obtain the entity coverage factor of the target candidate object group; The demand matching value of the target candidate object group is determined based on the name matching factor and entity coverage factor of the target candidate object group; The name matching factor of the target candidate object group is the sum of the name matching degrees of all the target candidate object groups.

Images