Semantic alignment and data query method and device for data elements

Abstract

The invention provides a semantic alignment and data query method and device for data elements, and relates to the technical field of computers. The semantic alignment method for the data elements is executed by the electronic equipment and comprises the steps that a field sequence is extracted from to-be-processed text data, and the field sequence comprises a plurality of fields; determining a first similarity between each field and each attribute in a plurality of attributes contained in the data element standard specification; selecting a candidate attribute corresponding to each field from the plurality of attributes according to the first similarity; constructing a plurality of candidate attribute sequences corresponding to the field sequence according to the candidate attributes corresponding to the plurality of fields; determining a second similarity between the field sequence and each candidate attribute sequence; and selecting an attribute sequence matched with the field sequence from the plurality of candidate attribute sequences according to the second similarity. By means of the method, data semantic alignment processing can be efficiently and accurately carried out.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a method and device for semantic alignment and data query of data elements. Background Art

[0002] The core purpose of data element processing is to improve the interoperability and value of data. In order to realize data element processing, it is necessary to understand the content of non-standardized data (such as forms) and semantically align it with the data element standard specifications. Taking forms as an example, the semantic alignment can be understood as the process of matching the fields in forms of different sources and formats with the data elements of "attributes" in the data element standard specifications. The above processing process not only realizes the connection of data logical semantics, but also lays the foundation for cross-departmental and cross-system data sharing and collaboration.

[0003] In the related technologies, most of them are based on manual methods to achieve semantic alignment between non-standard data and data element standard specifications. In addition, in the related technologies, there are also automatic extraction technologies based on machine learning models. For example, machine learning models are used to annotate data entities. Summary of the invention

[0004] The present disclosure provides a method and device for semantic alignment and data query of data elements.

[0005] According to a first aspect of the present disclosure, a semantic alignment method for data elements is provided, which is executed by an electronic device and includes: extracting a field sequence from text data to be processed, the field sequence including multiple fields; determining a first similarity between each field in the multiple fields and each attribute in a multiple attributes included in a data element standard specification; selecting a candidate attribute corresponding to each field from the multiple attributes based on the first similarity; constructing a plurality of candidate attribute sequences corresponding to the field sequence based on the candidate attributes corresponding to the multiple fields; determining a second similarity between the field sequence and each candidate attribute sequence in the multiple candidate attribute sequences; and selecting an attribute sequence matching the field sequence from the multiple candidate attribute sequences based on the second similarity.

[0006] In some embodiments, determining the first similarity between each of the multiple fields and each of the multiple attributes included in the data element standard specification includes: using a text representation language model to embed each of the fields to obtain an embedded representation vector of each field; determining a first vector similarity between the embedded representation vector of each field and the embedded representation vector of each attribute; determining a second vector similarity between the embedded representation vector of each field and the embedded representation vector of the entity or relationship to which each attribute belongs; and determining the first similarity based on the first vector similarity and the second vector similarity.

[0007] In some embodiments, the data element standard specification also includes entities and relationships, and determining the second similarity between the field sequence and each candidate attribute sequence in the multiple candidate attribute sequences includes: constructing a structural representation graph of each candidate attribute sequence; determining the connectivity of the structural representation graph; and determining the second similarity between the field sequence and each candidate attribute sequence based on the connectivity of the structural representation graph and the first similarities of all candidate attributes in each candidate attribute sequence.

[0008] In some embodiments, determining the first similarity based on the first vector similarity and the second vector similarity includes: determining a first weighted similarity based on the first vector similarity and the importance weight of each attribute in the data element standard specification; determining a second weighted similarity based on the second vector similarity and the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification; determining the first similarity based on the first weighted similarity and the second weighted similarity.

[0009] In some embodiments, the importance weight of each attribute in the data element standard specification is determined as follows: using an importance evaluation algorithm to determine the importance evaluation value of each attribute and the importance evaluation value of the attribute term corresponding to each attribute; determining the importance weight of each attribute in the data element standard specification based on the average of the importance evaluation value of each attribute and the importance evaluation value of the attribute term corresponding to each attribute.

[0010] In some embodiments, the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification is determined as follows: using an importance evaluation algorithm to determine the importance evaluation value of the entity or relationship to which each attribute belongs, and the importance evaluation value of the term corresponding to the entity or relationship to which each attribute belongs; determining the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification based on the average of the importance evaluation value of the entity or relationship to which each attribute belongs and the importance evaluation value of the term corresponding to the entity or relationship to which each attribute belongs.

[0011] In some embodiments, the importance assessment algorithm is a term frequency-inverse document frequency algorithm.

[0012] In some embodiments, the first weighted similarity is positively correlated with the first vector similarity and the importance weight of each attribute in the data element standard specification; and / or, the second weighted similarity is positively correlated with the second vector similarity and the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification; and / or, the first similarity is positively correlated with the first weighted similarity and the second weighted similarity.

[0013] In some embodiments, the first weighted similarity is determined by multiplying the first vector similarity by the importance weight of each attribute in the data element standard specification; and / or, the second weighted similarity is determined by multiplying the second vector similarity by the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification; and / or, the first similarity is determined by multiplying the first weighted similarity by the second weighted similarity.

[0014] In some embodiments, the embedding representation vector of each attribute is determined according to the following method: using a text representation language model, each attribute and the attribute term corresponding to each attribute are respectively embedded to obtain a first embedding vector of each attribute and a second embedding vector of the attribute term corresponding to each attribute; based on the first embedding vector of each attribute and the second embedding vector of the attribute term corresponding to each attribute, the embedding representation vector of each attribute is determined.

[0015] In some embodiments, the second similarity is positively correlated with the connectivity of the structure representation graph and the first similarities corresponding to all candidate attributes in each candidate attribute sequence.

[0016] In some embodiments, determining the second similarity between the field sequence and each candidate attribute sequence based on the connectivity of the structure representation graph and the first similarities corresponding to all candidate attributes in each candidate attribute sequence includes: determining the comprehensive attribute similarity based on the sum of the first similarities corresponding to all candidate attributes in each candidate attribute sequence; determining the second similarity between the field sequence and each candidate attribute sequence based on the product of the comprehensive attribute similarity and the connectivity of the structure representation graph.

[0017] In some embodiments, constructing the structural representation graph of each candidate attribute sequence includes: for each candidate attribute in the candidate attribute sequence, if the candidate attribute belongs to an attribute of an entity, setting a node; if the candidate attribute belongs to an attribute of a relationship, setting an edge; iteratively executing the setting step to obtain the structural representation graph.

[0018] In some embodiments, selecting the candidate attribute corresponding to each field from the multiple attributes based on the first similarity includes: taking the N attributes among the multiple attributes that have the largest first similarity with each field as the candidate attributes corresponding to each field, where N is an integer value greater than 1 and less than M, and M is the total number of the multiple attributes.

[0019] In some embodiments, constructing a plurality of candidate attribute sequences corresponding to the field sequence according to the candidate attributes corresponding to the plurality of fields includes: arranging and combining the candidate attributes corresponding to the plurality of fields to obtain the plurality of candidate attribute sequences.

[0020] According to a second aspect of the present disclosure, a data query method is provided, comprising: determining an attribute sequence that matches the field sequence according to the semantic alignment method for data elements as described above, and storing the correspondence between the fields in the field sequence and the attributes in the attribute sequence; in response to a data query request, determining the attributes to be queried in the data element standard specification involved in the request; determining the fields that match the attributes to be queried according to the attributes to be queried and the correspondence; and performing a data query according to the fields that match the attributes to be queried to obtain a query result.

[0021] According to a third aspect of the present disclosure, an electronic device is provided, comprising: a memory; and a processor coupled to the memory, the processor being configured to execute the semantic alignment method for data elements as described above, or the data query method as described above, based on instructions stored in the memory.

[0022] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored, and when the instructions are executed by a processor, the semantic alignment method for data elements as described above, or the data query method as described above is implemented.

[0023] According to a fifth aspect of the present disclosure, a computer program product is provided, on which computer program instructions are stored, and when the instructions are executed by a processor, the semantic alignment method for data elements as described above, or the data query method as described above is implemented.

[0024] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings, which constitute a part of the specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0026] Figure 1 is a flowchart of a method for semantic alignment of data elements according to some embodiments of the present disclosure;

[0027] Figure 2a is a schematic diagram of text data to be processed according to some embodiments of the present disclosure;

[0028] Figure 2b is a schematic diagram of text data to be processed according to some embodiments of the present disclosure;

[0029] Figure 3a is a schematic diagram of a data element standard specification according to some embodiments of the present disclosure;

[0030] Figure 3b is a schematic diagram of a data element standard specification according to some embodiments of the present disclosure;

[0031] Figure 4 is a schematic diagram of a process of determining a first similarity according to some embodiments of the present disclosure;

[0032] Figure 5 is a schematic diagram of a process of determining a second similarity according to some embodiments of the present disclosure;

[0033] Figure 6 is a flowchart of a method for semantic alignment of data elements according to other embodiments of the present disclosure;

[0034] Figure 7 is a schematic diagram of the calculation principle of the importance weight of data elements according to some embodiments of the present disclosure;

[0035] Figure 8is a flowchart of a data query method according to some embodiments of the present disclosure;

[0036] Fig. 9 is a structural schematic diagram of a semantic alignment device for data elements according to some embodiments of the present disclosure;

[0037] Fig.10 is a structural schematic diagram of a data query device according to some embodiments of the present disclosure;

[0038] Fig.11 is a schematic diagram of the structure of an electronic device according to some embodiments of the present disclosure;

[0039] Fig.12 is a schematic diagram of the structure of a computer system according to some embodiments of the present disclosure.

[0040] The present disclosure can be more clearly understood from the following detailed description with reference to the accompanying drawings. DETAILED DESCRIPTION

[0041] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure unless otherwise specifically stated.

[0042] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0043] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

[0044] Technologies, methods, and apparatus known to ordinary technicians in the relevant field may not be discussed in detail, but where appropriate, such technologies, methods, and apparatus should be considered part of the authorization specification.

[0045] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0046] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0047] In order to make the objectives, technical solutions and advantages of the present disclosure more clearly understood, the present disclosure is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0048] Data element processing (or semantic alignment of data elements) not only realizes the logical semantics of data, but also lays the foundation for cross-departmental and cross-system data sharing and collaboration. For example, in the field of government services, the above-mentioned processing can convert non-standard data into data elements in standard specifications, which can greatly simplify the process of citizens' affairs and significantly improve administrative efficiency and service quality. In enterprise management, the above-mentioned processing can convert non-standard data into data elements in standard specifications, which can help break information silos, promote collaborative work among departments, and improve the accuracy and timeliness of decision-making.

[0049] In the related art, semantic alignment of data elements based on manual methods is heavily dependent on the personal ability and work habits of the staff, and the processing efficiency is low. In addition, in the related art, data semantic alignment based on automatic extraction technology has a processing accuracy that is greatly affected by data quality and data complexity, making it difficult to meet high-precision processing requirements, and is overly dependent on external data sources, limiting the scope of application and flexibility of the algorithm.

[0050] In view of this, the present disclosure proposes a method and device for semantic alignment and data query of data elements, which can accurately, efficiently and flexibly realize semantic alignment processing of various text data to be processed and data element standard specifications.

[0051] Figure 1 FIG. 1 is a flow chart of a method for semantic alignment of data elements according to some embodiments of the present disclosure. Figure 1 As shown, the semantic alignment method for data elements includes steps S11 to S16.

[0052] In step S11, a field sequence is extracted from the text data to be processed.

[0053] The text data to be processed may be a form, a spreadsheet or other forms of text data. For example, the text data to be processed is Figure 2a Employee leave form in , or Figure 2b The basic information registration form of the enterprise, or the questionnaire, etc.

[0054] When the text data to be processed is form data, a field sequence is extracted from the form data in a set order. The field sequence includes multiple fields. For example, the fields are extracted in order from top to bottom and from left to right to obtain a field sequence.

[0055] For example, for Figure 2a The form extracts multiple fields from top to bottom. The extracted field sequence is ["leave number", "filling time", "name", "department", "leave type", "leave reason", "leave time", "total number of days", "human resources department"].

[0056] For example, for Figure 2b The form extracts multiple fields from top to bottom and from left to right. The extracted field sequence is ["Company Name (Seal)", "Organization Code", "(Main Item) Qualification", "Qualification Level", "Company Address", "Postal Code"].

[0057] In step S12, a first similarity between each of the plurality of fields and each of the plurality of attributes included in the data element standard specification is determined.

[0058] Data element standards and specifications are a set of data element specifications. Through data element standards and specifications, data governance and data lineage analysis can be achieved. Data element standards and specifications define three types of data elements: entities, relationships, and attributes, which can be expressed in the form of knowledge graphs.

[0059] Entities, such as people, places, objects, organizations, etc. Furthermore, entities can be further divided into multiple entity categories. For example, people can include historical figures, places can include houses, roads, parking lots, administrative divisions, etc., objects can include vehicles, ID cards, social security cards, air, etc., and organizations can include companies, schools, social groups, etc.

[0060] Relationship refers to the connection between entities. For example, relationships include similar relationships, mutual relationships, relationships between three or more entities, etc. Among them, similar relationships include relationships between people and people, and relationships between things; mutual relationships include relationships between people and places, people and things, and people and things. Furthermore, data element standards and specifications may also include relationship terms. Relationship terms are words used to describe the relationship between entities, which are more specific relationships. For example, relationship terms between people include: service, conflict, collaboration, management, guidance, blood relationship, etc., relationship terms between things include: subordination, connection, space, etc., and relationship terms between people and organizations include subordination, jurisdiction, etc.

[0061] Attributes are information used to describe the characteristics of an entity or relationship. For example, attributes include entity attributes and relationship attributes. Entity attributes are used to describe information about the characteristics of the entity itself, such as a person's name and age, a place's location and area, etc. Relationship attributes are used to describe information about the characteristics of the relationship between entities, such as the time, category, price, and evaluation of a service relationship. Furthermore, data element standards and specifications may also include attribute terms. Attribute terms are words used to describe entity attributes or relationship attributes, and are more specific attributes. For example, the attribute terms corresponding to a person's "name" include former name, nickname, etc.

[0062] In some examples, the data element standard specification includes a city knowledge system. Figure 3a and Figure 3bAs shown in the figure, the urban knowledge system provides a consistent expression basis for public data by abstracting and condensing urban business content, taking five types of entities, namely people, places, things, objects, and organizations, and the relationships and attributes between entities as the basic data elements.

[0063] In step S12, for each field in the field sequence, a first similarity between the field and each of the multiple attributes included in the data element standard specification is determined. For example, assuming that the field sequence has 10 fields and the data element standard specification has 20 attributes, the first similarity between each field and each of the 20 attributes is calculated.

[0064] There are many ways to determine the first similarity. The following two ways are used as examples for illustration. The first way to determine the first similarity is as follows: Figure 4 As shown, it includes steps S121 to S124.

[0065] Step S121: using the text representation language model, embedding each field to obtain an embedding representation vector for each field.

[0066] For example, the text representation language model is a pre-trained word embedding model. In step S121, the field is input into the model to obtain an embedding representation vector of the field.

[0067] Furthermore, before executing step S121, the following may be included: preprocessing the field. For example, preprocessing the field by segmenting, removing stop words, standardizing, etc. Then, the preprocessed field is input into the pretrained model to obtain an embedded representation vector of the field.

[0068] Step S122: Determine a first vector similarity between the embedded representation vector of each field and the embedded representation vector of each attribute.

[0069] Among them, the embedded representation vector of the attribute, and the embedded representation vector of the entity or relationship to which the attribute belongs mentioned later, can be obtained by pre-processing. In this way, when executing steps S122 and S123, the embedded representation vector of the attribute, and the embedded representation vector of the entity or relationship to which the attribute belongs can be directly obtained from the storage module. In addition, the text representation language model can also be used to process the attribute, entity, or relationship in real time to obtain the embedded representation vector of the attribute, entity, or relationship.

[0070] Specifically, the embedded representation vector of the attribute can be determined according to the following exemplary method: using the text representation language model, each attribute and the attribute term corresponding to each attribute are embedded to obtain the first embedding vector of each attribute and the second embedding vector of the attribute term corresponding to each attribute; based on the first embedding vector of each attribute and the second embedding vector of the attribute term corresponding to each attribute, the embedded representation vector of each attribute is determined.

[0071] For example, after obtaining the first embedding vector of the attribute and the second embedding vector of the attribute term corresponding to the attribute, the first embedding vector and the second embedding vector are bitwise averaged to obtain the embedding representation vector of the attribute. In addition, the embedding representation vector of the entity to which the attribute belongs and the embedding representation vector of the attribute relationship can also be determined in a similar manner.

[0072] In step S122, the first vector similarity can be determined in a variety of ways. For example, the cosine similarity between the embedded representation vector of each field and the embedded representation vector of each attribute is calculated and used as the first vector similarity. In addition, other ways can be used to determine the first vector similarity. For example, the first vector similarity is calculated based on the distance (such as the Euclidean distance) between the embedded representation vector of the field and the embedded representation vector of the attribute.

[0073] Step S123: Determine a second vector similarity between the embedded representation vector of each field and the embedded representation vector of the entity or relationship to which each attribute belongs.

[0074] In step S123, the second vector similarity can be determined in a variety of ways. For example, if a certain attribute belongs to an attribute of an entity, the cosine similarity between the embedded representation vector of the field and the embedded representation vector of the entity to which the attribute belongs is calculated, and is used as the second vector similarity; if a certain attribute belongs to an attribute of a relationship, the cosine similarity between the embedded representation vector of the field and the embedded representation vector of the relationship to which the attribute belongs is calculated, and is used as the second vector similarity. In addition, other methods can be used to calculate the second vector similarity. For example, the second vector similarity is calculated based on the distance between the embedded representation vector of the field and the embedded representation vector of the entity or relationship to which the attribute belongs.

[0075] Step S124: determining a first similarity according to the first vector similarity and the second vector similarity.

[0076] Step S124 can be implemented in a variety of ways, and two implementations are used as examples for description.

[0077] In a first implementation, step S124 includes: taking the product of the first vector similarity and the second vector similarity as the first similarity.

[0078] In a second embodiment, step S124 includes: determining a first weighted similarity based on the first vector similarity and the importance weight of each attribute in the data element standard specification; determining a second weighted similarity based on the second vector similarity and the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification; determining the first similarity based on the first weighted similarity and the second weighted similarity.

[0079] Among them, the importance weight of the attribute can be determined according to the following exemplary method: using the importance evaluation algorithm to determine the importance evaluation value of each attribute and the importance evaluation value of the attribute term corresponding to each attribute; calculating the importance evaluation value of each attribute and the average of the importance evaluation value of the attribute term corresponding to each attribute, and determining the importance weight of each attribute in the data element standard specification accordingly.

[0080] Among them, the importance weight of the entity or relationship to which the attribute belongs can be determined according to the following exemplary method: using the importance evaluation algorithm, determine the importance evaluation value of the entity or relationship to which each attribute belongs, and the importance evaluation value of the term corresponding to the entity or relationship to which each attribute belongs; calculate the importance evaluation value of the entity or relationship to which each attribute belongs, and the average of the importance evaluation value of the term corresponding to the entity or relationship to which each attribute belongs, and determine the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification accordingly. For example, the importance evaluation algorithm is the term frequency-inverse document frequency (TF-IWF) algorithm.

[0081] Further, in the second embodiment, the first weighted similarity, the second weighted similarity, and the first similarity may also satisfy one or more of the following: the first weighted similarity is positively correlated with the first vector similarity and the importance weight of the attribute; the second weighted similarity is positively correlated with the second vector similarity and the importance weight of the entity or relationship to which the attribute belongs; the first similarity is positively correlated with the first weighted similarity and the second weighted similarity. For example, the first weighted similarity is determined based on the product of the first vector similarity and the importance weight of the attribute; the second weighted similarity is determined based on the product of the second vector similarity and the importance weight of the entity or relationship to which the attribute belongs; the first similarity is determined based on the product of the first weighted similarity and the second weighted similarity.

[0082] In the disclosed embodiment, by first embedding the attributes in the fields and data element standard specifications, the semantic information of the fields and attributes can be captured, so that the matching process of the fields and attributes is no longer limited to literal similarity, but is improved to a richer semantic level of similarity, so that fields with different expressions but similar meanings can be identified, thereby improving the accuracy and flexibility of matching. Moreover, by considering the influence of the importance weights of different attributes in the data standard elements and the entities or relationships to which different attributes belong on the first similarity in the matching process of the fields and attributes, the matching process can more accurately identify and process key information, give priority to more important data elements, and further improve the accuracy of field and attribute matching.

[0083] The second method for determining the first similarity includes: using a text representation language model to embed each field to obtain an embedded representation vector for each field; determining the first vector similarity between the embedded representation vector of each field and the embedded representation vector of each attribute; and determining the first similarity based on the first vector similarity.

[0084] Furthermore, in the second method of determining the first similarity, there are multiple ways to determine the first similarity based on the first vector similarity. For example, the first vector similarity is used as the first similarity. For another example, a first weighted similarity is first determined based on the first vector similarity and the importance weight of the attribute, and then the first weighted similarity is used as the first similarity. Back Figure 1 In step S13, a candidate attribute corresponding to each field is selected from the multiple attributes according to the first similarity.

[0085] Step S13 can be implemented in many ways. In an exemplary implementation, N attributes among the multiple attributes that have the largest first similarity with the field are used as candidate attributes corresponding to the field, where N is an integer greater than 1 and less than M, and M is the total number of the multiple attributes.

[0086] In another exemplary implementation, an attribute among multiple attributes whose first similarity with a field is greater than a similarity threshold is used as a candidate attribute corresponding to the field.

[0087] In order to further improve data processing efficiency, in some embodiments, the semantic alignment method for data elements also includes: after step S11, for each field, traversing the attributes and attribute terms in the data element standard specification; when there is an attribute that is literally the same as the field, using the attribute as a candidate attribute of the field; when there is an attribute term that is literally the same as the field, using the attribute to which the attribute term belongs as a candidate attribute of the field; when there is no attribute or attribute term that is literally the same as the field, determining the candidate attributes of the field according to steps S12 and S13.

[0088] In step S14, multiple candidate attribute sequences corresponding to the field sequence are constructed according to the candidate attributes corresponding to the multiple fields.

[0089] In some examples, step S14 includes: arranging and combining candidate attributes corresponding to multiple fields to obtain multiple candidate attribute sequences, wherein the number of candidate attributes in each candidate attribute sequence is the same as the number of fields in the field sequence, and the candidate attributes in each candidate attribute sequence correspond to the fields in the field sequence one by one.

[0090] For example, assuming that the field sequence includes 5 fields and each field has two candidate attributes, 32 candidate attribute sequences can be obtained by permuting and combining the candidate attributes corresponding to the 5 fields.

[0091] In step S15, a second similarity between the field sequence and each candidate attribute sequence in the plurality of candidate attribute sequences is determined.

[0092] There are many ways to determine the second similarity. Figure 5 The method shown is used as an example for explanation. Figure 5 As shown, the method of determining the second similarity includes steps S141 to S143.

[0093] Step S141: construct a structural representation graph for each candidate attribute sequence.

[0094] In some examples, a structural representation graph is constructed as follows: for each candidate attribute in a candidate attribute sequence, if the candidate attribute belongs to an entity attribute, a node is set; if the candidate attribute belongs to a relationship attribute, an edge is set; the setting step is iteratively executed to obtain a structural representation graph of the candidate attribute sequence.

[0095] Step S142: Determine the connectivity of the structure representation graph.

[0096] The connectivity of a graph is a concept in graph theory, including the connectivity of nodes and the connectivity of edges. In some examples, the connectivity of a structure representation graph is specifically the node connectivity of the structure representation graph.

[0097] In some examples, the node connectivity of the structure representation graph is calculated as follows: Step 1, input the structure representation graph, which can be represented by an adjacency matrix or an adjacency list; Step 2, determine the node pairs: determine the node pairs (u, v) whose connectivity needs to be calculated; Step 3, use the maximum flow algorithm to calculate the maximum flow: take the node pair (u, v) as the source node and the sink node, and use the maximum flow algorithm (such as the Ford-Folkson algorithm or the Edmonds-Karp algorithm) to calculate the maximum flow from the source node to the sink node; Step 4, solve the minimum cut: according to the maximum flow minimum cut theorem, the maximum flow value is the minimum number of cuts from the source node to the sink node, and the minimum number of cuts is the node connectivity of the node pair (u, v); Step 5, repeat the calculation: repeat the above process for all node pairs in the graph, and take the minimum connectivity among all node pairs as the node connectivity of the structure representation graph.

[0098] Step S143: Determine the second similarity between the field sequence and each candidate attribute sequence according to the connectivity of the structure representation graph and the first similarities of all candidate attributes in each candidate attribute sequence.

[0099] In some examples, the second similarity is positively correlated with the connectivity of the structure representation graph and the first similarities corresponding to all candidate attributes in the candidate attribute sequence.

[0100] Further, in the above example, the second similarity can be determined based on the following method: determine the comprehensive attribute similarity according to the sum of the first similarities corresponding to all candidate attributes in each candidate attribute sequence; determine the second similarity between the field sequence and each candidate attribute sequence according to the product of the comprehensive attribute similarity and the connectivity of the structural representation graph. For example, first calculate the sum of the first similarities corresponding to all candidate attributes in the candidate attribute sequence, and use it as the comprehensive attribute similarity; then calculate the product of the comprehensive attribute similarity and the connectivity of the structural representation graph of the candidate attribute sequence, and use it as the second similarity between the field sequence and the candidate attribute sequence.

[0101] In the disclosed embodiment, by calculating the first similarity between the field and the attribute, and calculating the second similarity between the field sequence and the candidate attribute sequence, the local match is combined with the global match, which can improve the accuracy and robustness of the match, facilitate better processing of complex form structures and diversified field naming, and reduce matching errors and omissions. Further, by considering not only the first similarity of all candidate attributes in the candidate attribute sequence when calculating the second similarity, but also the connectivity of the structural representation graph of the candidate attribute sequence, the matching attribute sequence selected thereby is not only semantically similar to the field sequence, but also structurally reasonable, thereby improving the reliability and consistency of the matching attribute sequence selected thereby.

[0102] Back to Figure 1 In step S16, an attribute sequence matching the field sequence is selected from a plurality of candidate attribute sequences according to the second similarity.

[0103] For example, in step S16, the candidate attribute sequence with the second largest similarity among the multiple candidate attribute sequences is used as the attribute sequence that matches the field sequence.

[0104] In the disclosed embodiments, on the one hand, by using electronic devices to automatically realize the semantic alignment processing of fields in the text data to be processed and attributes in the data element standard specification, developers can be freed from the heavy data governance work and the semantic alignment efficiency for data elements is improved; on the other hand, by combining local matching with global matching, the accuracy of matching between fields and attributes can be improved. Moreover, compared with the data semantic alignment method based on automatic extraction technology in related technologies, it can reduce the dependence on external data sources and improve the flexibility and applicability of the algorithm process.

[0105] Figure 6 FIG. 1 is a flow chart of a method for semantic alignment of data elements according to other embodiments of the present disclosure. Figure 6 As shown, the semantic alignment method for data elements includes three steps: weight calculation, local node matching, and global sequence screening.

[0106] Step 1: Weight calculation. In some embodiments, the data element standard specification is specifically a city knowledge system. Figure 7 The process shown calculates the importance assessment value of each data element in the data element standard specification.

[0107] First, if Figure 7 As shown, the urban knowledge system is treated as a complete article. Also, the contextual information is appended to the data elements such as entities, relationships, attributes, and attribute terms in the urban knowledge system, and they are treated as paragraphs. For example, for the attribute term "former name", the result after appending the context is "person-name-former name". For example, for the relationship term "home address", the result after appending the context is "person-place-permanent residence-home address".

[0108] Then, a weight model (such as the TF-IWF algorithm) is used to calculate the importance evaluation value of each data element in the urban knowledge system.

[0109] For example, for the relational term "people-land-permanent residence-rental address" after the context is appended, assuming that it appears 10 times in the urban knowledge system, then TF = 10; further, assuming that there are a total of 1000 paragraphs, and 8 paragraphs contain "people-land-permanent residence-rental address", then IWF = log(1000 / 8) = 2.10; then, the product of TF and IWF 21.0 is used as the importance evaluation value of "people-land-permanent residence-rental address".

[0110] Back to Figure 6 Step 2: Local node matching. This step includes five sub-steps: embedding processing, attribute matching, entity relationship matching, calculating the first similarity, and selecting candidate attributes.

[0111] Firstly, all data elements in the urban knowledge system and input form fields are embedded respectively using the pre-trained text representation language model to obtain the embedded representation vectors of each data element in the urban knowledge system and the embedded representation vectors of the form fields.

[0112] Specifically, for form fields, the form fields can be preprocessed first, and then the pre-trained model can be used to embed the preprocessed form fields to obtain the embedding representation vector of the form fields; for data elements in the urban knowledge system, the data elements and their context can be taken as a whole to calculate the average embedding vector. The embedding process for data elements includes step a and step b.

[0113] Step a, pre-processing the data elements after the text is added, such as word segmentation, stop word removal, and standardization. For example, for the data elements "person-place-permanent residence-rental address" after the text is added, the word list ["person", "place", "permanent residence", "rental address"] is obtained through word segmentation.

[0114] Step b, use the pre-trained model to embed each word in the word list, and then take the average of the embedding vectors of each word in each dimension to obtain the average embedding vector. For example, for the data elements "person-place-permanent residence-rental address" appended with the text, the embedding vector of "person", the embedding vector of "place", the embedding vector of "permanent residence", and the embedding vector of "rental address" are averaged in each dimension to obtain the average embedding vector.

[0115] Then, attribute matching is performed. In this step, the average embedding vector of the attribute (i.e., the first embedding vector) and the average embedding vector of the attribute term corresponding to the attribute (i.e., the second embedding vector) are first averaged in each dimension to obtain the embedding representation vector of the attribute. For example, assuming that the attribute "name" of a person contains the three attribute terms "name", "nickname", and "former name", the average embedding vector of "name", the average embedding vector of "nickname", and the average embedding vector of "former name" are averaged in each dimension, and the result of the operation is used as the embedding representation vector of the attribute "name". Then, the cosine similarity between the embedding representation vector of the field and the embedding representation vector of the attribute is calculated and used as the first vector similarity.

[0116] Next, entity relationship matching is performed. In this step, if the attribute is a relationship attribute, the average embedding vector of the attribute relationship and the average embedding vector of the relationship term corresponding to the attribute relationship are averaged in each dimension to obtain the embedding representation vector of the attribute relationship. Then, the cosine similarity between the embedding representation vector of the field and the embedding representation vector of the attribute relationship is calculated and used as the second vector similarity. If the attribute is an entity attribute, similar processing is performed on the average embedding vector of the entity to which the attribute belongs and the average embedding vector of the entity classification corresponding to the entity to which the attribute belongs to obtain the second vector similarity.

[0117] Then, the first similarity is calculated. In this step, the first vector similarity between the embedded representation vector of the field and the embedded representation vector of the attribute is multiplied by the importance weight of the attribute to obtain a first weighted similarity; the second vector similarity between the embedded representation vector of the field and the embedded representation vector of the entity or relationship to which the attribute belongs is multiplied by the importance weight of the entity or relationship to obtain a second weighted similarity; the first weighted similarity is multiplied by the second weighted similarity to obtain the first similarity. Among them, the importance weight of the attribute is obtained by averaging the importance evaluation value of the attribute and the importance evaluation value of the attribute term corresponding to the attribute. The importance weight of the entity and the importance weight of the relationship can also be obtained in a similar manner.

[0118] Finally, select candidate attributes. In this step, the first N attributes with the largest first similarity to each field can be selected from all attributes in the urban knowledge system and used as candidate attributes for the field. Where N is an integer greater than 1 and less than the total number of attributes in the urban knowledge system.

[0119] Step 3: Global sequence screening: This step includes three sub-steps: constructing a candidate attribute sequence, calculating the second similarity, and selecting a matching attribute sequence.

[0120] First, the candidate attributes corresponding to the various form fields in the form field sequence are arranged and combined to obtain a plurality of candidate attribute sequences.

[0121] Then, a structural representation graph of each candidate attribute sequence is constructed; the connectivity of the structural representation graph is determined; and the product of the first similarities corresponding to all candidate attributes in the candidate attribute sequence is multiplied by the connectivity of the structural representation graph of the candidate attribute sequence to obtain a second similarity between the candidate attribute sequence and the form field sequence.

[0122] Finally, the candidate attribute sequence with the second largest similarity among the multiple candidate attribute sequences is used as the attribute sequence that matches the form field sequence.

[0123] In the disclosed embodiments, the accuracy and efficiency of data semantic alignment can be improved through the above process. Through the above processing, the foundation is laid for cross-departmental and cross-system data sharing and collaboration. For example, in the field of government services, by processing data into a standard data format, the process of citizens' affairs can be greatly simplified, and the effect of one-time filling and multiple reuse can be achieved, which significantly improves administrative efficiency and service quality. In enterprise management, by processing data into a standard data format, it helps to break down information silos, promote collaborative work between departments, and improve the accuracy and timeliness of decision-making.

[0124] Figure 8 FIG. 1 is a flow chart of a data query method according to some embodiments of the present disclosure. Figure 8 As shown, the data query method includes steps S81 to S84.

[0125] In step S81, an attribute sequence matching the field sequence is determined, and the correspondence between the fields and the matching attributes is stored.

[0126] In some examples, a sequence of attributes that matches a field is determined according to a semantic alignment method for data elements as previously described.

[0127] In some examples, the fields in the field sequence are taken as a column, and the attributes matching them are taken as a column, which are stored in the data table accordingly. At the same time, the data table also includes the field values ​​corresponding to the fields in the field sequence. For example, data table 1 stores the field "name" in the form, the attribute "name" in the data element standard specification, and the value "Zhang San" of the form field. Data table 2 stores the field "name" in the form, the attribute "name" in the data element standard specification, and the value "Li Si" of the form field.

[0128] In the embodiment of the present disclosure, the above correspondence is stored to facilitate subsequent data query.

[0129] In step S82, in response to the data query request, the attributes to be queried in the data element standard specification involved in the request are determined.

[0130] For example, by parsing the data query request, the attributes to be queried in the data element standard specification are determined.

[0131] In step S83, a field matching the attribute to be queried is determined according to the attribute to be queried and the corresponding relationship.

[0132] For example, assuming that the attribute to be queried is "name", by querying the above correspondence, it is determined that the fields that match it include "name" and "name".

[0133] In step S84, data query is performed according to the fields matching the attributes to be queried to obtain query results.

[0134] In some examples, a structured query statement is generated based on a field that matches an attribute to be queried, and a data table is queried based on the structured query statement to obtain a query result.

[0135] For example, suppose that data table 1 contains a field "name" and its field value; suppose that data table 2 contains a field "name" and its field value. By generating a structured query statement to query the "name" field in table 1 and the "name" field in table 2, the query result is obtained.

[0136] In the disclosed embodiment, the above process can be used to query data from different sources and in different formats, thereby improving the efficiency, portability and accuracy of data query.

[0137] Fig. 9 is a schematic diagram of a structure of a semantic alignment device for data elements according to some embodiments of the present disclosure. Fig. 9 As shown, the semantic alignment device 90 for data elements includes an extraction module 91, a first determination module 92, a first selection module 93, a construction module 94, a second determination module 95 and a second selection module 96.

[0138] The extraction module 91 is configured to extract a field sequence from the text data to be processed, wherein the field sequence includes a plurality of fields.

[0139] The first determination module 92 is configured to determine a first similarity between each field in the plurality of fields and each attribute in the plurality of attributes included in the data element standard specification.

[0140] The first selection module 93 is configured to select a candidate attribute corresponding to each field from a plurality of attributes according to the first similarity.

[0141] The construction module 94 is configured to construct a plurality of candidate attribute sequences corresponding to the field sequence according to the candidate attributes corresponding to the plurality of fields.

[0142] The second determination module 95 is configured to determine a second similarity between the field sequence and each candidate attribute sequence in the plurality of candidate attribute sequences.

[0143] The second selection module 96 is configured to select an attribute sequence matching the field sequence from a plurality of candidate attribute sequences according to the second similarity.

[0144] In the embodiments of the present disclosure, the above device can improve the efficiency and accuracy of semantic alignment processing for data elements.

[0145] Fig.10 is a schematic diagram of the structure of a data query device according to some embodiments of the present disclosure. Fig.10 As shown, the data query device 100 includes a storage module 101 , a first determination module 102 , a second determination module 103 , and a query module 104 .

[0146] The storage module 101 is configured to determine the attribute sequence that matches the field sequence according to the semantic alignment method for data elements as described above, and store the correspondence between the fields in the field sequence and the attributes in the attribute sequence.

[0147] The first determination module 102 is configured to determine, in response to a data query request, a to-be-queried attribute in a data element standard specification involved in the request.

[0148] The second determination module 103 is configured to determine a field matching the attribute to be queried according to the attribute to be queried and the above-mentioned corresponding relationship.

[0149] The query module 104 is configured to perform a data query according to the fields matching the attributes to be queried to obtain a query result.

[0150] In the embodiments of the present disclosure, the above devices can improve the efficiency, convenience and accuracy of data query.

[0151] Fig.11 is a schematic diagram of the structure of an electronic device according to some embodiments of the present disclosure. Fig.11 As shown, the electronic device 110 includes a memory 111 and a processor 112 coupled to the memory 111. The memory 111 is used to store instructions for executing the semantic alignment method for data elements or the data query method corresponding to the embodiment. The processor 112 is configured to execute the semantic alignment method for data elements or the data query method in any of the embodiments of the present disclosure based on the instructions stored in the memory 111.

[0152] Fig.12 Schematic diagram of the structure of a computer system according to some embodiments of the present disclosure. Fig.12 As shown, the semantic alignment device or data query device for data elements can be embodied in the form of a general-purpose computing device. The computer system 120 includes a memory 121, a processor 122, and a bus 123 connecting different system components.

[0153] The memory 121 may include, for example, a system memory, a non-volatile storage medium, etc. The system memory may store, for example, an operating system, an application, a boot loader, and other programs. The system memory may include a volatile storage medium, such as a random access memory (RAM) and / or a cache memory. The non-volatile storage medium may store, for example, instructions for executing at least one of the corresponding embodiments of the semantic alignment method or the data query method for data elements. The non-volatile storage medium includes, but is not limited to, a disk memory, an optical memory, a flash memory, and the like.

[0154] The processor 122 can be implemented by a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistors, etc. Discrete hardware components. Accordingly, each module in the modules such as the extraction module and the first determination module can be implemented by a central processing unit (CPU) running instructions in a memory to execute corresponding steps, or can be implemented by a dedicated circuit to execute corresponding steps.

[0155] The bus 123 may use any of a variety of bus structures, including, but not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MCA) bus, and a Peripheral Component Interconnect (PCI) bus.

[0156] The interfaces 124, 125, 126 in the computer system 120 and the memory 121 and the processor 122 can be connected via a bus 123. The input / output interface 124 provides a connection interface for input / output devices such as a display, a mouse, and a keyboard. The network interface 125 provides a connection interface for various networked devices. The storage interface 126 provides a connection interface for external storage devices such as a floppy disk, a USB flash drive, and an SD card.

[0157] Here, various aspects of the present disclosure are described with reference to flowcharts and / or block diagrams of methods, devices, and computer program products according to embodiments of the present disclosure. It should be understood that each frame of the flowchart and / or block diagram and the combination of frames can be implemented by computer-readable program instructions.

[0158] These computer-readable program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable device to produce a machine, so that the processor executes the instructions to produce means for implementing the functions specified in one or more blocks in the flowchart and / or block diagram.

[0159] These computer-readable program instructions may also be stored in a computer-readable memory, which cause the computer to work in a specific manner to produce an article of manufacture, including instructions for implementing the functions specified in one or more blocks in the flowchart and / or block diagram.

[0160] The present disclosure can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects.

[0161] The semantic alignment and data query method and device for data elements in the above-mentioned embodiments can improve the efficiency and accuracy of data semantic alignment.

[0162] So far, the semantic alignment and data query method and device for data elements according to the present disclosure have been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.

Claims

1. A method for semantic alignment of data elements, performed by an electronic device, comprising: Extracting a field sequence from the text data to be processed, wherein the field sequence includes a plurality of fields; Determining a first similarity between each of the plurality of fields and each of the plurality of attributes included in the data element standard specification; According to the first similarity, selecting a candidate attribute corresponding to each field from the multiple attributes; Constructing a plurality of candidate attribute sequences corresponding to the field sequence according to the candidate attributes corresponding to the plurality of fields; determining a second similarity between the field sequence and each candidate attribute sequence of the plurality of candidate attribute sequences; According to the second similarity, an attribute sequence matching the field sequence is selected from the multiple candidate attribute sequences.

2. The method for semantic alignment of data elements according to claim 1, wherein: The data element standard specification further includes entities and relationships, and determining a first similarity between each of the plurality of fields and each of the plurality of attributes included in the data element standard specification includes: Using a text representation language model, embedding each field to obtain an embedding representation vector for each field; Determine a first vector similarity between the embedded representation vector of each field and the embedded representation vector of each attribute; Determine a second vector similarity between the embedding representation vector of each field and the embedding representation vector of the entity or relationship to which each attribute belongs; The first similarity is determined according to the first vector similarity and the second vector similarity.

3. The semantic alignment method for data elements according to claim 1, wherein: The data element standard specification further includes entities and relationships, and determining the second similarity between the field sequence and each candidate attribute sequence in the plurality of candidate attribute sequences includes: Constructing a structural representation graph of each candidate attribute sequence; Determining the connectivity of the structure representation graph; The second similarity between the field sequence and each candidate attribute sequence is determined according to the connectivity of the structure representation graph and the first similarities of all candidate attributes in each candidate attribute sequence.

4. The method for semantic alignment of data elements according to claim 2, wherein: The determining the first similarity according to the first vector similarity and the second vector similarity comprises: Determining a first weighted similarity according to the first vector similarity and the importance weight of each attribute in the data element standard specification; Determine a second weighted similarity according to the second vector similarity and the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification; The first similarity is determined according to the first weighted similarity and the second weighted similarity.

5. The method for semantic alignment of data elements according to claim 4, wherein: The importance weight of each attribute in the data element standard specification is determined as follows: Determine the importance evaluation value of each attribute and the importance evaluation value of the attribute term corresponding to each attribute by using an importance evaluation algorithm; The importance weight of each attribute in the data element standard specification is determined according to the importance evaluation value of each attribute and the average value of the importance evaluation value of the attribute term corresponding to each attribute.

6. The method for semantic alignment of data elements according to claim 4, wherein: The importance weight of the entity or relationship to which each attribute belongs in the data element standard specification is determined as follows: Determine the importance evaluation value of the entity or relationship to which each attribute belongs, and the importance evaluation value of the term corresponding to the entity or relationship to which each attribute belongs, using an importance evaluation algorithm; The importance weight of the entity or relationship to which each attribute belongs in the data element standard specification is determined based on the average of the importance evaluation value of the entity or relationship to which each attribute belongs and the importance evaluation value of the term corresponding to the entity or relationship to which each attribute belongs.

7. The method for semantic alignment of data elements according to claim 5 or 6, wherein: The importance evaluation algorithm is a term frequency-inverse document frequency algorithm.

8. The method for semantic alignment of data elements according to claim 4, wherein: The first weighted similarity is positively correlated with the first vector similarity and the importance weight of each attribute in the data element standard specification; and / or The second weighted similarity is positively correlated with the second vector similarity and the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification; and / or The first similarity is positively correlated with the first weighted similarity and the second weighted similarity.

9. The method for semantic alignment of data elements according to claim 8, wherein: The first weighted similarity is determined by multiplying the first vector similarity by the importance weight of each attribute in the data element standard specification; and / or, the second weighted similarity is determined by multiplying the second vector similarity by the importance weight of the entity or relationship to which each attribute belongs in the data element standard specification; and / or, the first similarity is determined by multiplying the first weighted similarity by the second weighted similarity.

10. The method for semantic alignment of data elements according to claim 2, wherein: The embedding representation vector of each attribute is determined according to the following method: Using a text representation language model, embedding each attribute and the attribute term corresponding to each attribute respectively to obtain a first embedding vector for each attribute and a second embedding vector for the attribute term corresponding to each attribute; An embedding representation vector of each attribute is determined according to the first embedding vector of each attribute and the second embedding vector of the attribute term corresponding to each attribute.

11. The method for semantic alignment of data elements according to claim 3, wherein: The second similarity is positively correlated with the connectivity of the structure representation graph and the first similarities corresponding to all candidate attributes in each candidate attribute sequence.

12. The method for semantic alignment of data elements according to claim 11, wherein: Determining the second similarity between the field sequence and each candidate attribute sequence according to the connectivity of the structure representation graph and the first similarities corresponding to all candidate attributes in each candidate attribute sequence includes: Determining the comprehensive attribute similarity according to the sum of the first similarities corresponding to all candidate attributes in each candidate attribute sequence; The second similarity between the field sequence and each of the candidate attribute sequences is determined according to the product of the comprehensive attribute similarity and the connectivity of the structure representation graph.

13. The method for semantic alignment of data elements according to claim 3, wherein: The constructing of the structural representation graph of each candidate attribute sequence comprises: For each candidate attribute in the candidate attribute sequence, if the candidate attribute belongs to an attribute of an entity, a node is set; if the candidate attribute belongs to an attribute of a relationship, an edge is set; the setting step is iteratively performed to obtain the structural representation graph.

14. The method for semantic alignment of data elements according to claim 1, wherein: The selecting, according to the first similarity, a candidate attribute corresponding to each field from the multiple attributes comprises: N attributes among the multiple attributes having the largest first similarity with each field are used as candidate attributes corresponding to each field, where N is an integer value greater than 1 and less than M, and M is the total number of the multiple attributes.

15. The method for semantic alignment of data elements according to claim 1, wherein: The step of constructing a plurality of candidate attribute sequences corresponding to the field sequence according to the candidate attributes corresponding to the plurality of fields comprises: The candidate attributes corresponding to the multiple fields are arranged and combined to obtain the multiple candidate attribute sequences.

16. A data query method, comprising: According to any one of claims 1 to 15, the semantic alignment method for data elements is used to determine an attribute sequence that matches the field sequence, and store the corresponding relationship between the fields in the field sequence and the attributes in the attribute sequence; In response to a data query request, determining a to-be-queried attribute in the data element standard specification involved in the request; Determine, according to the attribute to be queried and the corresponding relationship, a field matching the attribute to be queried; Data query is performed according to the fields matching the attributes to be queried to obtain query results.

17. An electronic device comprising: Memory; as well as A processor coupled to the memory, the processor being configured to execute the semantic alignment method for data elements as described in any one of claims 1 to 15, or the data query method as described in claim 16, based on instructions stored in the memory.

18. A computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the semantic alignment method for data elements as described in any one of claims 1 to 15, or the data query method as described in claim 16.

19. A computer program product having computer program instructions stored thereon, which, when executed by a processor, implement the semantic alignment method for data elements as described in any one of claims 1 to 15, or the data query method as described in claim 16.

Images