Knowledge graph-based audit evaluation information generation method and device and electronic equipment

Abstract

Embodiments of the present disclosure disclose a knowledge graph-based audit evaluation information generation method, device and electronic equipment. A specific implementation of the method includes: obtaining a target audit report text; performing text cleaning on the target audit report text to generate text information; generating a target knowledge graph corresponding to the text information; determining a target knowledge graph corresponding to an isomorphic knowledge graph, obtaining a set of isomorphic knowledge graphs; generating audit evaluation information corresponding to the target audit report text according to the target knowledge graph, the set of isomorphic knowledge graphs and a pre-trained quality evaluation model; and sending the audit evaluation information to a target display terminal for display. This implementation improves audit efficiency and the accuracy of the generated audit evaluation information.

Description

Technical Field

[0001] Embodiments of this disclosure relate to the field of computer technology, and more specifically to a method, apparatus, and electronic device for generating audit assessment information based on knowledge graphs. Background Technology

[0002] Auditing refers to a technical means of reviewing audit reports in accordance with auditing rules. Currently, the common method for auditing audit reports is to manually audit the reports to generate corresponding audit assessment information.

[0003] However, when using the above method, the following technical problems often arise:

[0004] First, with the increase in the number of audit reports, it is difficult to audit them in a timely manner using manual methods, resulting in low audit efficiency;

[0005] Second, manual auditing relies heavily on human experience, which makes it difficult to guarantee the accuracy of the generated audit assessment information. Summary of the Invention

[0006] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this disclosure propose a method, apparatus, and electronic device for generating audit assessment information based on knowledge graphs to address one or more of the technical problems mentioned in the background section above.

[0008] In a first aspect, some embodiments of this disclosure provide a method for generating audit assessment information based on a knowledge graph. The method includes: acquiring a target audit report text; performing text cleaning on the target audit report text to generate text information; generating a target knowledge graph corresponding to the text information; determining a homogeneous knowledge graph corresponding to the target knowledge graph to obtain a set of homogeneous knowledge graphs; generating audit assessment information corresponding to the target audit report text based on the target knowledge graph, the set of homogeneous knowledge graphs, and a pre-trained quality assessment model; and sending the audit assessment information to a target display terminal for display.

[0009] Optionally, the above-mentioned text cleaning of the target audit report text to generate text information includes: reading the target audit report text to generate candidate text information; performing entity recognition on the candidate text information to generate a first entity information set, wherein the first entity information in the first entity information set includes: entity location information and entity information; determining the entity similarity between the entity information included in the first entity information and each entity information in the target entity library to generate a similarity value; filtering out entity information whose corresponding similarity value is within the target interval from the target entity library as candidate entity information to obtain a candidate entity information set; performing entity replacement on the first entity information included in the candidate text information according to the entity location information included in the first entity information and the target candidate entity information, wherein the target candidate entity information is the candidate entity information in the candidate entity information set that meets the first screening condition; and determining the candidate text information after entity replacement as the above-mentioned text information.

[0010] Optionally, generating the target knowledge graph corresponding to the aforementioned text information includes: performing entity recognition on the aforementioned text information to generate a second entity information set; removing entities from the second entity information in the aforementioned second entity information set to generate removed entity information, thus obtaining a removed entity information set; determining the relation information corresponding to each entity information in the removed entity information set to obtain a relation information group set; and generating the aforementioned target knowledge graph based on the removed entity information set and the relation information group set.

[0011] Optionally, determining the isomorphic knowledge graph corresponding to the target knowledge graph to obtain a set of isomorphic knowledge graphs includes: selecting knowledge graphs that meet the second selection criteria from the knowledge graph database as candidate knowledge graphs to obtain a set of candidate knowledge graphs, wherein the second selection criteria are that the number of entity nodes contained in the knowledge graph is greater than or equal to the number of entity nodes contained in the target knowledge graph, and the number of relation edges contained in the knowledge graph is greater than or equal to the number of relation edges contained in the target knowledge graph; performing the following processing steps on each entity node in the target knowledge graph: constructing feature vectors for the entity nodes. The process involves generating a first entity node feature vector; constructing feature vectors for at least one relation edge connected to the entity node to generate a first relation edge feature vector; for each candidate knowledge graph in the candidate knowledge graph set, the following steps are performed: constructing feature vectors for each entity node in the candidate knowledge graph to generate a second entity node feature vector, resulting in a second entity node feature vector sequence; constructing feature vectors for each target relation edge group in the candidate knowledge graph to generate a second relation edge feature vector sequence, wherein the target relation edge group is at least one relation edge connected to the entity node of the candidate knowledge graph; determining the feature similarity between each first entity node feature vector in the obtained first entity node feature vector sequence and each second entity node feature vector in the obtained second entity node feature vector sequence, resulting in an entity node similarity value group sequence; determining the feature similarity between each first relation edge feature vector in the obtained first relation edge feature vector sequence and each second relation edge feature vector in the obtained second relation edge feature vector sequence, resulting in a relation edge similarity value group sequence; and selecting the entity node with the largest value from each entity node similarity value group in the above entity node similarity value group sequence. The similarity value is used as the first candidate value to obtain the first candidate value set. From each relation edge similarity value group in the above relation edge similarity value group sequence, the relation edge similarity value with the largest value is selected as the second candidate value to obtain the second candidate value set. The first candidate value in the first candidate value set and the second candidate value in the second candidate value set are weighted and summed to generate the graph similarity value corresponding to the above candidate knowledge graph. From the above candidate knowledge graph set, the candidate knowledge graph whose graph similarity value satisfies the third screening condition is selected as the isomorphic knowledge graph to obtain the above isomorphic knowledge graph set.

[0012] Optionally, the quality assessment model includes: a feature extraction model set, a first classification model, a feature fusion model, and a second classification model; and the generation of audit assessment information corresponding to the target audit report text based on the target knowledge graph, the isomorphic knowledge graph set, and the pre-trained quality assessment model includes: inputting the target knowledge graph and the isomorphic knowledge graph in the isomorphic knowledge graph set into the feature extraction model in the feature extraction model set in parallel to generate a first feature graph feature information and a second feature graph feature information set; inputting the first feature graph feature information and the second feature graph feature information set into the first classification model to generate a classification result; inputting the first feature graph feature information and at least one second feature graph feature information corresponding to the classification result into the feature fusion model to generate a third feature graph feature information; and inputting the third feature graph feature information into the second classification model to generate the audit assessment information.

[0013] Optionally, before sending the audit assessment information to the target display terminal for display, the method further includes: sending a communication request to a communication terminal connected to the target display terminal; in response to receiving the communication connection request sent by the communication terminal, establishing an end-to-end communication link with the communication terminal; and in response to determining that the communication link has been successfully established, transmitting the encrypted text corresponding to the audit assessment information to the communication terminal through the communication link.

[0014] Optionally, before transmitting the encrypted text corresponding to the audit assessment information to the communication terminal via the communication link, the method further includes: constructing an encryption key based on the communication address and sending timestamp of the communication terminal; and encrypting the audit assessment information using the encryption key to generate the encrypted text.

[0015] Secondly, some embodiments of this disclosure provide an audit assessment information generation apparatus based on a knowledge graph. The apparatus includes: an acquisition unit configured to acquire a target audit report text; a text cleaning unit configured to perform text cleaning on the target audit report text to generate text information; a first generation unit configured to generate a target knowledge graph corresponding to the text information; a determination unit configured to determine a homogeneous knowledge graph corresponding to the target knowledge graph to obtain a set of homogeneous knowledge graphs; a second generation unit configured to generate audit assessment information corresponding to the target audit report text based on the target knowledge graph, the set of homogeneous knowledge graphs, and a pre-trained quality assessment model; and a display unit configured to send the audit assessment information to a target display terminal for display.

[0016] Optionally, the text cleaning unit is further configured to: read the target audit report text to generate candidate text information; perform entity recognition on the candidate text information to generate a first entity information set, wherein the first entity information in the first entity information set includes: entity location information and entity information; determine the entity similarity between the entity information included in the first entity information and each entity information in the target entity library to generate a similarity value; filter out entity information whose corresponding similarity value is within the target interval from the target entity library as candidate entity information to obtain a candidate entity information set; perform entity replacement on the first entity information included in the candidate text information according to the entity location information included in the first entity information and the target candidate entity information, wherein the target candidate entity information is the candidate entity information in the candidate entity information set that meets the first screening condition; and determine the candidate text information after entity replacement as the text information.

[0017] Optionally, the first generation unit is further configured to: perform entity recognition on the above text information to generate a second entity information set; remove entities from the second entity information in the above second entity information set to generate removed entity information, thus obtaining a removed entity information set; determine the relation information corresponding to each entity information in the removed entity information set to obtain a relation information group set; and generate the above target knowledge graph based on the removed entity information set and the relation information group set.

[0018] Optionally, the determining unit is further configured to: select knowledge graphs that meet the second selection criteria from the knowledge graph library as candidate knowledge graphs, thereby obtaining a set of candidate knowledge graphs, wherein the second selection criteria are that the number of entity nodes contained in the knowledge graph is greater than or equal to the number of entity nodes contained in the target knowledge graph, and the number of relation edges contained in the knowledge graph is greater than or equal to the number of relation edges contained in the target knowledge graph; perform the following processing steps on each entity node in the target knowledge graph: construct feature vectors for the entity node to generate a first entity node feature vector; construct feature vectors for at least one relation edge connected to the entity node to generate a first relation edge feature vector; for each candidate knowledge graph in the set of candidate knowledge graphs, perform the following processing steps: construct feature vectors for each entity node in the candidate knowledge graph to generate a second entity node feature vector, thereby obtaining a sequence of second entity node feature vectors; construct feature vectors for each target relation edge group in the candidate knowledge graph to generate a second relation edge feature vector sequence, wherein the target relation edge group is at least one relation edge connected to the entity node of the candidate knowledge graph; The feature similarity between each first entity node feature vector in the obtained first entity node feature vector sequence and each second entity node feature vector in the obtained second entity node feature vector sequence is determined to obtain an entity node similarity value group sequence. The feature similarity between each first relation edge feature vector in the obtained first relation edge feature vector sequence and each second relation edge feature vector in the obtained second relation edge feature vector sequence is determined to obtain a relation edge similarity value group sequence. The entity node similarity value with the largest value is selected from each entity node similarity value group in the above entity node similarity value group sequence as the first candidate value, resulting in a first candidate value set. The relation edge similarity value with the largest value is selected from each relation edge similarity value group in the above relation edge similarity value group sequence as the second candidate value, resulting in a second candidate value set. A weighted sum is performed on the first candidate value in the above first candidate value set and the second candidate value in the above second candidate value set to generate the graph similarity value corresponding to the above candidate knowledge graph. Candidate knowledge graphs whose graph similarity values ​​satisfy the third selection condition are selected from the above candidate knowledge graph set as isomorphic knowledge graphs, resulting in the above isomorphic knowledge graph set. In practice, the knowledge graphs corresponding to different audit report texts often contain some different feature information. Therefore, by determining the set of isomorphic knowledge graphs corresponding to the target knowledge graph, the richness of features can be increased, making the generated audit assessment information more accurate. Thus, determining the set of isomorphic knowledge graphs corresponding to the target knowledge graph is particularly important. First, isomorphic knowledge graphs should contain similar numbers of relation edges and entity nodes. Therefore, this disclosure uses a second screening condition to perform a coarse screening of the knowledge graphs to filter out useless knowledge graphs.Furthermore, isomorphic knowledge graphs should have similar graph structures. Therefore, this disclosure determines isomorphic knowledge graphs by constructing feature vectors for entity nodes and the relational edges connected to them, and calculating similarity based on the constructed feature vectors. This method significantly improves the accuracy of the obtained isomorphic knowledge graphs, thereby improving the accuracy of the generated audit evaluation information.

[0019] Optionally, the quality assessment model includes: a set of feature extraction models, a first classification model, a feature fusion model, and a second classification model; and the second generation unit is further configured to: input the target knowledge graph and the isomorphic knowledge graph in the set of isomorphic knowledge graphs in parallel into the feature extraction models in the set of feature extraction models to generate a first feature graph feature information and a second set of feature graph feature information; input the first feature graph feature information and the second set of feature graph feature information into the first classification model to generate a classification result; input the first feature graph feature information and at least one second set of feature graph feature information corresponding to the classification result into the feature fusion model to generate a third set of feature graph feature information; and input the third set of feature graph feature information into the second classification model to generate the audit assessment information.

[0020] Optionally, before sending the audit assessment information to the target display terminal for display, the apparatus further includes: sending a communication request to a communication terminal connected to the target display terminal; in response to receiving the communication connection request sent by the communication terminal, establishing an end-to-end communication link with the communication terminal; and in response to determining that the communication link has been successfully established, transmitting the encrypted text corresponding to the audit assessment information to the communication terminal through the communication link.

[0021] Optionally, before transmitting the encrypted text corresponding to the audit assessment information to the communication terminal via the communication link, the device further includes: constructing an encryption key based on the communication address and sending timestamp of the communication terminal; and encrypting the audit assessment information using the encryption key to generate the encrypted text.

[0022] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0023] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0024] The above embodiments of this disclosure have the following beneficial effects: The knowledge graph-based audit assessment information generation method of some embodiments of this disclosure greatly improves audit efficiency while ensuring the accuracy of the generated audit assessment information. Specifically, the reasons for low audit efficiency and the inability to guarantee the accuracy of the generated audit assessment information are: First, with the increase in the number of audit reports, it is difficult to audit the audit reports in a timely manner using manual methods, leading to low audit efficiency. Second, manual auditing relies heavily on human experience, making it impossible to guarantee the accuracy of the generated audit assessment information. Based on this, the knowledge graph-based audit assessment information generation method of some embodiments of this disclosure first obtains the target audit report text. Second, it performs text cleaning on the target audit report text to generate text information. Through text cleaning, text content irrelevant to entity extraction and relationship extraction is removed. Next, a target knowledge graph corresponding to the above text information is generated. The entities contained in the target audit report text and the relationships between entities are extracted from this graph. Then, the isomorphic knowledge graph corresponding to the target knowledge graph is determined, resulting in a set of isomorphic knowledge graphs. By obtaining the isomorphic knowledge graphs, knowledge graphs related to the target knowledge graph are determined. Furthermore, based on the aforementioned target knowledge graph, the aforementioned isomorphic knowledge graph set, and the pre-trained quality assessment model, audit assessment information corresponding to the aforementioned target audit report text is generated. The isomorphic knowledge graph set enriches the graph features of the target knowledge graph, thereby improving the accuracy of the generated audit assessment information. Finally, the aforementioned audit assessment information is sent to the target display terminal for display. This method solves the problems associated with manual auditing, significantly improving audit efficiency and the accuracy of the generated audit assessment information. Attached Figure Description

[0025] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0026] Figure 1 This is a schematic diagram illustrating an application scenario of the knowledge graph-based audit assessment information generation method according to some embodiments of this disclosure;

[0027] Figure 2 This is a flowchart of some embodiments of the knowledge graph-based audit assessment information generation method according to this disclosure;

[0028] Figure 3 This is a flowchart of some other embodiments of the knowledge graph-based audit assessment information generation method according to this disclosure;

[0029] Figure 4 This is a schematic diagram of the structure of some embodiments of the knowledge graph-based audit assessment information generation apparatus according to the present disclosure;

[0030] Figure 5 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0031] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0032] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0033] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0034] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0035] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0036] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0037] Figure 1 This is a schematic diagram illustrating an application scenario of the knowledge graph-based audit assessment information generation method according to some embodiments of this disclosure.

[0038] exist Figure 1In the application scenario, firstly, the computing device 101 can acquire the target audit report text 102; then, the computing device 101 can perform text cleaning on the target audit report text 102 to generate text information 103; next, the computing device 101 can generate the target knowledge graph 104 corresponding to the text information 103; further, the computing device 101 can determine the isomorphic knowledge graph corresponding to the target knowledge graph 104 to obtain the isomorphic knowledge graph set 105; in addition, the computing device 101 can generate audit evaluation information 107 corresponding to the target audit report text 102 based on the target knowledge graph 104, the isomorphic knowledge graph set 105, and the pre-trained quality evaluation model 106; finally, the computing device 101 can send the audit evaluation information 107 to the target display terminal 108 for display.

[0039] It should be noted that the aforementioned computing device 101 can be either hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device. When the computing device is software, it can be installed within the hardware devices listed above. It can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are made here.

[0040] It should be understood that Figure 1 The number of computing devices shown is merely illustrative. Any number of computing devices can be used depending on implementation needs.

[0041] Continue to refer to Figure 2 The diagram illustrates a flow 200 of some embodiments of the knowledge graph-based audit assessment information generation method according to this disclosure. The knowledge graph-based audit assessment information generation method includes the following steps:

[0042] Step 201: Obtain the target audit report text.

[0043] In some embodiments, the entity executing the knowledge graph-based audit assessment information generation method (e.g.) Figure 1 The computing device 101 shown can obtain the aforementioned target audit report text via a wired or wireless connection. The aforementioned target audit report text may be the audit report text to be audited and evaluated.

[0044] Step 202: Clean the target audit report text to generate text information.

[0045] In some embodiments, the executing entity may perform text cleaning on the target audit report text to generate text information. For example, the executing entity may remove information from the audit report text other than entities and entity correspondences to generate the text information. Alternatively, the executing entity may remove punctuation marks, tabs, and other symbols from the audit report text to generate the text information.

[0046] Step 203: Generate the target knowledge graph corresponding to the text information.

[0047] In some embodiments, the executing entity can generate a target knowledge graph corresponding to the text information. This target knowledge graph is a graph constructed based on the entities contained in the text information and the corresponding relationships between those entities. The data structure of the target knowledge graph is a graph structure.

[0048] As an example, firstly, the aforementioned execution entity can extract entities and relationships from the aforementioned text information. Then, based on the extracted entities and their corresponding relationships, the execution entity can construct the aforementioned target knowledge graph, with entities as nodes and their corresponding relationships as edges.

[0049] Step 204: Determine the isomorphic knowledge graph corresponding to the target knowledge graph to obtain a set of isomorphic knowledge graphs.

[0050] In some embodiments, the execution entity can determine the isomorphic knowledge graph corresponding to the target knowledge graph, thereby obtaining a set of isomorphic knowledge graphs. The isomorphic knowledge graphs in this set are those containing nodes similar to the target knowledge graph and having similar graph structures.

[0051] As an example, the aforementioned execution entity determines the isomorphic knowledge graph corresponding to the target knowledge graph, and obtains a set of isomorphic knowledge graphs, which may include the following steps:

[0052] The first step is to select knowledge graphs that meet the selection criteria from the knowledge graph database as preprocessed knowledge graphs, thus obtaining a set of preprocessed knowledge graphs.

[0053] The graph selection criterion mentioned above is that the knowledge graph contains a target number of target nodes. A target node is a node in the knowledge graph library that is identical to the node in the target knowledge graph. This target number represents the minimum number of identical nodes in the knowledge graph library that are identical to the target knowledge graph. This target number can be determined by the following formula:

[0054] TN = TP × TGN

[0055] Where TN represents the number of targets. TP represents the percentage of targets. TGN represents the number of nodes in the target knowledge graph. The percentage of targets is the ratio of the number of target nodes to the number of nodes in the target knowledge graph.

[0056] The second step involves performing the following sub-processing steps for each preprocessed knowledge graph in the preprocessed knowledge graph set:

[0057] The first sub-step is to select the target node.

[0058] The target node is a node that is included in both the preprocessed knowledge graph and the target knowledge graph.

[0059] The second sub-step involves performing tree structure transformation on the preprocessed knowledge graph and the target knowledge graph, respectively, using the target node as the root node, to generate the first multi-branch tree and the second multi-branch tree.

[0060] The first multi-way tree mentioned above is the tree corresponding to the preprocessed knowledge graph. The second multi-way tree mentioned above is the tree corresponding to the target knowledge graph.

[0061] The third sub-step is to determine whether the first multi-way tree and the second multi-way tree are similar.

[0062] The fourth sub-step, in response to determining that the first multi-way tree and the second multi-way tree are similar, identifies the preprocessed knowledge graph as an isomorphic knowledge graph.

[0063] Step 205: Generate audit assessment information corresponding to the target audit report text based on the target knowledge graph, the isomorphic knowledge graph set, and the pre-trained quality assessment model.

[0064] In some embodiments, the executing entity can generate audit assessment information corresponding to the target audit report text based on the target knowledge graph, a set of isomorphic knowledge graphs, and a pre-trained quality assessment model. The quality assessment model can be a model used to generate the audit assessment information corresponding to the target audit report text. The audit assessment information can be evaluation information characterizing whether the target audit report text conforms to audit rules.

[0065] As an example, firstly, the executing entity can extract features from the target knowledge graph to generate a first feature. Then, the executing entity can extract features from each isomorphic knowledge graph in the isomorphic knowledge graph set to generate a second feature, resulting in a second feature set. The first feature and the second features in the second feature set have the same feature length. Next, the executing entity can superimpose the first feature and each second feature in the second feature set to generate a superimposed feature. Further, the executing entity can divide each value in the superimposed feature by a target value to generate a target feature. The target value is the number of second features in the second feature set plus one. Finally, the executing entity can input the target feature into the quality assessment model to generate the audit assessment information. For example, the quality assessment model could be an LSTM (Long Short Term Memory) model.

[0066] Step 206: Send the audit assessment information to the target display terminal for display.

[0067] In some embodiments, the executing entity may transmit the audit assessment information to the target display terminal for display via a wired or wireless connection. The target display terminal may be a terminal with display functionality. For example, the target display terminal may be a computer, or it may be a handheld terminal with display functionality.

[0068] The above embodiments of this disclosure have the following beneficial effects: The knowledge graph-based audit assessment information generation method of some embodiments of this disclosure greatly improves audit efficiency while ensuring the accuracy of the generated audit assessment information. Specifically, the reasons for low audit efficiency and the inability to guarantee the accuracy of the generated audit assessment information are: First, with the increase in the number of audit reports, it is difficult to audit the audit reports in a timely manner using manual methods, leading to low audit efficiency. Second, manual auditing relies heavily on human experience, making it impossible to guarantee the accuracy of the generated audit assessment information. Based on this, the knowledge graph-based audit assessment information generation method of some embodiments of this disclosure first obtains the target audit report text. Second, it performs text cleaning on the target audit report text to generate text information. Through text cleaning, text content irrelevant to entity extraction and relationship extraction is removed. Next, a target knowledge graph corresponding to the above text information is generated. The entities contained in the target audit report text and the relationships between entities are extracted from this graph. Then, the isomorphic knowledge graph corresponding to the target knowledge graph is determined, resulting in a set of isomorphic knowledge graphs. By obtaining the isomorphic knowledge graphs, knowledge graphs related to the target knowledge graph are determined. Furthermore, based on the aforementioned target knowledge graph, the aforementioned isomorphic knowledge graph set, and the pre-trained quality assessment model, audit assessment information corresponding to the aforementioned target audit report text is generated. The isomorphic knowledge graph set enriches the graph features of the target knowledge graph, thereby improving the accuracy of the generated audit assessment information. Finally, the aforementioned audit assessment information is sent to the target display terminal for display. This method solves the problems associated with manual auditing, significantly improving audit efficiency and the accuracy of the generated audit assessment information.

[0069] Further reference Figure 3 The document illustrates a flow 300 of another embodiment of a knowledge graph-based audit assessment information generation method. This flow 300 includes the following steps:

[0070] Step 301: Obtain the target audit report text.

[0071] In some embodiments, the specific implementation of step 301 and its resulting technical effects can be found in [reference needed]. Figure 2 Step 201 in the corresponding embodiment will not be repeated here.

[0072] Step 302: Clean the target audit report text to generate text information.

[0073] In some embodiments, the entity executing the knowledge graph-based audit assessment information generation method (e.g.) Figure 1The computing device 101 shown performs text cleaning on the target audit report text to generate text information, which may include the following steps:

[0074] The first step is to read the target audit report text to generate candidate text information.

[0075] As an example, the aforementioned execution entity can use a full read approach to read the entire contents of the target audit report text at once to generate the candidate text information. For instance, the execution entity can read the target audit report text using the following code:

[0076] file = open("filename")

[0077] Inf = file.read()

[0078] Here, "filename" represents the file name of the target audit report text. "Inf" represents the text information. The executing entity can read the contents of the target audit report text all at once using the read() method.

[0079] The second step is to perform entity recognition on the above candidate text information to generate the first entity information set.

[0080] The aforementioned executing entity can perform entity recognition on the aforementioned candidate text information using an entity recognition model. The first entity information in the aforementioned first entity information set includes: entity location information and entity information. The entity location information represents the position of the entity corresponding to the first entity information within the aforementioned candidate text information. The entity information represents the entity at the position corresponding to the entity location information.

[0081] For example, the first entity information could be [entity location information: (12,23), entity information: "Zhang San"].

[0082] Third, for each piece of first entity information in the aforementioned first entity information set, perform the following processing steps:

[0083] The first sub-step involves determining the entity similarity between the entity information included in the first entity information and each entity information in the target entity database, in order to generate a similarity value.

[0084] The aforementioned executing entity can generate a similarity value by determining the cosine similarity between the entity information included in the first entity information and each entity information in the target entity database.

[0085] The second sub-step involves filtering out entity information whose similarity values ​​fall within the target range from the aforementioned target entity database, using these as candidate entity information to obtain a set of candidate entity information.

[0086] The target range mentioned above can be [100%, 85%].

[0087] The third sub-step involves replacing the first entity information included in the candidate text information with entities based on the entity location information and target candidate entity information included in the first entity information.

[0088] Specifically, the executing entity can replace the entity corresponding to the first entity information at the aforementioned entity location information with the entity corresponding to the aforementioned target candidate entity information. The aforementioned target candidate entity information refers to candidate entity information in the aforementioned candidate entity information set that satisfies the first screening condition. The first screening condition is that the similarity value corresponding to the candidate entity information is the largest.

[0089] The fourth sub-step is to determine the candidate text information after entity replacement as the aforementioned text information.

[0090] Step 303: Generate the target knowledge graph corresponding to the text information.

[0091] In some embodiments, the process of generating a target knowledge graph corresponding to text information by the aforementioned execution entity may include the following steps:

[0092] The first step is to perform entity recognition on the above text information to generate a second entity information set.

[0093] The second entity information in the aforementioned second entity information set represents the entities in the aforementioned text information. For example, the aforementioned executing entity can perform entity recognition on the aforementioned text information using a Hidden Markov Model (HMM).

[0094] The second step is to remove entities from the second entity information in the second entity information set to generate the removed entity information, thus obtaining the removed entity information set.

[0095] Among them, the removed entity information in the above-mentioned set of removed entity information is information corresponding to entities related to the audit.

[0096] The third step is to determine the relation information corresponding to each entity in the above-mentioned removed entity information set, thus obtaining a relation information group set.

[0097] In this context, the relationship information group set represents multiple relationships corresponding to one entity. The aforementioned execution entity can use the BERT (Bidirectional Encoder Representations from Transformers) model to determine the relationship information corresponding to each entity in the removed entity information set, thus obtaining the relationship information group set.

[0098] The fourth step is to generate the target knowledge graph based on the above-mentioned set of entity information after elimination and the above-mentioned set of relationship information groups.

[0099] Step 304: Determine the isomorphic knowledge graph corresponding to the target knowledge graph to obtain a set of isomorphic knowledge graphs.

[0100] In some embodiments, the execution entity determines the isomorphic knowledge graph corresponding to the target knowledge graph to obtain a set of isomorphic knowledge graphs, which may include the following steps:

[0101] The first step is to select knowledge graphs that meet the second selection criteria from the knowledge graph database as candidate knowledge graphs, thus obtaining a set of candidate knowledge graphs.

[0102] The second filtering criterion is that the number of entity nodes in the knowledge graph is greater than or equal to the number of entity nodes in the target knowledge graph, and the number of relation edges in the knowledge graph is greater than or equal to the number of relation edges in the target knowledge graph. The aforementioned knowledge graph library is a library that stores knowledge graphs from multiple sources.

[0103] The second step involves performing the following processing steps on each entity node in the target knowledge graph:

[0104] The first sub-step involves constructing feature vectors for the aforementioned entity nodes to generate the first entity node feature vector.

[0105] As an example, the aforementioned execution entity can perform one-hot encoding on the aforementioned entity nodes to generate the feature vector of the aforementioned first entity node.

[0106] As another example, the aforementioned execution entity can extract features from the aforementioned entity nodes using a convolutional neural network to generate the feature vector of the aforementioned first entity node.

[0107] The second sub-step involves constructing a feature vector for at least one relation edge connected to the aforementioned entity node to generate a first relation edge feature vector.

[0108] As an example, firstly, the aforementioned entity can generate multiple feature vectors by performing one-hot encoding on at least one relational edge connected to the aforementioned entity nodes. Then, the aforementioned execution entity can concatenate the multiple feature vectors to generate the aforementioned first relational edge feature vector.

[0109] Third, for each candidate knowledge graph in the above candidate knowledge graph set, perform the following processing steps:

[0110] The first sub-step involves constructing a feature vector for each entity node in the candidate knowledge graph to generate a second entity node feature vector, thus obtaining a sequence of second entity node feature vectors.

[0111] As an example, the aforementioned execution entity can perform one-hot encoding on each entity node in the candidate knowledge graph to generate a second entity node feature vector, thus obtaining a sequence of second entity node feature vectors.

[0112] The second sub-step involves constructing feature vectors for each target relation edge group in the aforementioned candidate knowledge graph to generate a second relation edge feature vector sequence.

[0113] Among them, the target relation edge group is at least one relation edge connected to the above candidate knowledge graph entity nodes.

[0114] As an example, the aforementioned execution entity can perform one-hot encoding on each target relation edge group in the aforementioned candidate knowledge graph to generate a second relation edge feature vector sequence.

[0115] As another example, the aforementioned execution entity can use a convolutional neural network to extract features from each target relation edge group in the aforementioned candidate knowledge graph to generate a second relation edge feature vector sequence.

[0116] The third sub-step involves determining the feature similarity between each first entity node feature vector in the obtained first entity node feature vector sequence and each second entity node feature vector in the above-mentioned second entity node feature vector sequence, thereby obtaining a sequence of entity node similarity value groups.

[0117] The feature similarity between the feature vectors of the first entity node and the feature vectors of the second entity node can be cosine similarity.

[0118] The fourth sub-step involves determining the feature similarity between each first relation edge feature vector in the obtained first relation edge feature vector sequence and each second relation edge feature vector in the above-mentioned second relation edge feature vector sequence, thereby obtaining a relation edge similarity value group sequence.

[0119] The feature similarity between the feature vectors of the first relation edge and the feature vectors of the second relation edge can be cosine similarity.

[0120] The fifth sub-step involves selecting the entity node similarity value with the largest value from each entity node similarity value group in the above entity node similarity value group sequence as the first candidate value, thus obtaining the first candidate value set.

[0121] The sixth sub-step involves selecting the relation edge similarity value with the largest value from each relation edge similarity value group in the above relation edge similarity value group sequence as the second candidate value, thus obtaining the second candidate value set.

[0122] The seventh sub-step involves performing a weighted summation of the first candidate value in the first candidate value set and the second candidate value in the second candidate value set to generate the graph similarity value corresponding to the aforementioned candidate knowledge graph.

[0123] The fourth step is to select candidate knowledge graphs whose graph similarity values ​​meet the third selection criteria from the above candidate knowledge graph set, and use them as isomorphic knowledge graphs to obtain the above isomorphic knowledge graph set.

[0124] The third screening criterion is that the map similarity value is greater than or equal to the map similarity value of the target position in the obtained map similarity value sequence. The map similarity value sequence is ordered from largest to smallest. The target position can be "4".

[0125] In practice, the knowledge graphs corresponding to different audit report texts often contain some different feature information. Therefore, by determining the set of isomorphic knowledge graphs corresponding to the target knowledge graph, the richness of features can be increased, making the generated audit evaluation information more accurate. Thus, determining the set of isomorphic knowledge graphs corresponding to the target knowledge graph is particularly important. First, isomorphic knowledge graphs should contain similar numbers of relation edges and entity nodes. Therefore, this disclosure uses a second screening condition to perform a coarse screening of the knowledge graph to filter out useless knowledge graphs. Furthermore, the graph structures of isomorphic knowledge graphs should be similar. Therefore, this disclosure determines isomorphic knowledge graphs by constructing feature vectors for entity nodes and relation edges connected to them, and calculating the similarity based on the constructed feature vectors. This method greatly improves the accuracy of the obtained isomorphic knowledge graphs, thereby improving the accuracy of the generated audit evaluation information.

[0126] Step 305: Generate audit assessment information corresponding to the target audit report text based on the target knowledge graph, the isomorphic knowledge graph set, and the pre-trained quality assessment model.

[0127] In some embodiments, the execution entity can generate audit assessment information corresponding to the target audit report text based on the target knowledge graph, a set of isomorphic knowledge graphs, and a pre-trained quality assessment model. The quality assessment model includes: a set of feature extraction models, a first classification model, a feature fusion model, and a second classification model. For example, the feature extraction model in the set of feature extraction models can be a VGG-16 model. The feature fusion model can be an LSTM model. The first and second classification models can be convolutional neural networks with classification layers.

[0128] The aforementioned executing entity generates audit assessment information corresponding to the target audit report text based on the target knowledge graph, the isomorphic knowledge graph set, and the pre-trained quality assessment model. This may include the following steps:

[0129] The first step is to input the target knowledge graph and the isomorphic knowledge graph in the above set of isomorphic knowledge graphs into the feature extraction model in the above set of feature extraction models in parallel to generate the first feature graph feature information and the second feature graph feature information set.

[0130] The first feature graph feature information is the feature information of the target knowledge graph extracted by the feature extraction model. The second graph feature information is the feature information of the isomorphic knowledge graph extracted by the feature extraction model.

[0131] The second step is to input the first feature map feature information and the second feature map feature information set into the first classification model to generate classification results.

[0132] The third step involves inputting the first feature map feature information and at least one second feature map feature information corresponding to the classification result into the feature fusion model to generate the third feature map feature information.

[0133] The fourth step is to input the aforementioned third map feature information into the aforementioned second classification model to generate the aforementioned audit assessment information.

[0134] Step 306: Send a communication request to the communication terminal connected to the target display terminal.

[0135] In some embodiments, the executing entity may send a communication request to a communication terminal connected to the target display terminal. The communication terminal is communicatively connected to the target display terminal. The communication terminal may be a terminal with information transmission and reception capabilities. The communication request may be a request to establish a communication connection with the communication terminal.

[0136] Step 307: In response to receiving a communication connection request from a communication terminal, establish an end-to-end communication link with the communication terminal.

[0137] In some embodiments, the executing entity may, in response to receiving a communication connection request from a communication terminal, establish an end-to-end communication link with the communication terminal. The communication connection request may be a request to establish a communication connection with the executing entity.

[0138] Step 308: In response to confirming that the communication link has been successfully created, transmit the encrypted text corresponding to the audit evaluation information to the communication terminal through the communication link.

[0139] In some embodiments, the executing entity may, in response to determining that the communication link has been successfully established, transmit the encrypted text corresponding to the audit evaluation information to the communication terminal via the communication link. The encrypted text corresponding to the audit evaluation information may be text generated by encrypting it using an encryption algorithm. The encryption algorithm may be a symmetric encryption algorithm.

[0140] Optionally, the aforementioned implementing entity may also perform the following steps:

[0141] The first step is to construct an encryption key based on the communication address and sending timestamp of the aforementioned communication terminal.

[0142] The aforementioned communication address can be the IP (Internet Protocol) address of the communication terminal or the MAC (Media Access Control Address).

[0143] As an example, firstly, the executing entity can concatenate the communication address and the sending timestamp into a string to generate a concatenated string. Then, the executing entity can hash the concatenated string using a hash algorithm to generate the encryption key.

[0144] The second step is to encrypt the audit assessment information using the aforementioned encryption key to generate the encrypted text.

[0145] The aforementioned executing entity may use the aforementioned encryption key as its private key and employ the public key to encrypt the aforementioned audit and evaluation information using an asymmetric encryption algorithm, thereby generating the aforementioned encrypted text.

[0146] Step 309: Send the audit assessment information to the target display terminal for display.

[0147] In some embodiments, the specific implementation of step 309 and its resulting technical effects can be found in [reference needed]. Figure 2 Step 206 in the corresponding embodiment will not be repeated here.

[0148] from Figure 3 It can be seen that, with Figure 2Compared to the descriptions of some corresponding embodiments, this disclosure first determines isomorphic knowledge graphs similar to the target knowledge graph by determining entity node similarity values ​​and relation edge similarity values, and then using weighted summation. This method achieves faster determination of isomorphic knowledge graphs. Furthermore, by using multiple parallel feature extraction models to extract features from both the target knowledge graph and the isomorphic knowledge graph, the resulting graph features are greatly enriched. This improves the accuracy of the generated audit assessment information.

[0149] Further reference Figure 4 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a knowledge graph-based audit assessment information generation device, which are similar to... Figure 2 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0150] like Figure 4 As shown, an audit assessment information generation device 400 based on a knowledge graph in some embodiments includes: an acquisition unit 401, a text cleaning unit 402, a first generation unit 403, a determination unit 404, a second generation unit 405, and a display unit 406. The acquisition unit is configured to acquire a target audit report text; the text cleaning unit is configured to perform text cleaning on the target audit report text to generate text information; the first generation unit is configured to generate a target knowledge graph corresponding to the text information; the determination unit is configured to determine a homogeneous knowledge graph corresponding to the target knowledge graph, obtaining a set of homogeneous knowledge graphs; the second generation unit is configured to generate audit assessment information corresponding to the target audit report text based on the target knowledge graph, the set of homogeneous knowledge graphs, and a pre-trained quality assessment model; and the display unit is configured to send the audit assessment information to a target display terminal for display.

[0151] In some optional implementations of certain embodiments, the text cleaning unit 402 is further configured to: read the target audit report text to generate candidate text information; perform entity recognition on the candidate text information to generate a first entity information set, wherein the first entity information in the first entity information set includes: entity location information and entity information; determine the entity similarity between the entity information included in the first entity information and each entity information in the target entity library to generate a similarity value; filter out entity information whose corresponding similarity value is within the target interval from the target entity library as candidate entity information to obtain a candidate entity information set; perform entity replacement on the first entity information included in the candidate text information according to the entity location information included in the first entity information and the target candidate entity information, wherein the target candidate entity information is the candidate entity information in the candidate entity information set that meets the first screening condition; and determine the candidate text information after entity replacement as the text information.

[0152] In some optional implementations of some embodiments, the first generation unit 403 is further configured to: perform entity recognition on the above-mentioned text information to generate a second entity information set; perform entity removal on the second entity information in the above-mentioned second entity information set to generate removed entity information, thereby obtaining a removed entity information set; determine the relation information corresponding to each entity information in the removed entity information set to obtain a relation information group set; and generate the above-mentioned target knowledge graph based on the removed entity information set and the relation information group set.

[0153] In some optional implementations of certain embodiments, the determining unit 404 is further configured to: filter knowledge graphs that meet the second filtering condition from the knowledge graph library as candidate knowledge graphs, thereby obtaining a set of candidate knowledge graphs, wherein the second filtering condition is that the number of entity nodes contained in the knowledge graph is greater than or equal to the number of entity nodes contained in the target knowledge graph, and the number of relation edges contained in the knowledge graph is greater than or equal to the number of relation edges contained in the target knowledge graph; perform the following processing steps on each entity node in the target knowledge graph: construct feature vectors for the entity node to generate a first entity node feature vector; construct feature vectors for at least one relation edge connected to the entity node to generate a first relation edge feature vector; for each candidate knowledge graph in the set of candidate knowledge graphs, perform the following processing steps: construct feature vectors for each entity node in the candidate knowledge graph to generate a second entity node feature vector, thereby obtaining a second entity node feature vector sequence; construct feature vectors for each target relation edge group in the candidate knowledge graph to generate a second relation edge feature vector sequence, wherein the target relation edge group is connected to the entity nodes of the candidate knowledge graph. At least one relation edge; determine the feature similarity between each first entity node feature vector in the obtained first entity node feature vector sequence and each second entity node feature vector in the obtained second entity node feature vector sequence, to obtain an entity node similarity value group sequence; determine the feature similarity between each first relation edge feature vector in the obtained first relation edge feature vector sequence and each second relation edge feature vector in the obtained second relation edge feature vector sequence, to obtain a relation edge similarity value group sequence; select the entity node with the largest value from each entity node similarity value group in the above entity node similarity value group sequence. The similarity value is used as the first candidate value, resulting in a first candidate value set. From each relation edge similarity value group in the above sequence, the relation edge similarity value with the largest value is selected as the second candidate value, resulting in a second candidate value set. The first candidate values ​​in the first candidate value set and the second candidate values ​​in the second candidate value set are weighted and summed to generate the graph similarity value corresponding to the above candidate knowledge graphs. From the above candidate knowledge graph set, the candidate knowledge graphs whose graph similarity values ​​satisfy the third selection condition are selected as isomorphic knowledge graphs, resulting in the above isomorphic knowledge graph set. In practice, the knowledge graphs corresponding to different audit report texts often contain some different feature information. Therefore, by determining the isomorphic knowledge graph set corresponding to the target knowledge graph, the richness of features can be increased, making the generated audit evaluation information more accurate. Therefore, how to determine the isomorphic knowledge graph set corresponding to the target knowledge graph is particularly important. First, the isomorphic knowledge graph should contain a similar number of relation edges and entity nodes.Therefore, this disclosure performs a coarse screening of the knowledge graph using a second screening criterion to filter out useless knowledge graphs. Furthermore, isomorphic knowledge graphs should have similar graph structures. Therefore, this disclosure determines isomorphic knowledge graphs by constructing feature vectors for entity nodes and the relational edges connected to them, and calculating the similarity based on the constructed feature vectors. This method significantly improves the accuracy of the obtained isomorphic knowledge graphs. This, in turn, improves the accuracy of the generated audit evaluation information.

[0154] In some optional implementations of certain embodiments, the quality assessment model includes: a set of feature extraction models, a first classification model, a feature fusion model, and a second classification model; and the second generation unit 405 is further configured to: input the target knowledge graph and the isomorphic knowledge graph in the set of isomorphic knowledge graphs in parallel into the feature extraction models in the set of feature extraction models to generate a first feature graph feature information and a second set of feature graph feature information; input the first feature graph feature information and the second set of feature graph feature information into the first classification model to generate a classification result; input the first feature graph feature information and at least one second set of feature graph feature information corresponding to the classification result into the feature fusion model to generate a third set of feature graph feature information; and input the third set of feature graph feature information into the second classification model to generate the audit assessment information.

[0155] In some optional implementations of certain embodiments, before sending the audit evaluation information to the target display terminal for display, the apparatus 400 further includes: sending a communication request to a communication terminal connected to the target display terminal; in response to receiving the communication connection request sent by the communication terminal, establishing an end-to-end communication link with the communication terminal; and in response to determining that the communication link has been successfully established, transmitting the encrypted text corresponding to the audit evaluation information to the communication terminal through the communication link.

[0156] In some optional implementations of certain embodiments, before transmitting the encrypted text corresponding to the audit evaluation information to the communication terminal via the communication link, the device 400 further includes: constructing an encryption key based on the communication address and sending timestamp of the communication terminal; and encrypting the audit evaluation information using the encryption key to generate the encrypted text.

[0157] It is understandable that the units described in the device 400 are related to the reference. Figure 2 The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method also apply to device 400 and the units contained therein, and will not be repeated here.

[0158] The following is for reference. Figure 5 It illustrates electronic devices suitable for implementing some embodiments of the present disclosure (such as...). Figure 1 The diagram shows the structure of the computing device 101)500. Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0159] like Figure 5 As shown, the electronic device 500 may include a processing unit (e.g., a central processing unit, a graphics processor, etc.) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage device 508 into a random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device 500. The processing unit 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0160] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 An electronic device 500 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 5 Each box shown can represent a device or multiple devices as needed.

[0161] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 509, or installed from storage device 508, or installed from ROM 502. When the computer program is executed by processing device 501, it performs the functions defined in the methods of some embodiments of this disclosure.

[0162] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: 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 some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal 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. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0163] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0164] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: acquire the target audit report text; perform text cleaning on the target audit report text to generate text information; generate a target knowledge graph corresponding to the text information; determine the isomorphic knowledge graph corresponding to the target knowledge graph to obtain a set of isomorphic knowledge graphs; generate audit assessment information corresponding to the target audit report text based on the target knowledge graph, the set of isomorphic knowledge graphs, and a pre-trained quality assessment model; and send the audit assessment information to a target display terminal for display.

[0165] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language 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 server. 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).

[0166] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0167] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including an acquisition unit, a text cleaning unit, a first generation unit, a determination unit, a second generation unit, and a display unit. The names of these units do not necessarily limit the specific unit; for example, the acquisition unit may also be described as "a unit for acquiring the target audit report text."

[0168] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0169] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for generating audit assessment information based on knowledge graphs, comprising: Obtain the target audit report text; The target audit report text is cleaned to generate text information; Generate the target knowledge graph corresponding to the text information; Determine the isomorphic knowledge graph corresponding to the target knowledge graph to obtain a set of isomorphic knowledge graphs, including: The knowledge graphs that meet the second screening condition are selected from the knowledge graph library as candidate knowledge graphs, and a set of candidate knowledge graphs is obtained. The second screening condition is that the number of entity nodes contained in the knowledge graph is greater than or equal to the number of entity nodes contained in the target knowledge graph, and the number of relation edges contained in the knowledge graph is greater than or equal to the number of relation edges contained in the target knowledge graph. Perform the following processing steps on each entity node in the target knowledge graph: The entity nodes are used to construct feature vectors to generate the first entity node feature vector; A feature vector is constructed for at least one relation edge connected to the entity node to generate a first relation edge feature vector; For each candidate knowledge graph in the candidate knowledge graph set, perform the following processing steps: For each entity node in the candidate knowledge graph, a feature vector is constructed to generate a second entity node feature vector, resulting in a sequence of second entity node feature vectors. For each target relation edge group in the candidate knowledge graph, a feature vector is constructed to generate a second relation edge feature vector sequence, wherein the target relation edge group is at least one relation edge connected to the entity node of the candidate knowledge graph; Determine the feature similarity between each first entity node feature vector in the obtained first entity node feature vector sequence and each second entity node feature vector in the second entity node feature vector sequence to obtain a sequence of entity node similarity value groups; Determine the feature similarity between each first relation edge feature vector in the obtained first relation edge feature vector sequence and each second relation edge feature vector in the second relation edge feature vector sequence to obtain a relation edge similarity value group sequence; From each entity node similarity value group in the entity node similarity value group sequence, the entity node similarity value with the largest value is selected as the first candidate value, thus obtaining the first candidate value set; From each relation edge similarity value group in the relation edge similarity value group sequence, the relation edge similarity value with the largest value is selected as the second candidate value, thus obtaining the second candidate value set; The first candidate value in the first candidate value set and the second candidate value in the second candidate value set are weighted and summed to generate the graph similarity value corresponding to the candidate knowledge graph; Candidate knowledge graphs whose graph similarity values ​​meet the third screening condition are selected from the candidate knowledge graph set and used as isomorphic knowledge graphs to obtain the isomorphic knowledge graph set. Based on the target knowledge graph, the isomorphic knowledge graph set, and the pre-trained quality assessment model, generate audit assessment information corresponding to the target audit report text; The audit assessment information is sent to the target display terminal for display.

2. The method according to claim 1, wherein, The text cleaning of the target audit report text to generate text information includes: Read the target audit report text to generate candidate text information; Entity recognition is performed on the candidate text information to generate a first entity information set, wherein the first entity information in the first entity information set includes: entity location information and entity information; Determine the entity similarity between the entity information included in the first entity information and each entity information in the target entity library to generate a similarity value; Entity information whose similarity value is within the target interval is selected from the target entity database and used as candidate entity information to obtain a set of candidate entity information; Based on the entity location information and target candidate entity information included in the first entity information, entity replacement is performed on the first entity information included in the candidate text information, wherein the target candidate entity information is the candidate entity information in the candidate entity information set that satisfies the first screening condition; The candidate text information after entity replacement is determined as the text information.

3. The method according to claim 1, wherein, The generation of the target knowledge graph corresponding to the text information includes: Entity recognition is performed on the text information to generate a second entity information set; The second entity information in the second entity information set is removed to generate the removed entity information, resulting in the removed entity information set. Determine the relation information corresponding to each entity information in the removed entity information set to obtain a relation information group set; The target knowledge graph is generated based on the set of entity information after removal and the set of relationship information groups.

4. The method according to claim 3, wherein, The quality assessment model includes: a set of feature extraction models, a first classification model, a feature fusion model, and a second classification model; and The step of generating audit assessment information corresponding to the target audit report text based on the target knowledge graph, the isomorphic knowledge graph set, and the pre-trained quality assessment model includes: The target knowledge graph and the isomorphic knowledge graph in the isomorphic knowledge graph set are input in parallel into the feature extraction model in the feature extraction model set to generate the first feature graph feature information and the second feature graph feature information set. The first feature map feature information and the second feature map feature information set are input into the first classification model to generate classification results; The first feature map feature information and at least one second map feature information corresponding to the classification result are input into the feature fusion model to generate the third map feature information; The third map feature information is input into the second classification model to generate the audit evaluation information.

5. The method according to claim 1, wherein, Before sending the audit assessment information to the target display terminal for display, the method further includes: Send a communication request to the communication terminal connected to the target display terminal; In response to receiving a communication connection request sent by the communication terminal, an end-to-end communication link is established with the communication terminal; In response to the confirmation that the communication link has been successfully established, the encrypted text corresponding to the audit evaluation information is transmitted to the communication terminal through the communication link.

6. The method according to claim 5, wherein, Before transmitting the encrypted text corresponding to the audit assessment information to the communication terminal via the communication link, the method further includes: Construct an encryption key based on the communication address and sending timestamp of the communication terminal; The audit assessment information is encrypted using the encryption key to generate the encrypted text.

7. A knowledge graph-based audit assessment information generation device, comprising: The acquisition unit is configured to acquire the target audit report text; The text cleaning unit is configured to perform text cleaning on the target audit report text to generate text information; The first generation unit is configured to generate the target knowledge graph corresponding to the text information; The determining unit is configured to determine the isomorphic knowledge graph corresponding to the target knowledge graph, obtaining a set of isomorphic knowledge graphs, including: filtering knowledge graphs that meet a second filtering condition from the knowledge graph library as candidate knowledge graphs, obtaining a set of candidate knowledge graphs, wherein the second filtering condition is that the number of entity nodes contained in the knowledge graph is greater than or equal to the number of entity nodes contained in the target knowledge graph, and the number of relation edges contained in the knowledge graph is greater than or equal to the number of relation edges contained in the target knowledge graph; performing the following processing steps on each entity node in the target knowledge graph: performing feature mapping on the entity node. The following steps are performed for each candidate knowledge graph in the candidate knowledge graph set: First entity node feature vector is constructed for each entity node in the candidate knowledge graph to generate a second entity node feature vector, resulting in a second entity node feature vector sequence; second entity node feature vector is constructed for each target relation edge group in the candidate knowledge graph to generate a second relation edge feature vector sequence, wherein the target relation edge group is at least one relation edge connected to an entity node in the candidate knowledge graph; feature similarity is determined between each first entity node feature vector in the obtained first entity node feature vector sequence and each second entity node feature vector in the obtained second entity node feature vector sequence, resulting in an entity node similarity value group sequence; feature similarity is determined between each first relation edge feature vector in the obtained first relation edge feature vector sequence and each second relation edge feature vector in the obtained second relation edge feature vector sequence, resulting in a relation edge similarity value group sequence; entity with the largest value is selected from each entity node similarity value group in the entity node similarity value group sequence. The node similarity value is used as the first candidate value to obtain a first candidate value set. From each relation edge similarity value group in the relation edge similarity value group sequence, the relation edge similarity value with the largest value is selected as the second candidate value to obtain a second candidate value set. The first candidate value in the first candidate value set and the second candidate value in the second candidate value set are weighted and summed to generate the graph similarity value corresponding to the candidate knowledge graph. From the candidate knowledge graph set, the candidate knowledge graphs whose graph similarity values ​​satisfy the third selection condition are selected as isomorphic knowledge graphs to obtain the isomorphic knowledge graph set. The second generation unit is configured to generate audit assessment information corresponding to the target audit report text based on the target knowledge graph, the isomorphic knowledge graph set, and the pre-trained quality assessment model. The display unit is configured to send the audit evaluation information to the target display terminal for display.

8. The knowledge graph-based audit assessment information generation device according to claim 7, wherein, The determining unit is further configured to: The knowledge graphs that meet the second screening condition are selected from the knowledge graph library as candidate knowledge graphs, and a set of candidate knowledge graphs is obtained. The second screening condition is that the number of entity nodes contained in the knowledge graph is greater than or equal to the number of entity nodes contained in the target knowledge graph, and the number of relation edges contained in the knowledge graph is greater than or equal to the number of relation edges contained in the target knowledge graph. Perform the following processing steps on each entity node in the target knowledge graph: The entity nodes are used to construct feature vectors to generate the first entity node feature vector; A feature vector is constructed for at least one relation edge connected to the entity node to generate a first relation edge feature vector; For each candidate knowledge graph in the candidate knowledge graph set, perform the following processing steps: For each entity node in the candidate knowledge graph, a feature vector is constructed to generate a second entity node feature vector, resulting in a sequence of second entity node feature vectors. For each target relation edge group in the candidate knowledge graph, a feature vector is constructed to generate a second relation edge feature vector sequence, wherein the target relation edge group is at least one relation edge connected to the entity node of the candidate knowledge graph; Determine the feature similarity between each first entity node feature vector in the obtained first entity node feature vector sequence and each second entity node feature vector in the second entity node feature vector sequence to obtain a sequence of entity node similarity value groups; Determine the feature similarity between each first relation edge feature vector in the obtained first relation edge feature vector sequence and each second relation edge feature vector in the second relation edge feature vector sequence to obtain a relation edge similarity value group sequence; From each entity node similarity value group in the entity node similarity value group sequence, the entity node similarity value with the largest value is selected as the first candidate value, thus obtaining the first candidate value set; From each relation edge similarity value group in the relation edge similarity value group sequence, the relation edge similarity value with the largest value is selected as the second candidate value, thus obtaining the second candidate value set; The first candidate value in the first candidate value set and the second candidate value in the second candidate value set are weighted and summed to generate the graph similarity value corresponding to the candidate knowledge graph; From the candidate knowledge graph set, candidate knowledge graphs whose graph similarity values ​​meet the third screening condition are selected as isomorphic knowledge graphs, thus obtaining the isomorphic knowledge graph set.

9. An electronic device, comprising: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 6.

10. A computer-readable medium having a computer program stored thereon, wherein, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 6.

Images