CN114756691A

CN114756691A - Structure chart generation method, model training method, map generation method and device

Info

Publication number: CN114756691A
Application number: CN202210432872.XA
Authority: CN
Inventors: 王鑫; 孙明明; 李平
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2022-04-22
Filing date: 2022-04-22
Publication date: 2022-07-15

Abstract

The disclosure provides a structure diagram generation method, a structure diagram generation model training method, a map generation method, a device, an electronic device, a storage medium and a program product, and relates to the technical field of data processing, in particular to the technical fields of maps, deep learning and the like. The specific implementation scheme is as follows: carrying out context coding on a statement to be processed to obtain a coding vector sequence; determining node information, topological structure information and side information for generating a structure diagram based on the coding vector sequence, wherein the node information is used for representing the attribute information of the nodes of the structure diagram, the topological structure information is used for representing whether sides exist among a plurality of nodes, and the side information is used for representing the incidence relation among the plurality of nodes; and generating a target structure chart aiming at the statement to be processed based on the node information, the topological structure information and the side information.

Description

Structure chart generation method, model training method, map generation method and device

Technical Field

The disclosure relates to the technical field of data processing, in particular to the technical fields of maps, deep learning and the like. In particular, the present invention relates to a structure diagram generation method, a training method of a structure diagram generation model, a map generation method, an apparatus, an electronic device, a storage medium, and a program product.

Background

Information extraction technology plays a very important role in the field of artificial intelligence, and more researches and applications in the field of artificial intelligence depend on the information extraction technology. For example, the knowledge graph is generated by extracting the content from the open domain information to enrich and supplement the content in the knowledge base, so as to be applied to the fields of retrieval, human-computer interaction and the like. For example, the content is extracted from the open domain information to generate a case map, so that the association relationship between a plurality of events can be clearly combed. How to accurately and effectively extract information from open field information becomes a research focus.

Disclosure of Invention

The disclosure provides a structure diagram generation method, a structure diagram generation model training method, a diagram generation method, a device, an electronic device, a storage medium and a program product.

According to an aspect of the present disclosure, there is provided a structure diagram generating method, including: carrying out context coding on the statement to be processed to obtain a coding vector sequence; determining node information, topological structure information and side information for generating a structure diagram based on the coding vector sequence, wherein the node information is used for representing attribute information of the nodes of the structure diagram, the topological structure information is used for representing whether edges exist among a plurality of nodes, and the side information is used for representing the association relation among the nodes; and generating a target structure diagram aiming at the statement to be processed based on the node information, the topological structure information and the side information.

According to another aspect of the present disclosure, a method for training a structure diagram generation model is provided, where the structure diagram generation model includes a context coding submodel, a node label generation submodel, a topology generation submodel, and an edge label generation submodel, and the method for training the structure diagram generation model includes: inputting a sample statement corresponding to a sample structure diagram into the context coding sub-model to obtain a sample coding vector sequence, wherein the sample structure diagram comprises a sample node label, a sample topological structure label and a sample edge label; inputting the sample coding vector sequence into the node label generation submodel to obtain sample node information; inputting the sample statement into the topological structure generation submodel to obtain sample topological structure information; inputting the sample statement into the side label generation submodel to obtain sample side information; and training the structure chart generation model based on the sample structure chart, the sample node information, the sample topological structure information and the sample side information to obtain a trained structure chart generation model.

According to another aspect of the present disclosure, there is provided an atlas generation method, including: generating a target map based on a structure map, wherein the structure map is generated by using the structure map generating method of the present disclosure.

According to another aspect of the present disclosure, there is provided a structural diagram generating apparatus including: the encoding module is used for carrying out context encoding on the statement to be processed to obtain an encoding vector sequence; the determining module is used for determining node information, topological structure information and edge information which are used for generating a structural diagram based on the coding vector sequence, wherein the node information is used for representing attribute information of nodes of the structural diagram, the topological structure information is used for representing whether edges exist among a plurality of nodes, and the edge information is used for representing the incidence relation among the nodes; and the structure chart generating module is used for generating a target structure chart aiming at the statement to be processed based on the node information, the topological structure information and the side information.

According to another aspect of the present disclosure, there is provided a training apparatus for a structure diagram generation model, where the structure diagram generation model includes a context coding submodel, a node label generation submodel, a topology generation submodel, and an edge label generation submodel, and the training apparatus for the structure diagram generation model includes: the first input module is used for inputting a sample statement corresponding to a sample structure chart into the context coding sub-model to obtain a sample coding vector sequence, wherein the sample structure chart comprises a sample node label, a sample topological structure label and a sample edge label; the second input module is used for inputting the sample coding vector sequence to the node label generation submodel to obtain sample node information; the third input module is used for inputting the sample statement to the topological structure generation submodel to obtain sample topological structure information; the fourth input module is used for inputting the sample statement to the side label generation submodel to obtain sample side information; and the training module is used for training the structure diagram generation model based on the sample structure diagram, the sample node information, the sample topological structure information and the sample side information to obtain a trained structure diagram generation model.

According to another aspect of the present disclosure, there is provided an atlas generating apparatus including: the map generation module is used for generating a target map based on a structural map, wherein the structural map is generated by using the structural map generation device disclosed by the disclosure.

According to another aspect of the present disclosure, there is provided an electronic device including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method according to the present disclosure.

According to another aspect of the present disclosure, there is provided a non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform a method as disclosed herein.

According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program which, when executed by a processor, implements a method as disclosed herein.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.

Drawings

The drawings are included to provide a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:

FIG. 1 schematically illustrates an exemplary system architecture to which the method and apparatus for architectural diagram generation may be applied, according to an embodiment of the disclosure;

FIG. 2 schematically shows a block diagram according to an embodiment of the disclosure;

FIG. 3 schematically illustrates a flow chart of a structure graph generation method according to an embodiment of the disclosure;

FIG. 4 schematically illustrates a schematic diagram of a structure graph generation model according to an embodiment of the disclosure;

FIG. 5 schematically shows a schematic diagram of generating a target structure diagram according to an embodiment of the disclosure;

FIG. 6 schematically illustrates a flow chart of a training method of a structure graph generative model according to an embodiment of the disclosure;

FIG. 7 schematically illustrates a flow chart of a method of atlas generation according to an embodiment of the disclosure;

FIG. 8 schematically shows a block diagram of a structure diagram generation apparatus according to an embodiment of the present disclosure;

FIG. 9 schematically illustrates a block diagram of a training apparatus for structure diagram generation of a model according to an embodiment of the disclosure;

figure 10 schematically illustrates a block diagram of an atlas generation apparatus according to an embodiment of the disclosure; and

fig. 11 schematically illustrates a block diagram of an electronic device suitable for implementing the structure diagram generation method according to an embodiment of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, in which various details of the embodiments of the disclosure are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

According to an embodiment of the present disclosure, a method for generating a structure diagram is provided, which may include: carrying out context coding on a statement to be processed to obtain a coding vector sequence; determining node information, topological structure information and side information for generating a structure diagram based on the coding vector sequence, wherein the node information is used for representing attribute information of the nodes of the structure diagram, the topological structure information is used for representing whether sides exist among a plurality of nodes, and the side information is used for representing the incidence relation among the plurality of nodes; and generating a target structure diagram aiming at the statement to be processed based on the node information, the topological structure information and the side information.

In the technical scheme of the disclosure, the processes of collecting, storing, using, processing, transmitting, providing, disclosing and applying the personal information of the related users are all in accordance with the regulations of related laws and regulations, necessary security measures are taken, and the customs of public sequences is not violated.

In the technical scheme of the disclosure, before the personal information of the user is obtained or collected, the authorization or the consent of the user is obtained.

Fig. 1 schematically illustrates an exemplary system architecture to which the method and apparatus for structure diagram generation may be applied, according to an embodiment of the disclosure.

It should be noted that fig. 1 is only an example of a system architecture to which the embodiments of the present disclosure may be applied to help those skilled in the art understand the technical content of the present disclosure, and does not mean that the embodiments of the present disclosure may not be applied to other devices, systems, environments or scenarios. For example, in another embodiment, an exemplary system architecture to which the method and apparatus for generating a structure diagram may be applied may include a terminal device, but the terminal device may implement the method and apparatus for generating a structure diagram provided in the embodiments of the present disclosure without interacting with a server.

As shown in fig. 1, a system architecture 100 according to this embodiment may include

terminal devices

101, 102, 103, a network 104, and a server 105. Network 104 is the medium used to provide communication links between

terminal devices

101, 102, 103 and server 105. Network 104 may include various connection types, such as wired and/or wireless communication links, and so forth.

A user may use

terminal devices

101, 102, 103 to interact with a server 105 over a network 104 to receive or send messages or the like. The

terminal devices

101, 102, 103 may have installed thereon various communication client applications, such as a knowledge reading-type application, a web browser application, a search-type application, an instant messaging tool, a mailbox client, and/or social platform software, etc. (by way of example only).

The

terminal devices

101, 102, 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like.

The server 105 may be a server providing various services, such as a background management server (for example only) providing support for content browsed by users using the

terminal devices

101, 102, 103. The background management server may perform processing such as structure diagram generation on the received statements to be processed from the

terminal devices

101, 102, and 103, and generate a target structure diagram.

It should be noted that the structure diagram generating method provided in the embodiment of the present disclosure may be generally executed by the

terminal device

101, 102, or 103. Correspondingly, the structural diagram generating device provided by the embodiment of the present disclosure may also be disposed in the

terminal device

101, 102, or 103.

Alternatively, the method for generating a structure diagram provided in the embodiment of the present disclosure may be generally executed by the server 105. Accordingly, the structure diagram generating apparatus provided in the embodiment of the present disclosure may be generally disposed in the server 105. The structure diagram generating method provided in the embodiment of the present disclosure may also be executed by a server or a server cluster that is different from the server 105 and is capable of communicating with the

terminal devices

101, 102, 103 and/or the server 105. Accordingly, the structural diagram generating apparatus provided in the embodiment of the present disclosure may also be disposed in a server or a server cluster that is different from the server 105 and is capable of communicating with the

terminal devices

101, 102, and 103 and/or the server 105.

It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for an implementation.

It should be noted that the sequence numbers of the respective operations in the following methods are merely used as a representation of the operations for description, and should not be construed as representing the execution order of the respective operations. The method need not be performed in the exact order shown, unless explicitly stated.

A structure diagram may refer to a general intermediate representation of information extraction, according to an embodiment of the disclosure. The structure graph may be generated based on the statement level corpus and the structure graph may be a word level structure graph. The architectural diagram may be utilized to generate an "open field informational representation," i.e., a phrase architectural diagram.

According to an embodiment of the present disclosure, the graph may include nodes and edges. The architectural graph can also include node information about the nodes, e.g., the node information can be used to characterize attribute information of the nodes of the architectural graph. The edges in the structure graph can be determined by the topology information, i.e. whether an edge exists between two nodes can be determined by the topology information. The structure graph may also include side information about the edges, which is used to characterize the association between any two nodes where an edge exists.

Fig. 2 schematically shows a block diagram according to an embodiment of the disclosure. In fig. 2, circular boxes represent word nodes; the dotted oval represents node information; the dotted rectangle represents side information; solid oval boxes represent phrase nodes.

As shown in FIG. 2, the structure diagram is a word level structure diagram for the "is that true" statement. Node 211 of structure graph 210 includes nodes "is", "that", and "true". The edges 212 of the graph include an edge between two nodes "is" and "that" and an edge between two nodes "is" and "true". Node information 213 includes, for example, descendent node information, such as "additional node wet," about node "is," for example, that characterizes a descendent node that would be generated by this node. Node information 213 also includes descendent node information, e.g., "no additions," about node "that, e.g., characterizes that no descendent node is to be generated by this node. The side information 214 may include information characterizing an association relationship between two nodes, e.g., "is" and "that", e.g., "cardinal predicate". Side information 214 may also include information characterizing an association between two nodes, e.g., "is" and "true," e.g., "next word.

As shown in fig. 2, by using the structure diagram generating method provided in the embodiment of the present disclosure, a statement level corpus can be converted into a word level structure diagram. The nodes in the word level structure graph may then be phrase combined to generate an "open domain information representation," such as phrase structure graph 220. Transforming the "open-realm information representation" may result in a graph 230, such as a knowledge graph or a fact graph.

According to the embodiment of the disclosure, the method for generating the structure diagram can improve the processing efficiency and the processing precision of the initial processing stage of various open domain information, and further improve the subsequent information extraction capability and the domain range of the rich map.

Fig. 3 schematically shows a flowchart of a structure diagram generation method according to an embodiment of the present disclosure.

As shown in fig. 3, the method includes operations S310 to S330.

In operation S310, a context is encoded for a statement to be processed, so as to obtain an encoded vector sequence.

In operation S320, node information, topology information, and side information for generating a structure diagram are determined based on the encoding vector sequence. The node information is used for representing attribute information of the nodes in the structure diagram, the topological structure information is used for representing whether edges exist among the nodes, and the edge information is used for representing the association relation among the nodes.

In operation S330, a target structure diagram for the to-be-processed sentence is generated based on the node information, the topology information, and the side information.

According to an embodiment of the present disclosure, the sentence to be processed may be a corpus at a sentence level. The type of the sentence to be processed is not limited, and may be, for example, a text corpus of english, chinese, or other languages.

According to an embodiment of the present disclosure, for operation S310, performing context coding on the statement to be processed to obtain a coded vector sequence may include: and inputting the sentence to be processed into the context coding sub-model to obtain a coding vector sequence. The structure of the context coding sub-model is not limited, and the context coding sub-model may be a deep learning model that can code the content in the sentence to be processed and can perform context coding processing by combining the context semantics of the content in the sentence to be processed.

According to the embodiment of the disclosure, the sentence to be processed can be split into a plurality of 'word level' single texts, the sentence to be processed taking Chinese as an example can be split into a plurality of single texts taking words as a unit, and the sentence to be processed taking English as an example can be split into a plurality of single texts taking words as a unit. The target structure graph may include a plurality of nodes. And the plurality of nodes correspond to the plurality of single texts in the sentence to be processed one by one. The sequence of code vectors may include a plurality of code vectors. The plurality of single texts correspond to the plurality of code vectors one to one.

According to an embodiment of the present disclosure, for operation S320, processing the encoding vector sequence, and determining node information, topology information, and side information for generating the structure diagram may include: and inputting the coding vector sequence into a node label generation sub-model to obtain node information. And inputting the coding vector sequence into a topological structure generation sub-model to obtain topological structure information. And inputting the coding vector sequence into an edge label generation sub-model to obtain edge information. The model structures of the node tag generation submodel, the topology structure generation submodel, and the edge tag generation submodel are not limited, and any deep learning model may be used as long as the above functions can be realized.

According to an embodiment of the present disclosure, for operation S330, generating the target structure diagram for the to-be-processed statement based on the node information, the topology structure information, and the side information may include: each single text in the sentence to be processed is a node. Node information characterizing the node attributes may be attached to each node. Based on the topology information, it is determined whether an edge exists between any two nodes in the statement to be processed, i.e., an edge is generated based on the topology information. And attaching the edge information used for representing the incidence relation between the two nodes to the two nodes with edges based on the edge information.

According to the embodiment of the disclosure, the target structure diagram can be generated by processing the sentence to be processed by utilizing the deep learning model. But is not limited thereto. Syntactic analysis or rules may also be utilized to generate the target structure graph based on the statement to be processed.

According to the embodiment of the disclosure, the target structure diagram is generated by using the deep learning model, data of various open fields can be processed, and the structure diagram generation range is further improved. In addition, the analysis difficulty of generating the structure chart can be reduced, the analysis precision is improved, and cascade errors caused by syntactic analysis or rules are avoided.

According to an embodiment of the present disclosure, for operation S320, determining node information, topology information, and side information for generating a structure diagram based on the coding vector sequence may include: and carrying out first transformation processing on the coding vector sequence to obtain a topological structure representation. Based on the topology representation, topology information is determined. And carrying out second transformation processing on the coding vector sequence to obtain a relation expression. Based on the relational representation, side information is determined. Based on the sequence of encoded vectors, node information is determined.

According to an embodiment of the present disclosure, the first transformation process may be a bilinear transformation process or a affine-bi-transformation process.

According to an embodiment of the present disclosure, the second transformation process may be a bilinear transformation process or a affine-bi-transformation process.

According to the embodiment of the disclosure, a structure diagram generation method based on a neural network is provided, wherein a structure diagram generation model is used for processing a to-be-processed statement, and a target structure diagram for the to-be-processed statement is generated.

FIG. 4 schematically shows a schematic diagram of a structure graph generation model according to an embodiment of the disclosure.

As shown in fig. 4, the structure diagram generation model may include a context coding submodel M410, a node tag generation submodel M420, a topology generation submodel M430, and an edge tag generation submodel M440.

As shown in fig. 4, the sentence to be processed D410 may be input into the context coding submodel M410, resulting in a coding vector sequence D420 containing context information. The encoding vector sequence D420 may be input to the node label generation submodel M420, resulting in node information D430. The sequence of encoding vectors D410 may be input into the topology generation submodel M430, resulting in topology information D440. The encoding vector sequence D420 is input to the side tag generation submodel M440, and side information D450 is obtained.

According to an embodiment of the present disclosure, the context coding submodel may include one or more of RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), GRU (Gate Recurrent Unit), CNN (Convolutional Neural Network), BERT (Bidirectional Encoder Representation), ERNIE (Enhanced reconstruction hydro-node-kNowledge Representation).

As shown in fig. 4, the node tag generation submodel M420 may include an attribute processing module M421. The encoding vector sequence D420 may be input into the attribute processing module M421, generating node information D430. That is, the attribute processing module may perform node attribute processing on the coded vector sequence to obtain node information about node attributes. According to an embodiment of the present disclosure, the attribute processing module may include: and the node full connection layer and the node activation layer are cascaded. The node activation layer may include a Softmax activation function or a Sigmoid activation function.

As shown in fig. 4, the topology generation submodel M430 may include a first transformation module M431 and a topology generation module M432, which are cascaded. The first transformation module M431 may be utilized to perform a first transformation process on the coded vector sequence D420, resulting in a topology representation. The topology generation module M432 determines topology information D440 based on the topology representation.

According to an embodiment of the present disclosure, the first transformation module may include: a bilinear transformation function or a dual affine transformation function. But is not limited thereto. Any other double transformation operating function may also be included.

According to an embodiment of the present disclosure, the topology generation module may include: a cascaded topology full connection layer and a topology activation layer. The topology activation layer may include a Softmax activation function or a Sigmoid activation function.

As shown in fig. 4, the edge label generation submodel M440 may include a second transformation module M441 and an edge label generation module M442 in cascade. The second transformation module M441 may be used to perform a second transformation process on the coded vector sequence D420, so as to obtain a relational expression. The side information D450 is determined by the side label generation module N442 based on the relational representation.

According to an embodiment of the present disclosure, the second transformation module may include: a bilinear transformation function or a dual affine transformation function. But is not limited thereto. Any other double transformation operating function may also be included.

According to an embodiment of the present disclosure, the edge tag generating module may include: and the side label activation layer is connected with the side label full connection layer. The edge tag activation layer may include a Softmax activation function or a Sigmoid activation function.

According to an embodiment of the present disclosure, performing a first transformation process on a coded vector sequence to obtain a topology representation may include: for any two nodes in the statement to be processed, a pair of code vectors corresponding to the two nodes is determined from the sequence of code vectors. The pair of code vectors is subjected to a first transformation process resulting in a representation of the topology between the two nodes.

For example, for any two word nodes, byte point i and byte point j, in the statement to be processed, an encoding vector i corresponding to the byte point i is determined from the encoding vector sequence, and an encoding vector j corresponding to the byte point j is formed by the encoding vector i and the encoding vector j. And performing double affine transformation on the encoding vector pair by using a first transformation processing module to obtain a first adjacency tensor, namely a topological structure representation, between two byte points of the word node i and the byte point j. And processing the topological structure representation by using a topological generation module, generating a result indicating whether an edge exists between two byte points of the word node i and the byte point j, for example, outputting a probability value between 0 and 1, regarding the probability value which is greater than a preset threshold value, for example, 0.8, as topological structure information of the existing edge, and regarding the probability value which is less than or equal to the preset threshold value as topological structure information of the nonexistent edge.

According to an embodiment of the present disclosure, performing a second transformation process on the coded vector sequence to obtain a relational representation may include: for any two nodes in the statement to be processed, a pair of code vectors corresponding to the two nodes is determined from the sequence of code vectors. And performing second transformation processing on the coding vector pair based on the class number of the side information to obtain a relation expression between the two nodes.

For example, for any two word nodes, byte point i and byte point j, in the statement to be processed, an encoding vector i corresponding to the byte point i is determined from the encoding vector sequence, and an encoding vector j corresponding to the byte point j is formed by the encoding vector i and the encoding vector j. Based on the category number of the side information, the encoding vector pair is subjected to double affine transformation by using a second transformation processing module, so that a second adjacency tensor, namely a relation expression, between two byte points of the word node i and the byte point j is obtained. And processing the relation representation by using an edge tag module, generating an edge information result aiming at all edge information categories between two byte points of the word node i and the byte point j, and taking the edge information result corresponding to the edge information category with the maximum confidence coefficient as the edge information.

FIG. 5 schematically shows a schematic diagram of generating a target structure diagram according to an embodiment of the disclosure. In fig. 5, circular boxes represent nodes; dotted oval represents node information; the dashed rectangle represents side information.

As shown in fig. 5, node information 520 of each of a plurality of nodes 510 for a to-be-processed statement "is that true" is obtained, topology information 530 for characterizing whether an edge exists between the plurality of nodes, and edge information 540 for characterizing an association relationship between two nodes.

As shown in FIG. 5, node information 520, topology information 530, and side information 540 may be combined to generate a target structure diagram for the to-be-processed statement "is that true".

According to the embodiment of the disclosure, the node information for characterizing the node attribute may include one or more of node part-of-speech information and derived node information for characterizing whether to generate a derived node. The node part-of-speech information may include part-of-speech information such as nouns, verbs, and adjectives, and may also include part-of-speech information such as subjects and predicates. The derived node information for characterizing whether to derive the node may include, for example, derived node information whether to attach the node "while".

According to an embodiment of the present disclosure, the side information may include an association relationship between a plurality of nodes with each other. Based on the side information and the topology structure information, the side information between the two nodes without sides is deleted, the side information between the two nodes with sides is reserved, and finally the target structure diagram is generated.

FIG. 6 schematically shows a flowchart of a training method of a structure graph generative model according to an embodiment of the present disclosure.

As shown in fig. 6, the method includes operations S610 to S650.

In operation S610, a sample statement corresponding to the sample structure diagram is input to the context coding sub-model, so as to obtain a sample coding vector sequence.

In operation S620, the sample coding vector sequence is input to the node label generation submodel, so as to obtain sample node information.

In operation S630, the sample statement is input to the topology generation submodel, and sample topology information is obtained.

In operation S640, the sample statement is input to the edge tag generation submodel, so as to obtain sample edge information.

In operation S650, a structure diagram generation model is trained based on the sample structure diagram, the sample node information, the sample topology structure information, and the sample side information, to obtain a trained structure diagram generation model.

According to the embodiment of the disclosure, the structure diagram generation model comprises a context coding submodel, a node label generation submodel, a topology structure generation submodel and an edge label generation submodel.

According to an embodiment of the present disclosure, the sample structure graph includes a sample node label, a sample topology label, and a sample edge label.

According to the embodiment of the present disclosure, for operation S650, based on the sample structure diagram, the sample node information, the sample topology structure information, and the sample side information, the structure diagram generation model is trained to obtain a trained structure diagram generation model, which may adopt a multi-task learning (multi-task learning) mode, specifically including: and inputting the sample node label and the sample node information into a node loss function to obtain a node loss value. And inputting the sample topological structure label and the sample topological structure information into a topological structure loss function to obtain a topological structure loss value. And inputting the sample edge label and the sample edge information into an edge loss function to obtain an edge loss value. And weighting and summing the node loss value, the topological structure loss value and the edge loss value to obtain a comprehensive loss value. And adjusting parameters of the structure chart generation model based on the comprehensive loss value until the comprehensive loss value is converged. The model at which the integrated loss value converges is used as a trained structure diagram to generate a model.

By using the training method for the structure diagram generation model provided by the embodiment of the disclosure, the context coding sub-model, the node label generation sub-model, the topology structure generation sub-model and the side label generation sub-model can be jointly trained by using the correlation among the topology structure information, the side information and the node information, so that the training speed is increased, and the precision of the trained structure diagram generation model is improved.

Fig. 7 schematically illustrates a flow chart of an atlas generation method according to an embodiment of the disclosure.

As shown in fig. 7, the method includes operation S710.

In operation S710, a target map is generated based on the structure map. The structure map is generated using a structure map generation method.

According to the embodiment of the disclosure, generating the target map based on the structure diagram may further include: and carrying out node combination on a plurality of nodes in the structure graph based on the topological structure information and the side information in the structure graph to generate a phrase structure graph. And extracting information from the phrase structure diagram to generate a target map.

According to embodiments of the present disclosure, the target profile may include one or more of a knowledge profile, a fact profile.

According to the embodiment of the disclosure, the structure diagram can be subjected to node combination by using a combination model to generate a phrase structure diagram. The combination model may include a deep learning model, as long as the deep learning model can process the structure graph to generate the phrase structure graph. But is not limited thereto. And combining part of nodes in the structure diagram according to rules to form phrase nodes so as to obtain the phrase structure diagram.

According to the embodiment of the disclosure, the phrase structure diagram can be subjected to information extraction by using an information extractor, so as to generate the target map. The information extractor may include a deep learning model, as long as the deep learning model can process the phrase structure diagram to generate the target map.

According to the embodiment of the disclosure, the structure diagram is used for generating the target map, the method is applied to the technical field of information extraction, the content of the knowledge map can be supplemented in the construction process of the knowledge map, and the richness of the knowledge map is improved. And the ability of combing the incidence relation among a plurality of events can be improved in the construction process of the affair map.

In summary, by using the structure diagram generation method provided by the embodiment of the present disclosure, statements to be processed in different fields can be quickly and accurately processed to generate a structure diagram. The structure chart is applied to the map generation field, and the generation speed of the new field map is improved.

Fig. 8 schematically shows a block diagram of a structure diagram generation apparatus according to an embodiment of the present disclosure.

As shown in FIG. 8, structure diagram generation apparatus 800 may include an encoding module 810, a determination module 820, and a structure diagram generation module 830.

The encoding module 810 is configured to perform context encoding on the statement to be processed to obtain an encoding vector sequence.

A determining module 820, configured to determine node information, topology information, and side information for generating the structure diagram based on the coding vector sequence. The node information is used for representing attribute information of the nodes in the structure diagram, the topological structure information is used for representing whether edges exist among the nodes, and the edge information is used for representing the incidence relation among the nodes.

And a structure diagram generating module 830, configured to generate a target structure diagram for the to-be-processed statement based on the node information, the topology information, and the side information.

According to an embodiment of the present disclosure, the determination module may include a first transformation unit, a second transformation unit, a first determination unit, a second determination unit, and a third determination unit.

And the first transformation unit is used for carrying out first transformation processing on the coding vector sequence to obtain topological structure representation.

And the second transformation unit is used for carrying out second transformation processing on the coding vector sequence to obtain the relational expression.

A first determining unit for determining node information based on the coded vector sequence.

A second determining unit for determining topology information based on the topology representation.

A third determining unit for determining the side information based on the relational representation.

According to an embodiment of the present disclosure. The statement to be processed comprises a plurality of nodes, the coding vector sequence comprises a plurality of coding vectors, and the plurality of nodes correspond to the plurality of coding vectors one to one.

According to an embodiment of the present disclosure, the first transform unit may include a first determination subunit, and a first transform subunit.

And the first determining subunit is used for determining a coding vector pair corresponding to any two nodes from the coding vector sequence aiming at the two nodes in the sentence to be processed.

And the first transformation subunit is used for carrying out first transformation processing on the coding vector pair to obtain a topological structure representation between the two nodes.

According to an embodiment of the present disclosure, the second transformation unit may include a second determination subunit, and a second transformation subunit.

And the second determining subunit is used for determining a coding vector pair corresponding to any two nodes in the statement to be processed from the coding vector sequence.

And the second transformation subunit is used for carrying out second transformation processing on the coding vector pair based on the class number of the side information to obtain a relation expression between the two nodes.

According to an embodiment of the present disclosure, the first determination unit includes an attribute processing subunit.

And the attribute processing subunit is used for determining the node information by performing node attribute processing on the coding vector sequence.

According to an embodiment of the present disclosure, the node information includes at least one of: node part-of-speech information, and derived node information for characterizing whether to generate derived nodes.

FIG. 9 schematically illustrates a block diagram of a training apparatus that constructs a model according to an embodiment of the disclosure.

As shown in FIG. 9, the apparatus 900 for training a structure-diagram-generating model may include a first input module 910, a second input module 920, a third input module 930, a fourth input module 940, and a training module 950.

The first input module 910 is configured to input a sample statement corresponding to the sample structure diagram to the context coding sub-model, so as to obtain a sample coding vector sequence. The sample structure graph comprises a sample node label, a sample topological structure label and a sample edge label. The structure diagram generation model comprises a context coding sub-model, a node label generation sub-model, a topological structure generation sub-model and an edge label generation sub-model.

And a second input module 920, configured to input the sample coding vector sequence to the node label generation submodel, so as to obtain sample node information.

A third input module 930, configured to input the sample statement to the topology generation submodel, so as to obtain sample topology information.

And a fourth input module 940, configured to input the sample statement to the edge tag generation submodel, so as to obtain sample edge information.

The training module 950 is configured to train the structure diagram generation model based on the sample structure diagram, the sample node information, the sample topology structure information, and the sample side information, to obtain a trained structure diagram generation model.

Figure 10 schematically shows a block diagram of an atlas generation apparatus according to an embodiment of the disclosure.

As shown in fig. 10, the atlas generation apparatus 10 may include an atlas generation module 1010.

And the map generation module 1010 is used for generating a target map based on the structure map. The configuration diagram is generated by a configuration diagram generating device.

The present disclosure also provides an electronic device, a readable storage medium, and a computer program product according to embodiments of the present disclosure.

According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method according to an embodiment of the disclosure.

According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium having stored thereon computer instructions for causing a computer to perform a method as in an embodiment of the present disclosure.

According to an embodiment of the disclosure, a computer program product comprising a computer program which, when executed by a processor, implements a method as in an embodiment of the disclosure.

FIG. 11 shows a schematic block diagram of an example electronic device 1100 that may be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processors, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 11, the device 1100 comprises a computing unit 1101, which may perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM)1102 or a computer program loaded from a storage unit 1108 into a Random Access Memory (RAM) 1103. In the RAM 1103, various programs and data necessary for the operation of the device 1100 may also be stored. The calculation unit 1101, the ROM 1102, and the RAM 1103 are connected to each other by a bus 1104. An input/output (I/O) interface 1105 is also connected to bus 1104.

A number of components in device 1100 connect to I/O interface 1105, including: an input unit 1106 such as a keyboard, a mouse, and the like; an output unit 1107 such as various types of displays, speakers, and the like; a storage unit 1108 such as a magnetic disk, optical disk, or the like; and a communication unit 1109 such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 1109 allows the device 1100 to exchange information/data with other devices through a computer network such as the internet and/or various telecommunication networks.

The computing unit 1101 can be a variety of general purpose and/or special purpose processing components having processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various dedicated Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, and the like. The calculation unit 1101 performs each method and process described above, such as a structure diagram generation method, a training method of a structure diagram generation model, or a map generation method. For example, in some embodiments, the architectural graph generation method, the training method for the architectural graph generation model, or the atlas generation method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 1108. In some embodiments, part or all of the computer program may be loaded and/or installed onto device 1100 via ROM 1102 and/or communication unit 1109. When loaded into RAM 1103 and executed by computing unit 1101, a computer program may perform one or more steps of the architectural diagram generation method, the training method of an architectural diagram generation model, or the atlas generation method described above. Alternatively, in other embodiments, the computing unit 1101 may be configured in any other suitable manner (e.g., by means of firmware) to perform the architectural graph generation method, the training method of the architectural graph generation model, or the atlas generation method.

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), Complex Programmable Logic Devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program code, when executed by the processor or controller, causes the functions/acts specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.

The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server with a combined blockchain.

It should be understood that various forms of the flows shown above, reordering, adding or deleting steps, may be used. For example, the steps described in the present disclosure may be executed in parallel, sequentially, or in different orders, and are not limited herein as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved.

The above detailed description should not be construed as limiting the scope of the disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A method for generating a structure diagram comprises the following steps:

carrying out context coding on a statement to be processed to obtain a coding vector sequence;

determining node information, topological structure information and side information for generating a structure diagram based on the coding vector sequence, wherein the node information is used for representing attribute information of nodes of the structure diagram, the topological structure information is used for representing whether edges exist among a plurality of nodes, and the side information is used for representing the association relationship among the nodes; and

And generating a target structure chart aiming at the statement to be processed based on the node information, the topological structure information and the side information.

2. The method of claim 1, wherein the determining node information, topology information, and side information for generating a structure graph based on the sequence of encoded vectors comprises:

performing first transformation processing on the coding vector sequence to obtain a topological structure representation;

performing second transformation processing on the coding vector sequence to obtain relational expression;

determining the node information based on the sequence of encoded vectors;

determining the topological structure information based on the topological structure representation; and

determining the side information based on the relational representation.

3. The method of claim 2, wherein the statement to be processed comprises a plurality of nodes, the sequence of code vectors comprises a plurality of code vectors, and the plurality of nodes are in one-to-one correspondence with the plurality of code vectors;

the performing a first transformation process on the coding vector sequence to obtain a topological structure representation includes:

for any two nodes in the statement to be processed, determining a code vector pair corresponding to the two nodes from the code vector sequence; and

And carrying out first transformation processing on the coding vector pair to obtain a topological structure representation between the two nodes.

4. The method of claim 2 or 3, wherein the statement to be processed comprises a plurality of nodes, the sequence of code vectors comprises a plurality of code vectors, and the plurality of nodes are in one-to-one correspondence with the plurality of code vectors;

performing a second transform on the coded vector sequence to obtain a relationship representation, including:

and performing second transformation processing on the coding vector pair based on the class number of the side information to obtain a relation expression between the two nodes.

5. The method of any of claims 1-4, wherein the determining the node information based on the sequence of encoded vectors comprises:

determining the node information by performing node attribute processing on the coding vector sequence;

wherein the node information comprises at least one of:

node part-of-speech information and derived node information used for representing whether derived nodes are generated or not.

6. A structure chart generation model training method comprises a context coding submodel, a node label generation submodel, a topology structure generation submodel and an edge label generation submodel, and comprises the following steps:

inputting a sample statement corresponding to a sample structure chart into the context coding sub-model to obtain a sample coding vector sequence, wherein the sample structure chart comprises a sample node label, a sample topological structure label and a sample edge label;

inputting the sample coding vector sequence into the node label generation submodel to obtain sample node information;

inputting the sample statement into the topological structure generation submodel to obtain sample topological structure information;

inputting the sample statement into the side label generation submodel to obtain sample side information; and

training the structure diagram generation model based on the sample structure diagram, the sample node information, the sample topological structure information and the sample side information to obtain a trained structure diagram generation model.

7. A method of atlas generation comprising:

generating a target map based on the structure map,

Wherein the structural map is generated using the structural map generation method according to any one of claims 1 to 5.

8. A structural chart generating apparatus comprising:

the encoding module is used for carrying out context encoding on the statement to be processed to obtain an encoding vector sequence;

the determining module is used for determining node information, topological structure information and side information which are used for generating the structure chart based on the coding vector sequence, wherein the node information is used for representing attribute information of the nodes of the structure chart, the topological structure information is used for representing whether edges exist among a plurality of nodes, and the side information is used for representing the association relation among the plurality of nodes; and

and the structure chart generation module is used for generating a target structure chart aiming at the statement to be processed based on the node information, the topological structure information and the side information.

9. The apparatus of claim 8, wherein the means for determining comprises:

the first transformation unit is used for carrying out first transformation processing on the coding vector sequence to obtain a topological structure representation;

the second transformation unit is used for carrying out second transformation processing on the coding vector sequence to obtain relational expression;

A first determining unit configured to determine the node information based on the coded vector sequence;

a second determining unit configured to determine the topological structure information based on the topological structure representation; and

10. The apparatus of claim 9, wherein the to-be-processed statement comprises a plurality of nodes, the sequence of code vectors comprises a plurality of code vectors, and the plurality of nodes are in one-to-one correspondence with the plurality of code vectors;

the first transform unit includes:

a first determining subunit, configured to determine, for any two nodes in the statement to be processed, a pair of coded vectors corresponding to the two nodes from the coded vector sequence; and

11. The apparatus according to claim 9 or 10, wherein the statement to be processed comprises a plurality of nodes, the sequence of code vectors comprises a plurality of code vectors, and the plurality of nodes are in one-to-one correspondence with the plurality of code vectors;

The second transform unit includes:

a second determining subunit, configured to determine, for any two nodes in the statement to be processed, a coded vector pair corresponding to the two nodes from the coded vector sequence; and

and the second transformation subunit is used for carrying out second transformation processing on the coding vector pair based on the category number of the side information to obtain a relation expression between the two nodes.

12. The apparatus of any of claims 8 to 11, wherein the first determining unit comprises:

an attribute processing subunit, configured to determine the node information by performing node attribute processing on the coded vector sequence;

wherein the node information comprises at least one of:

node part-of-speech information, and derived node information for characterizing whether to generate derived nodes.

13. A training device of a structure diagram generation model, wherein the structure diagram generation model comprises a context coding submodel, a node label generation submodel, a topological structure generation submodel and an edge label generation submodel, and the training device of the structure diagram generation model comprises:

the first input module is used for inputting a sample statement corresponding to a sample structure diagram into the context coding sub-model to obtain a sample coding vector sequence, wherein the sample structure diagram comprises a sample node label, a sample topological structure label and a sample edge label;

The second input module is used for inputting the sample coding vector sequence into the node label generation submodel to obtain sample node information;

the third input module is used for inputting the sample statement into the topological structure generation submodel to obtain sample topological structure information;

a fourth input module, configured to input the sample statement to the edge tag generation submodel, so as to obtain sample edge information; and

and the training module is used for training the structure diagram generation model based on the sample structure diagram, the sample node information, the sample topological structure information and the sample side information to obtain a trained structure diagram generation model.

14. An atlas generation apparatus comprising:

the map generation module is used for generating a target map based on the structure map,

wherein the configuration diagram is generated using the configuration diagram generating apparatus according to any one of claims 8 to 12.

15. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method for structure diagram generation as defined in any one of claims 1 to 5, a method for training a structure diagram generation model as defined in claim 6, or a method for atlas generation as defined in claim 7.

16. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the structural graph generating method according to any one of claims 1 to 5, the training method of the structural graph generating model according to claim 6, or the atlas generating method according to claim 7.

17. A computer program product comprising a computer program which, when executed by a processor, implements the architectural graph generation method of any one of claims 1 to 5, the training method of an architectural graph generation model of claim 6, or the atlas generation method of claim 7.