CN116601626A - Personal knowledge graph construction method and device and related equipment - Google Patents

Abstract

A method for constructing and updating personal knowledge graph in artificial intelligence field includes: acquiring initial user static attributes and initial user behavior attributes, and constructing a personal knowledge graph architecture according to the initial user static attributes and the initial user behavior attributes; acquiring initial user portrait data, initial user behavior data and entity relation sets; obtaining M initial static entity triples according to the initial user portrait data and the entity relation set; e initial behavior entity triples are obtained according to the initial user behavior data and the entity relation set; and generating a personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples and the personal knowledge graph framework. By adopting the method, the personal knowledge graph highly related to the personal characteristics of the user can be constructed, so that intelligent recommendation service is performed on the user according to the personal knowledge graph.

Description

Personal knowledge graph construction method and device and related equipment Technical Field

The application relates to the technical field of artificial intelligence, in particular to a personal knowledge graph construction method, a device and related equipment.

Background

Artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use the knowledge to obtain optimal results. In other words, artificial intelligence is a branch of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar manner to human intelligence. Artificial intelligence, i.e. research on design principles and implementation methods of various intelligent machines, enables the machines to have functions of sensing, reasoning and decision. Research in the field of artificial intelligence includes robotics, natural language processing, computer vision, decision and reasoning, man-machine interaction, recommendation and search, AI-based theory, and the like.

Knowledge Graph (KG) is an artificial intelligence domain method for describing concepts, entities and their relationships in the objective world in a structured form, and representing massive information generated on the internet in the form of "knowledge". The existing knowledge graph comprises two types: a general knowledge graph and a homeodomain knowledge graph. The universal knowledge graph has wide coverage range and is a structured encyclopedia knowledge base for all fields; the vertical domain knowledge graph is a knowledge base constructed by means of data of a specific domain and has specific industry significance.

The knowledge maps of the two types have wider coverage knowledge and low relevance with individual users; and the data volume is large, portability is poor, and maintenance is difficult.

Disclosure of Invention

The embodiment of the application discloses a personal knowledge graph construction method, a device and related equipment.

In a first aspect, an embodiment of the present application provides a method for constructing a personal knowledge graph, including:

acquiring initial user static attributes and initial user behavior attributes, and constructing a personal knowledge graph architecture according to the initial user static attributes and the initial user behavior attributes; the initial user static attribute is used for representing personal information of a user, the initial user behavior attribute is used for representing a field to which an entity type corresponding to user behavior belongs, the field is a set of entity types with the same characteristics, the entity types are a set of entities with the same characteristics or attributes, the entities are things which are associated with the user and are represented by nouns or words, the personal knowledge graph architecture represents a relationship between the user and the personal information, and the user and the field to which the entity type corresponding to the user behavior belongs; acquiring initial user portrait data, initial user behavior data and an entity relation set, wherein the initial user portrait data corresponds to the initial user static attribute, the initial user behavior data corresponds to the initial user behavior attribute, and the entity relation set comprises a plurality of relations which are used for representing the relation between the entities; obtaining M initial static entity triples according to the initial user portrait data and the entity relation set; e initial behavior entity triples are obtained according to the initial user behavior data and the entity relation set; the static entity triplets are used for describing attribute relations between users and static entities or between static entities, the behavior entity triplets are used for describing behavior relations between users and behavior entities or between behavior entities, and M and E are integers which are larger than or equal to zero; and generating a personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples and the personal knowledge graph framework.

It should be understood that the above-described domain corresponding to the user behavior refers to a domain corresponding to the operation content corresponding to the user behavior, and the domain may include music, movies, news, games, and the like; the entity comprises a static entity and a behavior entity, wherein the static entity is identified from user portrait data, and the behavior entity is identified from user behavior data; the entity triples are structured data forms for representing the relationship between two entities, and the entity triples comprise three types of static entity triples, behavior entity triples and target entity triples, and each entity triplet comprises two entities, entity types corresponding to the two entities and the relationship between the two entities.

It can be seen that, in the embodiment of the present application, a personal knowledge graph architecture is constructed by adopting static attributes representing personal information of a user and behavioral attributes representing behavioral characteristics of the user, where the architecture is an architecture taking a personal user as a center and representing characteristics of the personal user; then, respectively obtaining a static entity triplet representing the static attribute of the user and a behavior entity triplet representing the behavior attribute of the user based on the user portrait data and the behavior data; therefore, the personal knowledge graph generated according to the static entity triples, the behavior entity triples and the personal knowledge graph architecture can accurately represent the personal characteristics of the user.

In a possible implementation manner, the obtaining M initial static entity triples according to the initial user portrait data and the entity relation set includes: obtaining A static entities from initial user portrait data, wherein A is an integer greater than or equal to zero; and obtaining M initial static entity triples according to the A static entities and the entity relation set.

It can be seen that in the embodiment of the present application, the attribute value of the user is obtained from the user portrait data that characterizes the personal information of the user; then, according to the entity relation set and the acquired attribute values, structured data representing personal information of the user, namely a static entity triplet, is obtained; the structured static entity triples are more convenient for data storage and management, so that the efficiency of the personal knowledge graph construction process is improved.

In a possible implementation manner, the obtaining E initial behavior entity triples according to the initial user behavior data and the entity relationship set includes: acquiring B behavior entities and C behavior characters of a user from initial user behavior data, wherein the C behavior characters correspond to O behavior entities in the B behavior entities, the behavior characters are used for representing the operation of the user on the O behavior entities, and B, C and O are integers larger than or equal to zero; obtaining G initial behavior entity triples according to the B behavior entities and the entity relation set; obtaining H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities and the entity relation set, wherein E is equal to the sum of G and H, and G and H are integers greater than or equal to zero; the G initial behavior entity triples are used for describing the relationship among the B behavior entities; the H initial behavior entity triples are used to describe the relationship between the user and the O behavior entities.

It should be noted that the number of behavioural characters and the number of behavioural entities are not in one-to-one correspondence. One behavior character may correspond to a plurality of behavior entities, for example, the behavior character is "click", and the behavior entity corresponding to the behavior character may include "news 1", "news 2", "red date (song name)", and the like; the plurality of behavior characters may also correspond to one behavior entity, for example, the behavior characters may be "click" and "comment", and the behavior entities corresponding to the two behavior characters may be "news 1".

It can be seen that, in the embodiment of the present application, a plurality of behavior entities are obtained from the behavior data of the user, and then, based on the plurality of behavior entities and the entity relationship set, a plurality of structured behavior entity triples are obtained, the plurality of behavior entity triples are characterizations of behavior characteristics of the user, and the representation forms of the behavior entity triples and the static entity triples are the same, so that unification of data forms is realized, and data storage and management are facilitated, thereby improving efficiency of a personal knowledge graph construction process.

In a possible implementation manner, the personal knowledge graph construction method further includes: acquiring H operation times from initial user behavior data; the obtaining H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities and the entity relation set comprises the following steps: and generating H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities, the H operation times and the entity relation set.

It should be appreciated that, since one behavior character may correspond to a plurality of behavior entities, and a plurality of behavior characters may correspond to one behavior entity, the number of operation times may not be equal to the number of behavior characters or behavior entities, where the operation times are used to characterize a specific moment when a user operates on a behavior entity, and the H operation times and the H first behavior entity triples are in one-to-one correspondence.

It can be seen that, in the embodiment of the application, by acquiring the operation time corresponding to the behavior character of the user and fusing the operation time into the behavior entity triples, a plurality of behavior entity triples related to the behavior of the user are obtained, and the behavior entity triples have the operation time, so that the behavior characteristics of the user can be more accurately represented; and then the user personal knowledge graph obtained according to the behavior entity triples is more accurate and perfect.

In a possible implementation manner, the generating a personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples, and the personal knowledge graph architecture includes: generating X initial target entity triples according to M initial static entity triples and/or E initial behavior entity triples, wherein X is an integer greater than or equal to zero; each of the target entity triples comprises at least one target entity, the target entity triples describing a relationship between the target entities, or between a user and the target entities; and generating a personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples, the X initial target entity triples and the personal knowledge graph architecture.

It should be appreciated that the target entity triples are causally inferred based on static entity triples and/or behavioral entity triples, and that the target entity triples characterize commonalities between static entity triples and/or behavioral entity triples, i.e., deep personal characteristics of the user.

It can be seen that, in the embodiment of the present application, the target entity triples are used to represent deep user personal characteristics, and then the target entity triples are used to participate in generating a personal knowledge graph, so that the personal knowledge graph can accurately represent the deep personal characteristics of the user, and further, related content can be accurately pushed to the user according to the personal knowledge graph.

In one possible implementation, generating a personal knowledge graph from M initial static entity triples, E initial behavioral entity triples, and a personal knowledge graph architecture includes: generating a personal knowledge graph according to the M initial static entity triples, F initial behavior entity triples in the E initial behavior entity triples and the personal knowledge graph framework; the F initial behavior entity triples are obtained according to first user behavior data and entity relation sets, wherein the first user behavior data are behavior data, the difference value between the acquisition time and the current system time in the initial user behavior data is smaller than a preset threshold value, and F is an integer larger than or equal to zero.

It can be seen that, in the embodiment of the present application, since the collection time period of the first user behavior data for generating the F initial behavior entity triples is located in the collection time period of the initial user behavior data, and the collection time of the first user behavior data is close to the current system time, the F initial behavior entity triples can accurately represent the behavior characteristics of the user in the last period of time, so that the user personal knowledge graph generated subsequently according to the F initial behavior entity triples has better timeliness.

In a possible implementation manner, the generating a personal knowledge graph according to the M initial static entity triples, the F initial behavior entity triples in the E initial behavior entity triples, and the personal knowledge graph architecture includes: performing de-duplication on the F initial behavior entity triples to obtain I initial behavior entity triples, wherein I is an integer which is less than or equal to F and is greater than or equal to zero; when the I is smaller than or equal to the preset quantity Q, generating a personal knowledge graph according to M initial static entity triples, I initial behavioral entity triples and the personal knowledge graph framework, wherein Q is an integer larger than or equal to zero; when the I is larger than the preset quantity Q, grouping F initial behavior entity triples according to the I initial behavior entity triples to obtain O behavior entity triplet sets, wherein the O behavior entity triplet sets are in one-to-one correspondence with the I initial behavior entity triples, and the initial behavior entity triples contained in each behavior entity triplet set are identical to the initial behavior entity triples corresponding to the behavior entity triplet sets; sorting the I initial behavior entity triples according to the number of elements in each behavior entity triplet set to obtain a sorting result, wherein when the number of elements in each behavior entity triplet set is larger, the sorting position of the initial behavior entity triplet corresponding to each behavior entity triplet set in the sorting result is higher; and generating a personal knowledge graph according to the M initial static entity triples, the first Q initial behavior entity triples in the ordered I initial behavior entity triples and the personal knowledge graph framework.

It can be seen that in the embodiment of the present application, each behavioural entity triplet is firstly ordered according to the frequency of occurrence of each behavioural entity triplet in order from high to low; then Q behavioral entity triples are selected from the ordered behavioral entity triples; the high-frequency behavior characteristics of the user can be represented by obtaining the preset number of behavior entity triples in the mode, so that the behavior characteristics of the user can be accurately represented according to the personal knowledge graph generated by the Q entity triples; in addition, the preset quantity is set, so that the quantity of the behavioral entity triples for generating the personal knowledge graph can be controlled, the excessive scale of the personal knowledge graph is avoided, and the feasibility of using the personal knowledge graph on intelligent terminal equipment and the portability of the personal knowledge graph are ensured.

In one possible implementation, generating a personal knowledge graph from M initial static entity triples, E initial behavioral entity triples, X initial target entity triples, and a personal knowledge graph architecture includes: generating a personal knowledge graph according to M initial static entity triples, F initial behavior entity triples in E initial behavior entity triples, Y initial target entity triples in X initial target entity triples and a personal knowledge graph framework; the F initial behavior entity triples are obtained according to first user behavior data and entity relation sets, wherein the first user behavior data are behavior data, and the difference value between the acquisition time and the current system time in the initial user behavior data is smaller than a preset threshold value; the Y initial target entity triples are generated according to M initial static entity triples and/or F initial behavior entity triples, and Y is an integer greater than or equal to zero.

It can be seen that, in the embodiment of the present application, since the collection time period of the first user behavior data for generating the F initial behavior entity triples is located in the collection time period of the initial user behavior data, the Y initial target entity triples are generated according to the M static entity triples and the F initial behavior entity triples, and the collection time of the first user behavior data is close to the current system time, the Y initial target entity triples and the F initial behavior entity triples can respectively represent the deep personal characteristics and behavior characteristics of the user in the last period of time, so that the user personal knowledge graph generated according to the F initial behavior entity triples and the Y initial target entity triples can accurately represent the behavior characteristics and the deep personal characteristics of the user, and has better timeliness.

In one possible implementation, generating a personal knowledge graph from M initial static entity triples, F initial behavior entity triples of the E initial behavior entity triples, Y initial target entity triples of the X initial target entity triples, and a personal knowledge graph architecture includes: performing de-duplication on the Y initial target entity triples to obtain J initial target entity triples, wherein J is an integer which is less than or equal to Y and is greater than or equal to zero; when J is equal to the preset number Q, generating a personal knowledge graph according to M initial static entity triples, J initial target entity triples and the personal knowledge graph framework, wherein Q is an integer greater than or equal to zero; when J is greater than the preset quantity Q, grouping Y initial target entity triples according to J initial target entity triples to obtain J target entity triplet sets, wherein the J target entity triplet sets and the J initial target entity triples are in one-to-one correspondence, and the initial target entity triples contained in each target entity triplet set are identical to the initial target entity triples corresponding to the target entity triplet sets; according to the number of elements in each target entity triplet set, sorting J initial target entity triples to obtain a sorting result, wherein when the number of elements in each target entity triplet set is larger, the sorting position of the initial target entity triplet corresponding to each target entity triplet set is higher in the sorting result; generating a personal knowledge graph according to M initial static entity triples, the first Q initial target entity triples in the sorted J initial target entity triples and the personal knowledge graph framework; when J is smaller than the preset quantity Q, F initial behavior entity triples are grouped according to the I initial behavior entity triples to obtain O behavior entity triples, the O behavior entity triples correspond to the I initial behavior entity triples one by one, and the initial behavior entity triples contained in each behavior entity triples are identical to the initial behavior entity triples corresponding to the behavior entity triples; sorting the I initial behavior entity triples according to the number of elements in each behavior entity triplet set to obtain a sorting result, wherein when the number of the initial behavior entity triples contained in each behavior entity triplet set is larger, the sorting position of the initial behavior entity triplet corresponding to each behavior entity triplet set is higher in the result; and generating a personal knowledge graph according to the first Q-J initial behavior entity triples and the personal knowledge graph framework in the M initial static entity triples, the J initial target entity triples and the ordered I initial behavior entity triples.

It can be seen that in the embodiment of the present application, first, according to the occurrence frequency of each entity triplet, each entity triplet is ordered according to the order of the occurrence frequency of each entity triplet from top to bottom; then selecting from the ordered target entity triples, and selecting from the ordered behavior entity triples to select Q entity triples; because the target entity triples are used for representing the deep user personal characteristics, and the behavior entity triples are used for representing the user behavior characteristics, the Q entity triples obtained by adopting the method have higher correlation with the user, and the personal characteristics of the user can be more accurately represented according to the personal knowledge graph generated by the Q entity triples; meanwhile, the number of entity triples for generating the personal knowledge graph is controlled by setting the preset number, so that the excessive scale of the personal knowledge graph can be avoided, and the feasibility of using the personal knowledge graph on intelligent terminal equipment is ensured.

In a possible implementation manner, the personal knowledge graph construction method further includes: acquiring a target user static attribute and a target user behavior attribute, and updating the personal knowledge graph by utilizing the target user static attribute and the target user behavior attribute to obtain a target personal knowledge graph; acquiring second user portrait data and second user behavior data; obtaining K first static entity triples according to the second user portrait data and the entity relation set; obtaining L first behavior entity triples according to the second user behavior data and the entity relation set, wherein L and K are integers greater than or equal to zero; and updating the target personal knowledge graph according to the K first static entity triples and the L first behavioral entity triples.

It should be appreciated that the collection time periods for the target user static attribute, the target user behavior attribute, the second user portrayal data, and the second user behavior data are the same and later than the collection time periods for the initial user static attribute, the initial user behavior attribute, the initial user portrayal data, and the initial user behavior data described above.

It can be seen that, in the embodiment of the present application, since characteristics such as behaviors and personal preferences of a user may change over time, the personal knowledge graph is updated by using the static attribute of the target user and the behavioral attribute of the target user, the obtained target personal knowledge graph can accurately represent the category of the static characteristic and the behavioral characteristic of the user in a recent period, then the static entity triplet and the behavioral entity triplet that represent the personal information and the behavioral characteristic of the user in the recent period are respectively generated by using the second user portrait data and the second user behavioral data, and finally the target personal knowledge graph is updated by using the generated static entity triplet and behavioral entity triplet, so that the updated target personal knowledge graph can more accurately represent the personal characteristics of the user in the recent period.

In a possible implementation manner, the personal knowledge graph construction method further includes: acquiring third user portrait data and third user behavior data; obtaining R second static entity triples according to the third user portrait data and the entity relation set; obtaining S second behavior entity triples according to the second user behavior data and the entity relation set, wherein R and S are integers greater than or equal to zero; and updating the personal knowledge graph according to the R second static entity triples and the S second behavior entity triples.

It can be seen that, in the embodiment of the present application, since characteristics such as behaviors and personal preferences of a user may change over time, first, static entity triples and behavioral entity triples that characterize personal information and behavior characteristics of the user in a recent period of time are generated respectively by using third user portrait data and third user behavioral data, and then, personal knowledge maps are updated by using the generated static entity triples and behavioral entity triples, so that the updated personal knowledge maps can more accurately characterize the personal characteristics of the user in the recent period of time.

In a second aspect, an embodiment of the present application provides a personal knowledge graph construction apparatus, where the apparatus includes a functional module that performs part or all of the method according to the first aspect.

In a third aspect, an embodiment of the present application provides a personal knowledge graph construction apparatus, including: a memory for storing a program; a processor for executing the program stored in the memory, the processor being for executing part or all of the method of the first aspect when the program stored in the memory is executed.

In a fourth aspect, embodiments of the present application provide a computer readable medium storing program code for execution by a device, the program code comprising instructions for performing part or all of the method of the first aspect.

In a fifth aspect, embodiments of the present application provide a computer program product comprising instructions which, when run on a computer, cause the computer to perform some or all of the method of the first aspect described above.

In a sixth aspect, an embodiment of the present application provides a chip, where the chip includes a processor and a data interface, where the processor reads an instruction stored on a memory through the data interface, and performs some or all of the method in the first aspect.

Optionally, as an implementation manner, the chip may further include a memory, where the memory stores instructions, and the processor is configured to execute the instructions stored on the memory, where the instructions, when executed, are configured to perform some or all of the method in the first aspect.

In a seventh aspect, an embodiment of the present application provides an electronic device, where the electronic device includes some or all of the apparatus in any one of the second aspect or the third aspect.

Drawings

FIG. 1A is a system for constructing a personal knowledge graph, provided by an embodiment of the present application;

FIG. 1B is a diagram of another system for constructing a personal knowledge graph in accordance with an embodiment of the application;

FIG. 1C is a further system for constructing a personal knowledge graph in accordance with an embodiment of the application;

FIG. 2 is a system for training a neural network model provided by an embodiment of the present application;

FIG. 3 is a further system for constructing a personal knowledge graph in accordance with an embodiment of the application;

FIG. 4 is a schematic diagram of a convolutional neural network according to an embodiment of the present application;

FIG. 5 is a schematic diagram of another convolutional neural network according to an embodiment of the present application;

FIG. 6 is a schematic diagram of a representation and examples of an entity triplet according to an embodiment of the present application;

FIG. 7 is a schematic diagram of a representation and an example of an entity quad according to an embodiment of the present application;

FIG. 8 is a schematic diagram of another representation and example of an entity triplet provided by an embodiment of the present application;

FIG. 9 is a schematic flow chart of a method for constructing a personal knowledge graph according to an embodiment of the present application;

fig. 10 is a schematic structural diagram of a personal knowledge graph architecture according to an embodiment of the present application;

FIG. 11 is a schematic flow chart of user behavior data collection according to an embodiment of the present application;

FIG. 12 is a schematic diagram of a process for constructing a behavioral entity triplet according to an embodiment of the present application;

FIG. 13 is a schematic diagram of a fusion flow of a behavioral entity triplet and a behavioral entity triplet according to an embodiment of the present application;

FIG. 14 is a schematic diagram of a process for constructing a target entity triplet according to an embodiment of the present application;

FIG. 15 is a schematic diagram of a construction process of another target entity triplet according to an embodiment of the present application;

FIG. 16 is a schematic diagram of a process for constructing a target entity triplet according to an embodiment of the present application;

FIG. 17 is a schematic flow chart of expanding an entity triplet to be expanded to a personal knowledge graph according to an embodiment of the present application;

FIG. 18 is a flowchart of a personal knowledge graph updating method according to an embodiment of the present application;

FIG. 19 is a flowchart of another method for updating a personal knowledge graph according to an embodiment of the present application;

FIG. 20 is a schematic block diagram of a personal knowledge graph construction device, provided by an embodiment of the application;

fig. 21 is a schematic hardware structure of a personal knowledge graph construction device according to an embodiment of the present application.

Detailed Description

The following description of the technical solutions according to the embodiments of the present application will be given with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application. It should be noted that in the present application, the terms "action", "action relationship" and "operation behavior" are all terms for describing actions of a user on a behavior entity, that is, for characterizing a relationship between a user and a behavior entity, and may be equivalent in the embodiments of the present application.

As shown in fig. 1A, a system for constructing a personal knowledge graph includes a user device and a data processing device.

The user equipment comprises intelligent terminals such as a user, a mobile phone, a personal computer, a tablet or an information processing center. The user device is the initiating end of the personal knowledge graph construction process, and typically the user initiates the request through the user device.

The data processing device may be a device or a server having a data processing function, such as a cloud server, a web server, an application server, and a management server. The data processing equipment receives user behavior data from the intelligent terminal through the interactive interface, and then performs data processing in modes of machine learning, deep learning and the like through a memory for storing the data and a processor link for data processing. The memory may be a generic term comprising a database of locally stored as well as stored history data, which may be on the data processing device or on other network servers. The above user behavior data may be a semi-structured dotting log, as shown in fig. 11, or may be other data for recording user behaviors, which is not limited in particular in the embodiment of the present application.

Fig. 1B shows another application scenario of the system for constructing a personal knowledge graph according to an embodiment of the present application. In this scenario, the intelligent terminal directly serves as a data processing device, directly receives input from a user, and directly processes the input by hardware and/or software of the intelligent terminal, and a specific process is similar to that of fig. 1A, and reference may be made to the above description, which is not repeated herein.

As shown in fig. 1C, the user device may be a local device 101 or 102, the data processing device may be an executing device 110, where the data storage system 150 may be integrated on the executing device 110, or may be disposed on a cloud or other network server, and the executing device 110 may be integrated on the local device 101 or 102.

Referring to fig. 2, a system architecture 200 for training a neural network model is provided in accordance with an embodiment of the present application. The data collection device 260 is configured to collect training data and store the training data in the database 230, and the training device 220 generates the entity recognition model/causal inference model 201 based on the training data maintained in the database 230. How the training device 220 obtains the entity recognition model/causal inference model 201 based on the training data, where the entity recognition model can recognize the entities in the user behavior data and the entity types corresponding to the entities, and the causal inference model can mine the target entity triples reflecting the deep personal characteristics of the user based on the static entity triples and/or the behavioral entity triples will be described in more detail below.

Fig. 2 is a functional block diagram in a data processing process, and when the functional block diagram corresponds to the actual application scenario diagram in fig. 1A or fig. 1B, the user equipment 240 may be the user equipment in fig. 1A or fig. 1B, and the execution device 210 and the data storage system 250 may be integrated in the user equipment when the data processing capability of the user equipment in fig. 1A or fig. 1B is relatively strong. In some possible embodiments, the execution device 210 and the data storage system 250 may also be integrated on the data processing device in FIG. 1A or FIG. 1B. The database 230, training device 220, and data collection device 260 may be integrated on the data processing device of fig. 1A or 1B, respectively, and may be disposed on the cloud or other servers on the network.

In the field of knowledge graph, the data collection device 260 may be a terminal device, or may be a server or a cloud input/output interface, for obtaining interaction layers (interfaces) of user behavior data and user portrait data.

Optionally, the entity recognition model/causal inference model 201 is implemented based on a deep neural network.

The operation of each layer in the deep neural network can be expressed mathematicallyTo describe: the work of each layer in a physical layer deep neural network can be understood as completing the transformation of input space into output space (i.e., row space to column space of the matrix) by five operations on the input space (set of input vectors), including: 1. dimension increasing/decreasing; 2. zoom in/out; 3. rotating; 4. translating; 5. "bending". Wherein the operations of 1, 2 and 3 are as followsThe operation of 4 is completed by +b, and the operation of 5 is implemented by α (). The term "space" is used herein to describe two words because the object being classified is not a single thing, but rather a class of things, space referring to the collection of all individuals of such things. Where W is a weight vector, each value in the vector representing a weight value of a neuron in the layer neural network. The vector W determines the spatial transformation of the input space into the output space described above, i.e. the weights W of each layer control how the space is transformed. The purpose of training the deep neural network is to finally obtain a weight matrix (a weight matrix formed by a plurality of layers of vectors W) of all layers of the trained neural network. Thus, the training process of the neural network is essentially a way to learn and control the spatial transformation, and more specifically to learn the weight matrix.

Because the output of the deep neural network is expected to be as close to the truly desired value as possible, the weight vector of each layer of the neural network can be updated by comparing the predicted value of the current network with the truly desired target value and then based on the difference between the two (of course, there is typically an initialization process prior to the first update, i.e. pre-configuring parameters for each layer in the deep neural network), for example, if the predicted value of the network is higher, the weight vector is adjusted to be predicted to be lower and adjusted continuously until the neural network can predict the truly desired target value. Thus, it is necessary to define in advance "how to compare the difference between the predicted value and the target value", which is a loss function (loss function) or an objective function (objective function), which are important equations for measuring the difference between the predicted value and the target value. Taking the loss function as an example, the higher the output value (loss) of the loss function is, the larger the difference is, and then the training of the deep neural network becomes a process of reducing the loss as much as possible.

The entity recognition model/causal inference model 201 derived by the training device 220 may be applied to different systems or devices. In fig. 2, the executing device 210 is configured with an I/O interface 212, and performs data interaction with an external device, and a "user" may input data to the I/O interface 212 through the user device 240.

The execution device 210 may invoke data, code, etc. in the data storage system 250, receive user portrayal data and behavioral data from the I/O interface 212, and invoke the entity recognition model/causal inference model 201 generated in the training device 220 to generate static entity triples and behavioral entity triples.

The association function 213 pre-processes the received user behavior data to facilitate subsequent further processing of the user data.

Finally, the I/O interface 212 returns the static entity triples and the behavioral entity triples to the user device 240 for provision to the user.

Further, the training device 220 may generate corresponding entity recognition models/causal inference models 201 based on different data for different targets to provide better results to the user.

In the case shown in FIG. 2, a user may manually specify data in the input execution device 210, e.g., to operate in an interface provided by the I/O interface 212. In another case, the user device 240 may automatically input user image and behavior data to the I/O interface 212 and obtain a static entity triplet and a behavior entity triplet that are subsequently used to generate the personal knowledge graph, and if the user device 240 automatically inputs data to obtain authorization of the user, the user may set corresponding rights in the user device 240. The user may view the results output by the execution device 210 at the user device 240, and the specific presentation may be in a specific manner such as display, sound, action, etc. The user device 240 may also be used as a data collection terminal to store the collected training data in the database 230.

It should be noted that fig. 2 is only a schematic diagram of a system architecture for building a personal knowledge graph according to an embodiment of the present application, and the positional relationship among devices, apparatuses, modules, etc. shown in the drawing is not limited in any way, for example, in fig. 2, the data storage system 250 is an external memory with respect to the execution device 210, and in other cases, the data storage system 250 may be disposed in the execution device 210.

Training with convolutional neural networks

The convolutional neural network (convolutional neuron network, CNN) is a deep neural network with a convolutional structure, which is a deep learning architecture that is an algorithm for machine learning to perform multiple levels of learning at different levels of abstraction. As a deep learning architecture, CNN is a feed-forward artificial neural network in which individual neurons respond to sentences input therein.

As shown in fig. 4, convolutional Neural Network (CNN) 100 may include an input layer 110, a convolutional layer/pooling layer 120, where the pooling layer is optional, and a neural network layer 130.

Convolution layer/pooling layer 120:

convolution layer:

The convolutional/pooling layer 120 as shown in fig. 4 may include layers as examples 121-126, in one implementation, 121 being a convolutional layer, 122 being a pooling layer, 123 being a convolutional layer, 124 being a pooling layer, 125 being a convolutional layer, 126 being a pooling layer; in another implementation, 121, 122 are convolutional layers, 123 are pooling layers, 124, 125 are convolutional layers, and 126 are pooling layers. I.e. the output of the convolution layer may be used as input to a subsequent pooling layer or as input to another convolution layer to continue the convolution operation.

Taking the example of the convolution layer 121, the convolution layer 121 may include a plurality of convolution operators, which are also called kernels, and function in the knowledge-graph construction process as a filter for extracting specific information from the input text semantic information, where the convolution operators may be a weight matrix in nature, and the weight matrix is usually predefined.

The weight values in the weight matrices are required to be obtained through a large amount of training in practical application, and each weight matrix formed by the weight values obtained through training can extract information from the input text, so that the convolutional neural network 100 is helped to perform correct prediction.

When convolutional neural network 100 has multiple convolutional layers, the initial convolutional layer (e.g., 121) tends to extract more general features, which may also be referred to as low-level features; as the depth of the convolutional neural network 100 increases, features extracted by the later convolutional layers (e.g., 126) become more complex, such as features of high level semantics, which are more suitable for the problem to be solved.

Pooling layer:

since it is often desirable to reduce the number of training parameters, the convolutional layers often require periodic introduction of pooling layers, i.e., layers 121-126 as illustrated at 120 in FIG. 4, which may be one convolutional layer followed by one pooling layer, or multiple convolutional layers followed by one or more pooling layers. In the knowledge graph construction process, the only purpose of the pooling layer is to reduce the space size of data.

Fully connected neural network layer 130:

after processing by the convolutional layer/pooling layer 120, the convolutional neural network 100 is not yet sufficient to output the required output information. Because, as previously described, the convolutional layer/pooling layer 120 will only extract features and reduce the parameters imposed by the input data. However, in order to generate the final output information (the required class information or other relevant information), convolutional neural network 100 needs to utilize fully-connected neural network layer 130 to generate the output of one or a set of the required number of classes. Thus, multiple hidden layers (131, 132 to 13n as shown in fig. 4) and an output layer 140 may be included in the fully connected neural network layer 130, where parameters included in the multiple hidden layers may be pre-trained according to relevant training data of a specific task type, which may include text semantic recognition, classification, or generation, for example.

After the multi-layer hidden layer in the fully connected neural network layer 130, that is, the final layer of the whole convolutional neural network 100 is the output layer 140, the output layer 140 has a class cross entropy-like loss function, specifically for calculating the prediction error, once the forward propagation of the whole convolutional neural network 100 (such as the propagation from 110 to 140 in fig. 4 is forward propagation), the backward propagation (such as the propagation from 140 to 110 in fig. 4 is backward propagation) starts to update the weight values and the bias of the aforementioned layers, so as to reduce the loss of the convolutional neural network 100 and the error between the result output by the convolutional neural network 100 through the output layer and the ideal result.

It should be noted that, the convolutional neural network 100 shown in fig. 4 is only an example of a convolutional neural network, and in a specific application, the convolutional neural network may also exist in the form of other network models, for example, a plurality of convolutional layers/pooling layers shown in fig. 5 are parallel, and the features extracted respectively are all input to the fully connected neural network layer 130 for processing.

After receiving the user behavior data, the execution device 210 determines a plurality of behavior entities in the user behavior data and entity types to which each behavior entity belongs based on the entity identification model, and can quickly and accurately identify related entities in the user behavior data by introducing the entity identification model, thereby improving efficiency of constructing personal knowledge graph. Then, the executing device 210 may determine the relationship between the entities according to the identified entity and the entity relationship set, generate a plurality of behavioral entity triples based on the relationship between each two behavioral entities and the two behavioral entities, and quickly and accurately determine the relationship between the entities by introducing the entity relationship extraction model to form a normalized entity triplet, so that the entity triples can be conveniently and quickly screened subsequently, and the screened entity triples are filled into the personal knowledge graph architecture, thereby improving the efficiency of building the personal knowledge graph. In a possible implementation manner, the calculation module 211 may further accept the static entity triples transmitted from the user device, mine deep target entity triples according to the above behavioral entity triples, static entity triples and causal reasoning model, mine deep target entity triples related to the user by introducing the causal reasoning model, and fill the target entity triples into the personal knowledge graph architecture in the subsequent process, so as to improve the correlation between the personal knowledge graph and the user.

Assume that the user behavior data is: at 19 months 5, 2020, 20:03:34 evening, user-click-red-Li Keqin-popular-super quality SQ-VIP. The entity recognition model recognizes five behavior entities and corresponding entity types thereof: red day (song name), li Keqin (singer), popularity (style), super quality SQ (sound quality), VIP (user attribute), the foregoing are entity types of each behavioral entity in parentheses, and in this piece of user behavior data, the behavior character of the user is click, and the operation object is red day. According to the five identified behavior entities and the entity relation set, five behavior entity triples are obtained: 1) User-click (2020/05/19-20:03:34) -red day; 2) Red day-singer-Li Keqin; 3) Red day-music style-fashion; 4) Red day-tone quality-super quality SQ; 5) User-account type-VIP. The five behavioural entity triples obtained above are stored in the data storage system 250.

In the scenario shown in fig. 1B, the processor of the intelligent terminal runs the convolutional neural network as shown in fig. 4 and 5, in the scenario shown in fig. 1A and 1C, the data processing device may run the convolutional neural network as shown in fig. 4 and 5, and in the system shown in fig. 2, the calculation module 211 in the training device 220 and the execution device 210 may run the convolutional neural network as shown in fig. 4 and 5.

It should be appreciated that instead of using convolutional neural networks, the present application may also employ other network models to construct entity recognition models and causal inference models, such as transformer-based bi-directional encoder representation (bidirectional encoder representations from transformers, BERT) techniques, as the application is not specifically limited in this regard.

Referring to fig. 3, fig. 3 is a system 300 for constructing a personal knowledge graph, in accordance with an embodiment of the application. As shown in fig. 3, the system 300 includes an intelligent terminal 310 and a server 320. The intelligent terminal 310 comprises a data source 301, wherein the data source 301 is used for recording data generated by corresponding APP operation on the intelligent terminal 310; the data acquisition module 302 is configured to acquire relevant user data from the data source 301 and store the user data; the calculation module 303 inputs the collected user data into an entity identification model/causal inference model to obtain an entity triplet; the knowledge graph generation module 304 generates a personal knowledge graph of the user according to the obtained entity triples and the personal knowledge graph architecture; the recommendation engine module 305 precisely matches the relevant content according to the user's display or implicit intention on the server 320 by analyzing the user's personal knowledge graph and pushes the relevant content to the corresponding application APP in the data source 301.

Referring to fig. 9, fig. 9 is a flow chart of a method for constructing a personal knowledge graph according to an embodiment of the present application. As shown in fig. 9, the method includes:

s910: acquiring initial user static attributes and initial user behavior attributes, and constructing a personal knowledge graph architecture according to the initial user static attributes and the initial user behavior attributes; the method comprises the steps that an initial user static attribute is used for representing personal information of a user, an initial user behavior attribute is used for representing the field of an entity type corresponding to the user behavior, the field is a set of entity types with the same characteristics, the entity types are a set of entities with the same characteristics or attributes, the entities are things which are associated with the user and are represented by nouns or words, a personal knowledge graph architecture represents the relation between the user and the personal information, and the field of the entity type corresponding to the user behavior is related to the user.

The initial user static attribute comprises a plurality of attribute relations of the user, and the attribute relations are in one-to-one correspondence with a plurality of personal information of the user. As shown in fig. 10, fig. 10 is a personal knowledge graph architecture of the user, and fig. 10 includes 7 attribute relationships (corresponding to 7 pieces of personal information): spouse, graduation university, nationality, sex, birthday, height and weight. In the personal knowledge graph framework, the 7 attribute relationships of the user do not have corresponding static entities, namely, the specific data of spouse, graduation institution, nationality, gender, birthday, height and weight of the user are not stored in the personal knowledge graph framework of the user.

The above-mentioned initial user behavior attribute is used to characterize the domain to which the entity type corresponding to the user behavior belongs, and specifically, the initial user behavior attribute may include the domain (e.g., music and game) associated with the user through an action and the entity type included in the associated domain, but does not include a specific behavior entity. As shown in fig. 10, the initial user behavior attribute contains four fields: music, games, movies, and news, which also contain multiple entity types, respectively. The user associates the four fields by listening, playing, watching and browsing, respectively. The specific process for acquiring the initial user behavior attribute is as follows: firstly, counting according to behavior data of a large number of users to obtain the first D fields with the highest user behavior frequency, wherein D is an integer which is larger than or equal to zero; then respectively acquiring corresponding entity types in the D fields; and associating different fields and entity types in the fields with the user through actions to obtain initial user behavior attributes shown in fig. 10. It should be understood that the four actions included in the initial user behavior attribute, the four fields corresponding to the four actions, and the entity types included in the four fields in fig. 10 are only one example provided by the present application, and the present application is not limited to the action types and numbers, the types and numbers of fields, and the entity types included in the fields.

After the initial user static attribute and the initial user behavior attribute are obtained, the initial user static attribute and the initial user behavior attribute are fused, and a personal knowledge graph framework (shown in fig. 10) with the user as a core is generated.

It can be seen that, in the embodiment of the present application, the personal knowledge graph architecture is an architecture for representing the relationship between the user and the static entity representing the personal information of the user, and the domain to which the user and the user behavior belong, and the architecture uses the user as a core, so that the personal knowledge graph of the user generated according to the architecture can better reflect the personal characteristics of the user.

S920: the method comprises the steps of obtaining initial user portrait data, initial user behavior data and an entity relation set, wherein the initial user portrait data corresponds to initial user static attributes, the initial user behavior data corresponds to initial user behavior attributes, and the entity relation set comprises a plurality of relations which are used for representing relations among different entities.

The initial user portrait data corresponds to the initial user static attribute, specifically, the initial user portrait data includes a plurality of static entities, and the number of the plurality of static entities is less than or equal to the number of attribute relationships included in the initial user static attribute, for example, if the initial user static attribute includes 7 attribute relationships as shown in fig. 10, the number of static entities included in the initial user portrait data is less than or equal to 7, and the included static entities are in one-to-one correspondence with the corresponding attribute relationships. The initial user profile data is structured data, i.e., each user profile data contains two static entities and an associated character between the two static entities.

The initial user behavior data corresponds to the initial user behavior attribute, and in one possible embodiment, the initial user behavior data may be semi-structured data, i.e., a dotting log (as shown in fig. 11), and the initial user behavior data may include a plurality of dotting logs, each of which corresponds to an operation behavior of one user. Each dotting log comprises behavior characters of a user, operation time corresponding to the behavior characters, operation objects corresponding to the behavior characters, fields to which the operation objects belong, text contents representing attributes of the operation objects, and text contents representing attributes of the user. The behavior characters are used for representing the behavior relation between the user and the operation object, and the behavior characters can comprise clicking, searching, commenting and copying; the operational object may be a specific entity or a piece of text (e.g., search text and comment text); the field of operation objects may include music, video, books, and the like; the attribute of the operation object may be a noun or text content for specifically describing the operation object; the user attribute may be a noun or text content describing the user in a corresponding domain, e.g. in the music domain the user is a VIP user on the music application APP. It should be understood that the above listed behavior characters and objects of operation are only a few examples of the present application, and the present application is not limited thereto in particular.

The acquisition time period of the initial user portrait data is the same as the acquisition time period of the initial user behavior data. The set of entity relationships may comprise a plurality of relationships for characterizing relationships between different entities, and a plurality of entity types for characterizing relationships between different entity types. For example, the entity relationship set may include three relationships of clicking, singing and music types, and the three relationships may be used to characterize the relationship between the user and the operation object, the singer and the song name, and the music type to which the song name belongs, respectively.

Referring to fig. 11, as shown in example one of the user behavior data, the piece of user behavior data specifically includes: user-click-2020/05/19-20:03:34-red day-music-Li Keqin-popular-super quality SQ-VIP; the action character of the user is clicking, the operation object is music of red day, the field of the operation object is music, and the operation time is 2020/05/19-20:03:34. The red date of the operation object is a behavior entity, and the corresponding entity type is song name; text content characterizing the properties of the operation object includes Li Keqin, popular and super quality SQ; the text content characterizing the user attribute is VIP.

As shown in the second example of the user behavior data, the piece of user behavior data specifically includes: user-comment-2020/05/18-19:03:34- "Multi-position stars are antipesthetic singing" -news headline-antipesthetic thematic "-thank you for all-position stars"; the behavior characters of the user are comments, the operation object is a piece of text content, namely the belonging field of the operation object of 'multiple stars are antipesthetic singing' is news, and the text content representing the attribute of the operation object is comment text of the user, namely 'thanking for all stars'.

It should be understood that the description and the two examples of the user behavior data in fig. 11 are only illustrative examples of the present application, and the embodiments of the present application are not limited thereto in detail.

S930: obtaining M initial static entity triples according to the initial user portrait data and the entity relation set; e initial behavior entity triples are obtained according to the initial user behavior data and the entity relation set; the static entity triplets are used for describing attribute relations between users and static entities or between static entities, the behavior entity triplets are used for describing behavior relations between users and behavior entities or between behavior entities, and M and E are integers which are larger than or equal to zero.

In a possible implementation manner, the obtaining M initial static entity triples according to the initial user portrait data and the entity relation set includes: identifying A static entities and entity types to which the A static entities belong from initial user portrait data by using an entity identification model, wherein all attribute relationships or partial attribute relationships in the A static entities and the initial user static attributes are in one-to-one correspondence, namely the A is smaller than or equal to the number of attribute relationships in the initial user static attributes, and the A is an integer larger than or equal to zero; m initial static entity triples (the representation is shown in FIG. 6) are generated from the A static entities and the entity relationship sets. The entity relation set includes a plurality of relations, which refers to the relations between different entities and also refers to the relations between the entities included in one entity type and the entities included in another entity type. Further, the specific process of generating M initial static entity triples according to the a static entities and the entity relationship set includes: matching the entity type of the A static entities with the entity type in the entity relation set, and determining attribute relations (also called relations) between the A static entities or between the A static entities and the user; and obtaining M initial static entity triples according to the A static entities and the determined attribute relationship.

For example, the initial user portrait data for user 1 may include: (1) user 1-sex-male; (2) user 1-spouse-user 2; (3) user 2-graduation at university 1; (4) user 1-graduation at-university 1.

The generation of the initial static entity triples is described below with the four initial user portrait data enumerated above: the entity recognition model recognizes static entities from the initial user portrait data of the user 1, wherein the static entities comprise a male (gender), a user 2 (name of a person), a university 1 (name of a school), and the entity types of the static entities are arranged in brackets; and identifying the associated character "sex" between user 1 and "man", the associated character "spouse" between user 1 and user 2, the associated character "graduation" between user 1 and university 1, the associated character "graduation" between user 2 and university 1; then, matching the entity types corresponding to the static entities and the identified association characters between the static entities with the entity types in the entity relationship set and the relationships between the entity types in the entity relationship set respectively, so as to determine attribute relationships between the static associations or between the user 1 and the static entities; based on the determined attribute relationship and each static entity, four initial static entity triples are generated: user 1 (name) -gender (attribute relationship) -male (gender), user 1 (name) -spouse (attribute) -user 2 (name), user 2 (name) -graduation in (attribute) -university 1 (school name), user 1 (name) -graduation in (attribute) -university 1 (school name).

It should be understood that in the present application, an entity is a noun or data that characterizes a variety of specific things, and that the entity in the present application includes static entities, behavioral entities, and users. The relationship between entities is divided into attribute relationship and behavior relationship, the attribute relationship and static entity are adopted for description, so that user portrait data for representing personal information of a user and user behavior data for representing behavior of the user are distinguished, and the description mode is not particularly limited.

Referring to fig. 6, fig. 6 is a schematic diagram of a representation of an entity triplet according to an embodiment of the present application, where the initial static entity triplet may be represented by the form shown in fig. 6. As shown in fig. 6, the entity triples are used to characterize the relationship between two entities, and each entity triplet includes two entities, a relationship between the two entities, and an entity type tag to which the two entities belong. The two entities in fig. 6 are Li Keqin and hong Kong; li Keqin and hong Kong are in relation to each other; the entity type to which Li Keqin belongs is a person name, and the entity type to which hong Kong belongs is a place name.

The process of obtaining E initial behavior entity triples from the initial user behavior data and the entity relationship set, which includes two steps of S1210 and S1220, will be described in detail with reference to fig. 12 as follows:

S1210: and extracting personal knowledge, namely acquiring B behavior entities and C behavior characters from initial user behavior data, wherein the C behavior characters correspond to O behavior entities in the B behavior entities, the behavior characters are used for representing the operation of a user on the O behavior entities, the initial user behavior data are a plurality of dotting logs related to the user behaviors, B, C and H are integers which are greater than or equal to zero, and C, H is less than or equal to B.

Specifically, the initial user behavior data is input into an entity identification model, and the entity identification model can identify B behavior entities and entity types corresponding to the B behavior entities from a plurality of dotting logs contained in the initial user behavior data. Meanwhile, the entity recognition model can also recognize C behavior characters of the user and O behavior entities corresponding to the C behavior characters from the initial user behavior data, wherein the O behavior entities are contained in the B behavior entities.

For example, the process of entity identification is described using example one of the user behavior data in FIG. 11: the behavior entities that the entity identification model can identify from the dotting log include: red day, music, popularity, super quality SQ and VIP; the field to which the dotting log belongs is the music field; meanwhile, the entity identification model can also classify the identified behavior entities and determine the entity types to which the identified behavior entities belong: the entity type of red day is song name, popular entity type is music type, SQ entity type is tone quality, VIP entity type is user account attribute; further, the entity recognition model can also recognize behavior characters- "click" and entities- "red days" corresponding to the "click" from the piece of user behavior data.

In one possible embodiment, the initial user behavior data may be determined from the initial user behavior data and the personal knowledge graph architecture prior to identifying the behavior entity from the initial user behavior data using the entity identification model. The original user behavior data is also semi-structured data, i.e. a plurality of dotting logs (as shown in fig. 11). According to the domain of the personal knowledge graph architecture, the initial user behavior data is screened out from the original user behavior data, specifically, as the domains to which the multiple dotting logs contained in the original user behavior data respectively belong may be different from the domains defined in the initial user behavior attribute, according to the domains defined in the initial user behavior attribute, multiple dotting logs belonging to the domains defined in the initial user behavior attribute are selected out from the original user behavior data, and then the selected multiple dotting logs are used as the initial user behavior data of the input entity identification model.

S1220: the personal knowledge shows that G initial behavior entity triples are obtained according to the B behavior entities and the entity relation set; obtaining H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities and the entity relation set, wherein E is equal to the sum of G and H; the G initial behavior entity triples are used for describing the relationship between the B behavior entities, and the H initial behavior entity triples are used for describing the relationship between the user and the O behavior entities.

Specifically, based on the B behavioral entities identified in step S1210, two kinds of personal knowledge can be obtained: static knowledge and behavioral knowledge. The static knowledge is represented by G initial behavioral entity triples that describe the relationship between B behavioral entities, or the relationship between the user and B behavioral entities (non-operational relationship); the behavioral knowledge is represented by H initial behavioral entity triples that describe the relationship between the user and O behavioral entities (the operational relationship of the user to O behavioral entities). The initial behavioral entity triples described above may be represented in the form shown in fig. 6.

The process of generating an initial behavioral entity triplet from the B behavioral entities and the set of entity relationships will be described in detail below: (1) For the G initial behavior entity triples, firstly, entity types of the B behavior entities and entity types in an entity relationship set can be matched, so that the relationship between every two behavior entities in the B behavior entities is determined according to the relationship between the entity types in the entity relationship set, and then G initial behavior entity triples are generated based on every two behavior entities, the entity types of the two behavior entities and the relationship between the two behavior entities; (2) For the above-mentioned H initial behavior entity triples, firstly, the entity types to which the O behavior entities belong and the behavior characters between the user and the O behavior entities may be respectively matched with the entity types in the entity relationship set and the relationships between the entity types, so as to determine the operation relationships between the user and the O behavior entities; and then generating H initial behavior entity triples according to the determined operation relation, the user and the O behavior entities.

For example, the generation of the initial behavioural entity triplet is described using example one of the user behavioural data in fig. 11: (1) The entity types of the behavior entities identified in the example one are matched with the entity types in the entity relationship set, and the relationship among the behavior entities contained in different entity types in the example one can be determined at the moment because the relationship among the entity types is already defined in the entity relationship set: the relationship between red date (song name) and popularity (music type) is music type, the relationship between red date (song name) and SQ (sound quality) is sound quality, the relationship between user (name) and VIP (account attribute) is account attribute, the entity types of the entities in the brackets are the entity types, and then the relationship between every two behavior entities can be represented by adopting the representation form of the entity triples in FIG. 6, at this time, three initial behavior entity triples can be obtained, and the three initial behavior entity triples form part of static knowledge of the user; (2) The entity types corresponding to the user and the red date and the user behavior character 'click' acquired from the first example are respectively matched with the entity types and the relationships among the entity types in the entity relationship set, the operation relationship of the user on the red date is determined to be 'click', and an initial behavior entity triplet is generated based on the user (name), the red date (song name) and the click (operation relationship), wherein the initial behavior entity triplet forms a part of the user behavior knowledge.

In one possible implementation, the entity recognition model may also obtain the H operation times from the initial user behavior data. The obtaining H initial behavior entity triples according to the user, the C behavior characters and the O behavior entities includes: and generating H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities, the H operation times and the entity relation set.

Specifically, according to the user, the C behavior characters, the O behavior entities and the H operation times, H behavior entity quadruples are generated, where a representation manner of the behavior entity quadruples is shown in fig. 7, and each entity quadruple includes two entities, a relationship term representing a relationship between the two entities, a time term representing an occurrence time of the relationship, and entity type labels corresponding to the two entities. The representation mode of the entity tetrad is different from the representation mode of the entity triplet, and the time item for representing the behavior or action of the user is added in the tetrad, so that the behavior characteristic of the user can be reflected more accurately.

After the above-mentioned H behavioral entity triples are generated, the time items and the relationship items in the H behavioral entity triples may be spliced into one item, so as to obtain H behavioral entity triples with time items, and the specific process is shown in fig. 13, and then the formats of the entity triples for representing behavioral knowledge and static knowledge may be unified, where the specific representation manner of the H behavioral entity triples with time items is shown in fig. 8.

In the embodiment of the application, the expression form shown in fig. 8 is adopted to describe the above-mentioned H initial behavior entity triples, so that the formats of the H initial behavior entity triples and the G initial behavior entity triples can be unified, and the access of data is convenient, thereby improving the construction efficiency of the personal knowledge graph; in addition, as the time items are added in the H initial behavior entity triples, the behavior characteristics of the user in different time periods can be better reflected, and further, the user personal knowledge graph constructed according to the initial behavior entity triples is guaranteed to have better timeliness.

S940: and generating a personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples and the personal knowledge graph framework.

Specifically, in one possible implementation manner, the M initial static entity triples and the E initial behavioral entity triples are filled into a personal knowledge graph architecture, so as to generate a personal knowledge graph capable of reflecting the characteristics of the user.

In a possible implementation manner, the generating the personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples, and the personal knowledge graph architecture may include: and generating a personal knowledge graph reflecting the characteristics of the user according to the M initial static entity triples, F initial behavior entity triples in the E initial behavior entity triples and the personal knowledge graph architecture. Specifically, in one possible implementation, M initial static entity triples and F initial behavioral entity triples are filled into a personal knowledge graph architecture to obtain a personal knowledge graph of the user.

The E initial behavior entity triples are generated based on initial user behavior data, and the acquisition time of the initial user behavior data is any time between the time T1 and the current system time; the F initial behavior entity triples are generated based on first user behavior data, the acquisition time of the first user behavior data is any time between a time T2 and the current system time, the time T2 is later than or equal to a time T1, and the time T2 and the time T1 are both earlier than the current system time, so that the E initial behavior entity triples comprise F initial behavior entity triples. It should be understood that the specific time represented by T1 and T2 may be adjusted according to the application scenario, which is not particularly limited by the present application.

In a possible implementation manner, the specific process of generating the personal knowledge graph of the user according to the M initial static entity triples, the F initial behavior entity triples and the personal knowledge graph architecture may include: and selecting a preset number Q of initial behavior entity triples from the F initial behavior entity triples, filling the Q initial behavior entity triples and the M initial static entity triples into a personal knowledge graph framework together, generating a personal knowledge graph of a user, wherein Q is an integer greater than or equal to zero.

The process of selecting Q initial behavior entity triples from the F initial behavior entity triples will be described in detail below: firstly, de-duplicating the F initial behavior entity triples to obtain I initial behavior entity triples; the de-duplication refers to removing the same initial behavior entity triples in the F initial behavior entity triples, and the I initial behavior entity triples obtained after de-duplication are all different. The same judging conditions of the two initial behavior entity triples are as follows: when two behavior entities in one initial behavior entity triplet are respectively the same as two behavior entities in the other initial behavior entity triplet, and the relationship between the two behavior entities in the one initial behavior entity triplet is the same as the relationship between the two behavior entities in the other initial behavior entity triplet; it should be noted that the process of determining whether two initial behavior entity triples are identical does not involve a determination of the operation time. After obtaining the I initial behavior entity triples, Q initial behavior entity triples are selected from the I initial behavior entity triples, wherein I is an integer greater than or equal to zero.

The above process of selecting Q initial behavior entity triples from the I initial behavior entity triples is divided into two cases:

(1) When the I is smaller than or equal to the preset quantity Q, the I initial behavior entity triples are Q initial behavior entity triples needing to be selected.

(2) When the I is larger than the preset quantity Q, grouping F initial behavior entity triples according to the I initial behavior entity triples to obtain O behavior entity triplet sets, wherein the O behavior entity triplet sets are in one-to-one correspondence with the I initial behavior entity triples, and the initial behavior entity triples contained in each behavior entity triplet set are identical to the initial behavior entity triples corresponding to the behavior entity triplet sets; then, according to the number of elements in each behavior entity set, sorting the I initial behavior entity triples according to a rule that the number is from more to less to obtain a sorting result, wherein when the number of elements in the behavior entity triples is more, the initial behavior entity triples corresponding to the behavior entity triples are more front in the sorting result; and finally, selecting the first Q initial behavior entity triples from the sorting result.

In a possible implementation manner, the generating the personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples, and the personal knowledge graph architecture may include: generating X initial target entity triples according to M initial static entity triples and/or E initial behavior entity triples, wherein X is an integer greater than or equal to zero; each of the target entity triples contains at least one target entity, the target entity triples describing a relationship between the target entities, or between a user and the target entity. Generating a personal knowledge graph according to the M initial static entity triplets, the E initial behavior entity triplets, the X initial target entity triplets and the personal knowledge graph framework; specifically, in one possible implementation, M initial static entity triples, E initial behavior entity triples, and X initial target entity triples are filled into a personal knowledge graph architecture to generate a personal knowledge graph.

The process of generating X initial target entity triples according to the M initial static entity triples and/or the E initial behavior entity triples includes three implementation manners, where the target entity triples generated by the three implementation manners form X initial target entity triples, and the three implementation manners are described below by way of example:

(1) Generating an initial target entity triplet according to the initial static entity triplet: referring to fig. 14, fig. 14 is a schematic diagram illustrating a process for constructing a target entity triplet according to an embodiment of the present application. As shown in fig. 14, the entity triplet one, entity triplet two and entity triplet three in fig. 14 are input into a causal reasoning model, which can identify the common feature among the three entity triples, namely that both user 1 and user 2 are graduated from university 1, and then conclude that: the relationship between user 1 and user 2 is alumni and based on this conclusion, a target entity triplet one is generated as shown in fig. 14.

(2) Generating an initial target entity triplet according to the initial static entity triplet and the initial behavior entity triplet: referring to fig. 15, fig. 15 is a schematic diagram illustrating a construction process of another target entity triplet according to an embodiment of the present application. As shown in fig. 15, the entity triplet four, the entity triplet five and the entity triplet six in fig. 15 are input into a causal reasoning model, and the model can identify that the news category browsed by the user belongs to the current politics from the entity triplet four and the entity triplet five; the university specialty of the user is also identified from entity triplet six as being related to the current politics, then information is drawn that the user is interested in the current politics type, and a target entity triplet two is generated based on the information.

(3) Generating an initial target entity triplet according to the initial behavior entity triplet: referring to fig. 16, fig. 16 is a schematic diagram illustrating a construction process of another target entity triplet according to an embodiment of the present application. As shown in fig. 16, the entity triplet seven, entity triplet eight and entity triplet nine in fig. 16 are input into a causal reasoning model, which can identify information related to red date from the entity triplet one, such as Li Keqin for singer on red date, popular music of type, SQ for tone quality; identifying the text content of the news, i.e., "antipesthetic donation" from entity triplet two, e.g., donation stars Li Keqin, zhang Xueyou, and Zhou Jielun; identifying information related to the movie from entity triplet three, e.g., actors Zhang Zhilin and Li Keqin of the movie; the model then infers that the three actions of the user are relevant to Li Keqin by analyzing the various information identified above, thereby concluding that the user likes Li Keqin, and generating a target entity triplet three based on the conclusion.

In a possible implementation manner, the specific process of generating the personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples, the X initial target entity triples, and the personal knowledge graph architecture may include: and generating a personal knowledge graph according to the M initial static entity triples, F initial behavior entity triples in the E initial behavior entity triples, Y initial target entity triples in the X initial target entity triples and the personal knowledge graph architecture. Specifically, in one possible implementation, M initial static entity triples, F initial behavioral entity triples, and Y initial target entity triples are filled into a personal knowledge graph architecture to obtain a personal knowledge graph.

The E initial behavior entity triples are generated based on initial user behavior data, and the acquisition time of the initial user behavior data is any time between the time T1 and the current system time; the F initial behavior entity triples are generated based on first user behavior data, the acquisition time of the first user behavior data is any time between a time T2 and the current system time, the time T2 is later than or equal to a time T1, and the time T2 and the time T1 are both earlier than the current system time, so that the E initial behavior entity triples comprise F initial behavior entity triples. The Y initial target entity triples are generated according to M initial static entity triples and/or F initial behavior entity triples, so that the X initial target entity triples include the Y initial target entity triples.

In a possible implementation manner, the process of generating the personal knowledge graph according to the M initial static entity triples, the F initial behavior entity triples, the Y initial target entity triples, and the personal knowledge graph architecture may further include: q entity triples are selected from the F initial behavior entity triples and the Y initial target entity triples, and then the Q entity triples and the M initial static entity triples are filled into a personal knowledge graph architecture to obtain a personal knowledge graph.

The process of selecting Q entity triples from the F initial behavior entity triples and the Y initial target entity triples will be described in detail below: the method comprises the steps of firstly carrying out de-duplication on Y initial target entity triples to obtain J initial target entity triples, wherein J is an integer greater than or equal to zero, J is less than or equal to Y, and the de-duplication process of the Y initial target entity triples is the same as the de-duplication process of F initial behavior entity triples, and is not repeated here.

After obtaining J initial target entity triples, Q entity triples are selected from F initial behavior entity triples and J initial target entity triples, and the specific process is as follows:

(1) And when J is equal to the preset quantity Q, filling M initial static entity triples and J initial target entity triples into a personal knowledge graph framework to generate a personal knowledge graph.

(2) When J is greater than the preset quantity Q, grouping Y initial target entity triples according to J initial target entity triples to obtain J target entity triples; the process of obtaining the J target entity triplet sets and ordering the J initial target entity triples is the same as the process of obtaining the O behavior entity triplet sets and ordering the I initial behavior entity triples, and therefore will not be described in detail. After the sequencing result of the J initial target entity triples is obtained, filling the M initial static entity triples and the first Q initial target entity triples in the J initial target entity triples into a personal knowledge graph framework to generate a personal knowledge graph.

(3) And when J is smaller than the preset quantity Q and the sum of J and I is larger than Q, filling the first Q-J initial behavior entity triples in the M initial static entity triples, the J initial target entity triples and the ordered I initial behavior entity triples into a personal knowledge graph architecture to generate a personal knowledge graph.

(4) And when J is smaller than the preset quantity Q and the sum of J and I is smaller than or equal to Q, filling M initial static entity triples, J initial target entity triples and I initial behavior entity triples into a personal knowledge graph framework to generate a personal knowledge graph.

In a possible implementation manner, please refer to fig. 17, when the entity triples to be expanded are filled into the personal knowledge graph architecture, searching is performed in the universal knowledge graph and/or the domain-type knowledge graph in the server side or the cloud side, and whether any entity in the entity triples to be expanded exists in the universal knowledge graph or the domain-type knowledge graph is confirmed, if so, an identifier of any entity is stored in the intelligent terminal side, and the identifier can be used for quickly positioning any entity in the universal knowledge graph or the domain-type knowledge graph, so that information related to any entity can be quickly obtained; the entity triples to be expanded refer to the entity triples filled in the personal knowledge graph framework in each embodiment.

It can be seen that, in the embodiment of the present application, the personal knowledge graph architecture is constructed by adopting the static attribute representing the personal user information and the behavior data representing the domain to which the user behavior belongs, so that the architecture is an architecture that uses the user as a core and can reflect the personal information and behavior characteristics of the user; generating a behavior entity triplet representing the user personal information and static entity triplet and a behavior entity triplet representing the user behavior characteristic by using the user portrait data and the user behavior data respectively, and filling the static entity triplet and the behavior entity triplet into a personal knowledge graph architecture to obtain a personal knowledge graph reflecting the user personal characteristic; therefore, the related content can be accurately recommended to the user based on the generated personal knowledge graph. Meanwhile, by introducing time items into the behavior entity triples, the specific moment when the user behavior occurs can be reflected more accurately, so that the constructed personal knowledge graph is more accurate; in addition, the target entity triples reflecting the deep personal characteristics of the user are mined based on the static entity triples and the behavior entity triples, and the target entity triples are filled into the personal knowledge graph framework, so that the obtained personal knowledge graph can reflect the characteristics of the user more accurately and has stronger correlation with the user.

Referring to fig. 18, fig. 18 is a flow chart of a personal knowledge graph updating method provided by the application. As shown in fig. 18, the updating method includes four steps:

s1810: and acquiring the static attribute and the behavior attribute of the target user, and updating the personal knowledge graph by utilizing the static attribute and the behavior attribute of the target user to obtain the target personal knowledge graph.

Specifically, firstly, comparing attribute relationships contained in the static attributes of the target user with attribute relationships contained in the static attributes of the initial user, selecting a part which is not contained in the static attributes of the initial user from the attribute relationships contained in the static attributes of the target user to obtain an attribute relationship screening result, and expanding the attribute relationships in the screening result into a personal knowledge graph; similarly, comparing whether the operation behaviors, the fields and the entity types contained in the fields contained in the target user behavior attribute and the initial user behavior attribute are the same or not, and screening the operation behaviors, the fields and the entity types which are not contained in the initial user behavior attribute from the target user behavior attribute to obtain a screening result; and expanding the attribute relationship, the operation behavior, the field and the entity type in the two screening results into the personal knowledge graph to obtain the target personal knowledge graph.

The acquisition time periods of the target user static attribute, the target user behavior attribute, the second user portrait data and the second user behavior data are the same and are later than the acquisition time periods of the initial user static attribute, the initial user behavior attribute, the initial user portrait data and the initial user behavior data.

S1820: and acquiring second user portrait data and second user behavior data.

Specifically, the specific representation forms of the second user portrait data and the second behavior data are the same as the initial user portrait data and the initial user behavior data, respectively, which are described in step S920 and are not described herein. The second user portrayal data corresponds to the user static attribute in the target personal knowledge graph, the second user behavior data corresponds to the user behavior attribute in the target personal knowledge graph, and the acquisition time period of the second user portrayal data and the second user behavior data is later than the acquisition time period of the initial user portrayal data and the initial user behavior data.

S1830: obtaining K first static entity triples according to the second user portrait data and the entity relation set; and obtaining L first behavior entity triples according to the second user behavior data and the entity relation set, wherein L and K are integers greater than or equal to zero.

Specifically, the specific process of this step is the same as that of step S930, and will not be described again.

S1840: and updating the target personal knowledge graph according to the K first static entity triples and the L first behavioral entity triples.

Specifically, in this step, the process of generating the entity triples to be extended according to the K first static entity triples and the L first behavioral entity triples is the same as the corresponding process in step S940, and will not be described again; and after the entity triplet to be expanded is obtained, updating the target personal knowledge graph by utilizing the entity triplet to be expanded.

Referring to fig. 19, fig. 19 is a flow chart of another personal knowledge graph updating method provided by the present application. As shown in fig. 19, the updating method includes three steps:

s1910: and acquiring third user portrait data and third user behavior data.

S1920: obtaining R second static entity triples according to the third user portrait data and the entity relation set; and obtaining S second behavior entity triples according to the second user behavior data and the entity relation set, wherein R and S are integers greater than or equal to zero.

Specifically, steps S1910 and S1920 correspond to the specific processes in steps S920 and S930, and are not described again.

S1930: and updating the personal knowledge graph according to the R second static entity triples and the S second behavior entity triples.

Specifically, in this step, the process of generating the entity triples to be extended according to the K first static entity triples and the L first behavioral entity triples is the same as the corresponding process in step S940, and will not be described again; and updating the personal knowledge graph obtained in the step S940 by using the entity triplet to be expanded.

Referring to fig. 20, fig. 20 is a schematic block diagram of a personal knowledge graph construction apparatus 2000 according to an embodiment of the present application. As shown in fig. 20, the apparatus 2000 includes an acquisition unit 2010, a processing unit 2020, and a generation unit 2030.

An acquiring unit 2010, configured to acquire an initial user static attribute and an initial user behavior attribute; the initial user static attribute is used for representing personal information of a user, the initial user behavior attribute is used for representing the field of an entity type corresponding to the user behavior, the field is a set of entity types with the same characteristics, the entity types are a set of entities with the same characteristics or attributes, the entities are things associated with the user and represented by nouns or words; the method comprises the steps of obtaining initial user portrait data, initial user behavior data and an entity relation set, wherein the initial user portrait data corresponds to initial user static attributes, the initial user behavior data corresponds to initial user behavior attributes, and the entity relation set comprises a plurality of relations which are used for representing relations among entities.

A processing unit 2020, configured to construct a personal knowledge graph architecture according to the initial user static attribute and the initial user behavior attribute, where the personal knowledge graph architecture characterizes a relationship between a user and personal information, and a relationship between domains to which entity types corresponding to the user and the user behavior belong; the method comprises the steps of obtaining M initial static entity triples according to initial user portrait data and entity relation sets; e initial behavior entity triples are obtained according to the initial user behavior data and the entity relation set; the initial static entity triplets are used for describing attribute relations between users and static entities or between static entities, the initial behavior entity triplets are used for describing behavior relations between users and behavior entities or between behavior entities, and M and E are integers which are larger than or equal to zero.

The generating unit 2030 is configured to generate a personal knowledge graph according to the M initial static entity triples, the E initial behavioral entity triples, and the personal knowledge graph architecture.

It should be understood that the acquiring unit 2010, the processing unit 2020 and the generating unit 2030 in the apparatus 2000 may also be specifically configured to perform the corresponding methods in the embodiments as described in fig. 9, 18 and 19, and are not described here again.

Referring to fig. 21, fig. 21 is a schematic hardware structure diagram of a personal knowledge graph construction device according to an embodiment of the present application. The image processing apparatus 2100 shown in fig. 21 (the apparatus 2100 may be a computer device in particular) includes a memory 2101, a processor 2102, a communication interface 2103, and a bus 2104. The memory 2101, the processor 2102, and the communication interface 2103 are connected to each other by a bus 2104.

The memory 2101 may be a Read Only Memory (ROM), a static storage device, a dynamic storage device, or a random access memory (random access memory, RAM). The memory 2101 may store a program, and when the program stored in the memory 2101 is executed by the processor 2102, the processor 2102 and the communication interface 2103 are used to perform the respective steps of the image processing method of the embodiment of the present application.

The processor 2102 may employ a general-purpose central processing unit (central processing unit, CPU), microprocessor, application specific integrated circuit (application specific integrated circuit, ASIC), graphics processor (graphics processing unit, GPU) or one or more integrated circuits for executing associated programs to perform functions required by the elements of the image processing apparatus of an embodiment of the present application or to perform the image processing method of an embodiment of the present application.

The processor 2102 may also be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the image processing method of the present application may be performed by integrated logic circuits of hardware in the processor 2102 or by instructions in the form of software. The processor 2102 may also be a general purpose processor, a digital signal processor (digital signal processing, DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf programmable gate array (field programmable gate array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components. The disclosed methods, steps, and logic blocks in the embodiments of the present application may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present application may be embodied directly in the execution of a hardware decoding processor, or in the execution of a combination of hardware and software modules in a decoding processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in the memory 2101, and the processor 2102 reads information in the memory 2101, and in combination with hardware thereof, performs functions necessary for execution of units included in the image processing apparatus of the embodiment of the present application, or executes the image processing method of the embodiment of the method of the present application.

The communication interface 2103 enables communication between the apparatus 2100 and other devices or communication networks using a transceiver device such as, but not limited to, a transceiver. For example, training data (e.g., images to be processed in the embodiment of the application depicted in FIG. 10) may be acquired via the communication interface 2103.

The bus 2104 may include a path to transfer information between elements of the device 2100 (e.g., the memory 2101, the processor 2102, the communication interface 2103).

It should be noted that although the apparatus 2100 shown in fig. 21 shows only a memory, a processor, and a communication interface, those skilled in the art will appreciate that in a particular implementation, the apparatus 2100 also includes other devices necessary to achieve normal operation. Also, as will be appreciated by those of skill in the art, the apparatus 2100 may also include hardware devices that implement other additional functions, as desired. Furthermore, it will be appreciated by those skilled in the art that the apparatus 2100 may also include only the necessary components to implement an embodiment of the present application, and not necessarily all of the components shown in FIG. 21.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.

In the several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.

The above functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a random access memory (random access memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (24)

1. The personal knowledge graph construction method is characterized by comprising the following steps of:

   acquiring initial user static attributes and initial user behavior attributes, and constructing a personal knowledge graph architecture according to the initial user static attributes and the initial user behavior attributes; the initial user static attribute is used for representing personal information of a user, the initial user behavior attribute is used for representing a field to which an entity type corresponding to user behavior belongs, the field is a set of entity types with the same characteristics, the entity types are a set of entities with the same characteristics or attributes, the entities are things which are associated with the user and are represented by nouns or words, the personal knowledge graph architecture represents a relationship between the user and the personal information, and the user and the field to which the entity type corresponding to the user behavior belongs;

   acquiring initial user portrait data, initial user behavior data and an entity relation set, wherein the initial user portrait data corresponds to the initial user static attribute, the initial user behavior data corresponds to the initial user behavior attribute, and the entity relation set comprises a plurality of relations which are used for representing the relation between the entities;

   Obtaining M initial static entity triples according to the initial user portrait data and the entity relation set; e initial behavior entity triples are obtained according to the initial user behavior data and the entity relation set; the initial static entity triples are used for describing attribute relations between the users and static entities or between the static entities, the initial behavior entity triples are used for describing behavior relations between the users and behavior entities or between the behavior entities, and M and E are integers which are larger than or equal to zero;

   and generating the personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples and the personal knowledge graph framework.
2. The method of claim 1, wherein the obtaining M initial static entity triples from the initial user representation data and the set of entity relationships comprises:

   obtaining A static entities from the initial user portrait data, wherein A is an integer greater than or equal to zero;

   and obtaining the M initial static entity triples according to the A static entities and the entity relation set.
3. A method according to claim 1 or 2, wherein said deriving E initial behavioural entity triples from said initial user behavioural data and said set of entity relations comprises:

   b behavior entities and C behavior characters are obtained from the initial user behavior data, the C behavior characters correspond to O behavior entities in the B behavior entities, the behavior characters are used for representing the operation of the user on the O behavior entities, and the B, the C and the O are integers which are larger than or equal to zero;

   obtaining G initial behavior entity triples according to the B behavior entities and the entity relation set;

   obtaining H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities and the entity relation set, wherein G and H are integers greater than or equal to zero;

   the G initial behavior entity triples are used for describing the relation among the B behavior entities; the H initial behavior entity triplets are used for describing the relation between the user and the O behavior entities, and the E initial behavior entity triplets comprise the G initial behavior entity triplets and the H initial behavior entity triplets.
4. A method according to claim 3, wherein the method further comprises:

   acquiring H operation times from the initial user behavior data;

   the obtaining H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities and the entity relationship set includes:

   and generating the H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities, the H operation times and the entity relation set.
5. The method of any of claims 1-4, wherein the generating the personal knowledge-graph from the M initial static entity triples, the E initial behavioral entity triples, and the personal knowledge-graph architecture comprises:

   generating X initial target entity triples according to the M initial static entity triples and/or the E initial behavior entity triples, wherein X is an integer greater than or equal to zero; each of the target entity triples comprises at least one target entity, the target entity triples being used to describe a relationship between target entities or between the user and the target entities;

   And generating the personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples, the X initial target entity triples and the personal knowledge graph framework.
6. The method of any of claims 1-4, wherein the generating the personal knowledge-graph from the M initial static entity triples, the E initial behavioral entity triples, and the personal knowledge-graph architecture comprises:

   generating the personal knowledge graph according to the M initial static entity triples, F initial behavior entity triples in the E initial behavior entity triples and the personal knowledge graph framework;

   the F initial behavior entity triples are obtained according to first user behavior data and the entity relation set, the first user behavior data are behavior data, the difference value between the acquisition time and the current system time in the initial user behavior data is smaller than a preset threshold, and F is an integer larger than or equal to zero.
7. The method of claim 6, wherein generating the personal knowledge-graph from the M initial static entity triples, F initial behavior entity triples of the E initial behavior entity triples, and the personal knowledge-graph architecture comprises:

   Performing de-duplication on the F initial behavior entity triples to obtain I initial behavior entity triples, wherein I is an integer which is less than or equal to F and is greater than or equal to zero;

   when the I is smaller than or equal to a preset number Q, generating the personal knowledge graph according to the M initial static entity triples, the I initial behavior entity triples and the personal knowledge graph framework, wherein Q is an integer larger than or equal to zero;

   when the I is larger than the preset quantity Q, grouping the F initial behavior entity triples according to the I initial behavior entity triples to obtain O behavior entity triplet sets, wherein the O behavior entity triplet sets are in one-to-one correspondence with the I initial behavior entity triples, and the initial behavior entity triples contained in each behavior entity triplet set are identical to the initial behavior entity triples corresponding to the behavior entity triplet sets; sorting the I initial behavior entity triples according to the number of elements in each behavior entity triplet set to obtain a sorting result, wherein when the number of elements in each behavior entity triplet set is larger, the sorting position of the initial behavior entity triplet corresponding to each behavior entity triplet set in the sorting result is higher; and generating the personal knowledge graph according to the M initial static entity triples, the first Q initial behavior entity triples in the ordered I initial behavior entity triples and the personal knowledge graph framework.
8. The method of claim 5, wherein generating the personal knowledge-graph from the M initial static entity triples, the E initial behavioral entity triples, the X initial target entity triples, and the personal knowledge-graph architecture comprises:

   generating the personal knowledge graph according to the M initial static entity triples, F initial behavior entity triples in the E initial behavior entity triples, Y initial target entity triples in the X initial target entity triples and the personal knowledge graph framework;

   the F initial behavior entity triples are obtained according to first user behavior data and the entity relation set, wherein the first user behavior data are behavior data in which the difference value between the acquisition time and the current system time in the initial user behavior data is smaller than a preset threshold value; the Y initial target entity triples are generated according to the M initial static entity triples and/or the F initial behavior entity triples, and Y is an integer greater than or equal to zero.
9. The method of claim 8, wherein generating the personal knowledge graph from the M initial static entity triples, F initial behavior entity triples of the E initial behavior entity triples, Y initial target entity triples of the X initial target entity triples, and the personal knowledge graph architecture comprises:

   Performing de-duplication on the Y initial target entity triples to obtain J initial target entity triples, wherein J is an integer which is less than or equal to Y and is greater than or equal to zero;

   when the J is equal to a preset number Q, generating the personal knowledge graph according to the M initial static entity triples, the J initial target entity triples and the personal knowledge graph framework, wherein Q is an integer greater than or equal to zero;

   when the J is larger than the preset quantity Q, grouping the Y initial target entity triples according to the J initial target entity triples to obtain J target entity triplet sets, wherein the J target entity triplet sets are in one-to-one correspondence with the J initial target entity triples, and the initial target entity triples contained in each target entity triplet set are the same as the initial target entity triples corresponding to the target entity triplet sets; sorting the J initial target entity triples according to the number of elements in each target entity triplet set to obtain a sorting result, wherein when the number of elements in each target entity triplet set is larger, the sorting position of the initial target entity triplet corresponding to each target entity triplet set is higher in the sorting result; generating the personal knowledge graph according to the M initial static entity triples, the first Q initial target entity triples in the sorted J initial target entity triples and the personal knowledge graph framework;

   When the J is smaller than the preset quantity Q, grouping the F initial behavior entity triples according to the I initial behavior entity triples to obtain O behavior entity triplet sets, wherein the O behavior entity triplet sets are in one-to-one correspondence with the I initial behavior entity triples, and the initial behavior entity triples contained in each behavior entity triplet set are identical to the initial behavior entity triples corresponding to the behavior entity triplet sets; sorting the I initial behavior entity triples according to the number of elements in each behavior entity triplet set to obtain a sorting result, wherein when the number of the initial behavior entity triples contained in each behavior entity triplet set is larger, the sorting position of the initial behavior entity triplet corresponding to each behavior entity triplet set is higher in the result; and generating the personal knowledge graph according to the first Q-J initial behavior entity triples and the personal knowledge graph framework in the M initial static entity triples, the J initial target entity triples and the sequenced I initial behavior entity triples.
10. The method according to any one of claims 1-9, wherein the method further comprises:

    acquiring a target user static attribute and a target user behavior attribute, and updating the personal knowledge graph by utilizing the target user static attribute and the target user behavior attribute to obtain a target personal knowledge graph;

    acquiring second user portrait data and second user behavior data;

    obtaining K first static entity triples according to the second user portrait data and the entity relation set; obtaining L first behavior entity triples according to the second user behavior data and the entity relation set, wherein L and K are integers greater than or equal to zero;

    and updating the target personal knowledge graph according to the K first static entity triples and the L first behavioral entity triples.
11. The method according to any one of claims 1-9, wherein the method further comprises:

    acquiring third user portrait data and third user behavior data;

    obtaining R second static entity triples according to the third user portrait data and the entity relation set; obtaining S second behavior entity triples according to the second user behavior data and the entity relation set, wherein R and S are integers greater than or equal to zero;

    And updating the personal knowledge graph according to the R second static entity triples and the S second behavior entity triples.
12. A personal knowledge graph construction apparatus, the apparatus comprising:

    the acquisition unit is used for acquiring the initial user static attribute and the initial user behavior attribute; the initial user static attribute is used for representing personal information of a user, the initial user behavior attribute is used for representing the field to which an entity type corresponding to the user behavior belongs, the field is a set of entity types with the same characteristics, the entity types are a set of entities with the same characteristics or attributes, and the entities are things associated with the user and represented by nouns or words; acquiring initial user portrait data, initial user behavior data and an entity relation set, wherein the initial user portrait data corresponds to the initial user static attribute, the initial user behavior data corresponds to the initial user behavior attribute, and the entity relation set comprises a plurality of relations which are used for representing the relation between the entities;

    the processing unit is used for constructing a personal knowledge graph framework according to the initial user static attribute and the initial user behavior attribute, wherein the personal knowledge graph framework represents the relationship between the user and the personal information, and the entity type corresponding to the user and the user behavior belongs to the field; the method comprises the steps of obtaining M initial static entity triples according to the initial user portrait data and the entity relation set; e initial behavior entity triples are obtained according to the initial user behavior data and the entity relation set; the initial static entity triples are used for describing attribute relations between the users and the static entities or between the static entities, the initial behavior entity triples are used for describing behavior relations between the users and the behavior entities or between the behavior entities, and M and E are integers which are larger than or equal to zero;

    And the generating unit is used for generating the personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples and the personal knowledge graph framework.
13. The apparatus of claim 12, wherein in said deriving M initial static entity triples from said initial user representation data and said set of entity relationships, said processing unit is specifically configured to:

    obtaining A static entities from the initial user portrait data, wherein A is an integer greater than or equal to zero;

    and obtaining the M initial static entity triples according to the A static entities and the entity relation set.
14. The apparatus according to claim 12 or 13, wherein in terms of said deriving E initial behavioural entity triples from said initial user behavioural data and said set of entity relations, said processing unit is specifically configured to:

    b behavior entities and C behavior characters are obtained from the initial user behavior data, the C behavior characters correspond to O behavior entities in the B behavior entities, the behavior characters are used for representing the operation of the user on the O behavior entities, and the B, the C and the O are integers which are larger than or equal to zero;

    Obtaining G initial behavior entity triples according to the B behavior entities and the entity relation set;

    obtaining H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities and the entity relation set, wherein G and H are integers greater than or equal to zero;

    the G initial behavior entity triples are used for describing the relation among the B behavior entities; the H initial behavior entity triplets are used for describing the relation between the user and the O behavior entities, and the E initial behavior entity triplets comprise the G initial behavior entity triplets and the H initial behavior entity triplets.
15. The apparatus according to claim 14, wherein the processing unit is further specifically configured to:

    acquiring H operation times from the initial user behavior data;

    the obtaining H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities and the entity relationship set includes:

    and generating the H initial behavior entity triples according to the user, the C behavior characters, the O behavior entities, the H operation times and the entity relation set.
16. The apparatus according to any one of claims 12-14, wherein the generating unit is specifically configured to:

    generating X initial target entity triples according to the M initial static entity triples and/or the E initial behavior entity triples, wherein X is an integer greater than or equal to zero; each of the target entity triples comprises at least one target entity, the target entity triples being used to describe a relationship between target entities or between the user and the target entities;

    and generating the personal knowledge graph according to the M initial static entity triples, the E initial behavior entity triples, the X initial target entity triples and the personal knowledge graph framework.
17. The apparatus according to any one of claims 12-14, wherein the generating unit is specifically configured to:

    generating the personal knowledge graph according to the M initial static entity triples, F initial behavior entity triples in the E initial behavior entity triples and the personal knowledge graph framework;

    the F initial behavior entity triples are obtained according to first user behavior data and the entity relation set, the first user behavior data are behavior data, the difference value between the acquisition time and the current system time in the initial user behavior data is smaller than a preset threshold, and F is an integer larger than or equal to zero.
18. The apparatus according to claim 17, wherein in the aspect of generating the personal knowledge graph according to the M initial static entity triples, F initial behavior entity triples of the E initial behavior entity triples, and the personal knowledge graph architecture, the generating unit is specifically configured to:

    performing de-duplication on the F initial behavior entity triples to obtain I initial behavior entity triples, wherein I is an integer which is less than or equal to F and is greater than or equal to zero;

    when the I is smaller than or equal to a preset number Q, generating the personal knowledge graph according to the M initial static entity triples, the I initial behavior entity triples and the personal knowledge graph framework, wherein Q is an integer larger than or equal to zero;

    when the I is larger than the preset quantity Q, grouping the F initial behavior entity triples according to the I initial behavior entity triples to obtain O behavior entity triplet sets, wherein the O behavior entity triplet sets are in one-to-one correspondence with the I initial behavior entity triples, and the initial behavior entity triples contained in each behavior entity triplet set are identical to the initial behavior entity triples corresponding to the behavior entity triplet sets; sorting the I initial behavior entity triples according to the number of elements in each behavior entity triplet set to obtain a sorting result, wherein when the number of elements in each behavior entity triplet set is larger, the sorting position of the initial behavior entity triplet corresponding to each behavior entity triplet set in the sorting result is higher; and generating the personal knowledge graph according to the M initial static entity triples, the first Q initial behavior entity triples in the ordered I initial behavior entity triples and the personal knowledge graph framework.
19. The apparatus of claim 16, wherein the personal knowledge-graph is generated in the architecture according to the M initial static entity triples, the E initial behavioral entity triples, the X initial target entity triples, and the personal knowledge-graph, the generating unit is specifically configured to:

    generating the personal knowledge graph according to the M initial static entity triples, F initial behavior entity triples in the E initial behavior entity triples, Y initial target entity triples in the X initial target entity triples and the personal knowledge graph framework;

    the F initial behavior entity triples are obtained according to first user behavior data and the entity relation set, wherein the first user behavior data are behavior data in which the difference value between the acquisition time and the current system time in the initial user behavior data is smaller than a preset threshold value; the Y initial target entity triples are generated according to the M initial static entity triples and/or the F initial behavior entity triples, and Y is an integer greater than or equal to zero.
20. The apparatus of claim 19, wherein in the aspect of generating the personal knowledge graph spectrum according to the M initial static entity triples, the F initial behavior entity triples of the E initial behavior entity triples, the Y initial target entity triples of the X initial target entity triples, and the personal knowledge graph architecture, the generating unit is specifically configured to:

    Performing de-duplication on the Y initial target entity triples to obtain J initial target entity triples, wherein J is an integer which is less than or equal to Y and is greater than or equal to zero;

    when the J is equal to a preset number Q, generating the personal knowledge graph according to the M initial static entity triples, the J initial target entity triples and the personal knowledge graph framework, wherein Q is an integer greater than or equal to zero;

    when the J is larger than the preset quantity Q, grouping the Y initial target entity triples according to the J initial target entity triples to obtain J target entity triplet sets, wherein the J target entity triplet sets are in one-to-one correspondence with the J initial target entity triples, and the initial target entity triples contained in each target entity triplet set are the same as the initial target entity triples corresponding to the target entity triplet sets; sorting the J initial target entity triples according to the number of elements in each target entity triplet set to obtain a sorting result, wherein when the number of elements in each target entity triplet set is larger, the sorting position of the initial target entity triplet corresponding to each target entity triplet set is higher in the sorting result; generating the personal knowledge graph according to the M initial static entity triples, the first Q initial target entity triples in the sorted J initial target entity triples and the personal knowledge graph framework;

    When the J is smaller than the preset quantity Q, grouping the F initial behavior entity triples according to the I initial behavior entity triples to obtain O behavior entity triplet sets, wherein the O behavior entity triplet sets are in one-to-one correspondence with the I initial behavior entity triples, and the initial behavior entity triples contained in each behavior entity triplet set are identical to the initial behavior entity triples corresponding to the behavior entity triplet sets; sorting the I initial behavior entity triples according to the number of elements in each behavior entity triplet set to obtain a sorting result, wherein when the number of the initial behavior entity triples contained in each behavior entity triplet set is larger, the sorting position of the initial behavior entity triplet corresponding to each behavior entity triplet set is higher in the result; and generating the personal knowledge graph according to the first Q-J initial behavior entity triples and the personal knowledge graph framework in the M initial static entity triples, the J initial target entity triples and the sequenced I initial behavior entity triples.
21. The apparatus according to any one of claims 12-20, wherein:

    the acquisition unit is also used for acquiring the static attribute and the behavior attribute of the target user, the second user portrait data and the second user behavior data;

    the processing unit is further configured to update the personal knowledge graph by using the static attribute of the target user and the behavioral attribute of the target user to obtain a target personal knowledge graph; obtaining K first static entity triples according to the second user portrait data and the entity relation set; obtaining L first behavior entity triples according to the second user behavior data and the entity relation set, wherein L and K are integers greater than or equal to zero;

    the generating unit is further configured to update the target personal knowledge graph according to the K first static entity triples and the L first behavioral entity triples.
22. The apparatus according to any one of claims 12-20, wherein:

    the acquisition unit is also used for acquiring third user portrait data and third user behavior data;

    the processing unit is further used for obtaining R second static entity triples according to the third user portrait data and the entity relation set; obtaining S second behavior entity triples according to the second user behavior data and the entity relation set, wherein R and S are integers greater than or equal to zero;

    The generating unit is further configured to update the personal knowledge graph according to the R second static entity triples and the S second behavioral entity triples.
23. A computer readable storage medium storing program code for device execution, the program code comprising instructions for performing the method of any one of claims 1 to 11.
24. A chip comprising a processor and a data interface, the processor reading instructions stored on a memory via the data interface, performing the method of any of claims 1 to 11.