A knowledge graph application method based on big data

Abstract

The application discloses a kind of knowledge graph application methods based on big data, the method includes the following steps: step one: from the query submitted by user identifies the target entity and search intent that user wishes to find;Step two: the trigger word of target entity set is matched with the abstract trigger word of event class, based on the target entity matched out, the preference of user is predicted;Step three: each vertical field is based on the multi-angle recommendation of field in knowledge graph architecture;Step four: the application rule of multiple vertical fields is established, and the knowledge graph of target entity and matched event class item obtained from query is applied to multiple vertical fields, and the interest prediction value of user to event item recommendation is generated according to the user feature vector and event item feature vector extracted from knowledge graph, the present application has the characteristics of accurate matching and strong correlation recommendation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of knowledge graph application, and particularly relates to a knowledge graph application method based on big data. BACKGROUND

[0002] Most of the current search and recommendation systems are statistical models, and most of the traditional search and recommendation models are based on the historical behavior of users to generate results. From the historical search, click, shopping and browsing behavior of users, the statistical law of user behavior is formed. For users who do not conform to the statistical law, the traditional statistical model will only simply regard them as abnormal. However, the behavior of any user often has its motivation. The traditional model is difficult to capture the motivation behind the individual behavior due to lack of data, and can only simply transplant the behavior characteristics of the group to the individual. Therefore, the recommendation services provided by the existing recommendation system are all the same, which easily eliminates the individual characteristics of users or objects. Although there are a large number of personalized models, they only refine the application range of the statistical model to a certain extent and are difficult to change the essence of using group behavior characteristics to serve individuals. Therefore, the traditional statistical model mostly generates results based on historical behavior. For users who do not conform to the statistical law, they will be simply regarded as abnormal and cannot obtain the motivation behind the user. It is difficult to understand the user's intention and accurately match. Therefore, it is necessary to design a knowledge graph application system based on big data for accurate matching and strong correlation recommendation. SUMMARY

[0003] The present application relates to the technical field of knowledge graph application, and particularly relates to a knowledge graph application method based on big data.

[0004] In order to solve the above technical problems, the present application provides the following technical scheme: a knowledge graph application method based on big data, comprising the following steps:

[0005] Step one: identifying the target entity and search intention that the user hopes to find from the query submitted by the user;

[0006] Step two: matching the trigger word of the set target entity with the abstract trigger word of the event class, extracting personalized features based on the matched target entity, and predicting the user's preference;

[0007] Step three: multi-angle recommendation in the field based on the knowledge graph architecture of each vertical field;

[0008] Step four: establishing application rules of multiple vertical fields, and applying the knowledge graph of the target entity obtained from the query and the matched event class project to multiple vertical fields.

[0009] According to the above technical scheme, the step of identifying the target entity and search intention that the user hopes to find from the query submitted by the user comprises:

[0010] Obtaining the query data of the user under the search page of the system, jumping to the search main interface of the user through analyzing the webpage and taking the query keyword of the user as the medium, obtaining the target entity and the related content thereof according to the self-defined concept entity and the relationship between the entities in the application field, describing the attribute information of the entity itself by using the data attribute, and finding out the target entity and the related content thereof in the existing knowledge graph through the set query statement;

[0011] Obtaining the corresponding information class intent and transaction class intent according to the target entity and the related content thereof accessed by the user.

[0012] According to the technical scheme, the step of matching the trigger word of the target entity with the abstract trigger word of the event class comprises:

[0013] Setting the trigger word of the output target entity, and matching the trigger word with the abstract trigger word of the event class;

[0014] When the matching result is consistent, the object element and the environment element of the target entity are continuously matched;

[0015] When the matching result is inconsistent, the next event class in the event knowledge graph is continuously searched and matched;

[0016] When the trigger word, the object element and the environment element of the event are consistent with a certain event class in the event knowledge graph, the event class is stored, and the next event class in the event knowledge graph is continuously matched;

[0017] When there are multiple event classes in the finally matched event class set and the event classes have a classification relationship therebetween, the most bottom sub-event class is found and is taken as the event class to which the event instance belongs;

[0018] When the finally matched event class set is empty, it is judged that the target entity cannot find the corresponding event class in the event knowledge graph.

[0019] According to the technical scheme, the step of extracting the personalized feature based on the matched target entity and predicting the preference of the user comprises:

[0020] The target entity data and the search intent data matched and output are preprocessed;

[0021] An interactive feature matrix is formed based on the interaction of the project and the entity, each interactive matrix is combined with the associated entity, and the interactive matrix and the knowledge graph corresponding to the target entity based on the associated semantics are obtained;

[0022] The associated data information of the user node on the knowledge graph is fused to obtain the personalized feature of the user;

[0023] The correlation data information of the interaction nodes is aggregated in the knowledge graph to obtain personalized features of user correlation interaction data.

[0024] The user personalized features and the user correlation interaction data features are alternately extracted to obtain a user feature vector and a user correlation interaction feature vector.

[0025] The preference of the user is predicted through a prediction function.

[0026] According to the technical solution, the steps of the multi-angle recommendation of each vertical field based on the knowledge graph architecture include:

[0027] The final feature vector representation of the user target entity and the event item is extracted from the knowledge graph, and the preference is predicted by performing inner product operation on the user feature vector and the event item feature vector. Each value in the multi-dimensional vector can be regarded as an evaluation value of a type, that is, the first value in the vector represents a certain event item corresponding to the user entity target. The greater the value, the higher the preference of the user to the related event item. The greater the value of the event item feature vector, the higher the recommendation value of the event item.

[0028] According to the technical solution, the steps of applying the knowledge graph of the target entity obtained from the query and the matched event class item to the multi-vertical field include:

[0029] The knowledge graph application rules of the multi-vertical field are constructed, the target entity data of different fields is collected based on the data collected by the user in the corresponding vertical field webpage query end, the collected data set of the field is serialized, and the data is further cleaned and preprocessed to construct the knowledge triplets conforming to the knowledge link in the field knowledge graph. After feature extraction, the target json data is converted, and the json data is transmitted to the front and back ends for logical interaction, and the visualization of the knowledge graph application is formed according to the interaction and matching prediction.

[0030] According to the technical solution, the knowledge graph application system includes:

[0031] The entity semantic association module is configured to identify semantic associations between a target entity and corresponding event items that a user wishes to find from a query submitted by the user.

[0032] The knowledge graph analysis and matching module is configured to analyze and match a trigger word of the target entity and an abstract trigger word of an event class using a knowledge graph.

[0033] The knowledge graph application module is configured to apply the knowledge graph of the target entity obtained from the query and the matched event class item to the multi-vertical field.

[0034] Compared with the prior art, the application has the beneficial effects that: the application identifies the target entity and the search intention that the user wants to find from the query submitted by the user, matches the trigger word of the set target entity with the abstract trigger word of the event class, and extracts the personalized features based on the matched target entity, predicts the preference of the user, so that the multi-angle recommendation in the field is made based on the knowledge graph architecture in each vertical field; the knowledge graph of the target entity and the matched event class item obtained from the query is applied to the multi-vertical field, so that the user interest prediction value of the event item recommendation is generated based on the user feature vector and the event item feature vector extracted from the knowledge graph, so that the recommendation matching for the target entity is more accurate, and the recommendation relevance is stronger. BRIEF DESCRIPTION OF DRAWINGS

[0035] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, which together with the embodiments of the application are used to explain the application, and do not constitute a limitation on the application. In the drawings:

[0036] Figure 1 A flowchart of a knowledge graph application method based on big data provided by the first embodiment of the application;

[0037] Figure 2 A module composition schematic diagram of a knowledge graph application system based on big data provided by the second embodiment of the application. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0039] Embodiment one: Figure 1 A flowchart of a knowledge graph application method based on big data provided by the first embodiment of the application, the present embodiment can be applied to the scene of knowledge graph application, and the method can be executed by a knowledge graph application system based on big data provided by the present embodiment, as shown in the figure, the method specifically includes the following steps: Figure 1

[0040] Step one: identify the target entity and the search intention that the user wants to find from the query submitted by the user;

[0041] ​In the embodiment of the present application, the query data of the user is obtained under the search page of the system, the search main interface of the user is jumped to through analyzing the webpage and taking the user query keyword as a medium, the concept entity and the relationship between the entities are defined according to the application field, the attribute information of the entity itself is described by using the data attribute, the target entity and the related content are found out in the existing knowledge graph through the set query statement, the most relevant entity to the query is found out by giving the query represented by a group of entities, a concept is generated to explain the potential connection between the related entity and the entity in the query, the target optimization is performed by finding the most relevant entity, the introduction of the most relevant entity to the query can maintain the semantics of the original entity in the query as much as possible, the found result is sorted by entities, the result content is reasonably classified, and the user is presented in an organized manner, the related content outside the target entity is expanded, and the related entity is effectively presented to the user to improve the search diversity.

[0042] For example, the corresponding information class intent and transaction class intent are obtained according to the target entity and the related content of the user access query, that is, the information about a certain topic that the user wants to obtain is built, and then the transaction activity information including shopping and downloading internet resources is built through the network as a medium, the subject classification of entity search is solved by sorting the candidate entity categories, the syntax analysis is performed based on the classified subject, the part-of-speech tagging or syntax modification relationship of the keywords in the query is output, the semantic role corresponding to the keywords in the query is output, and the semantic association between the keywords in the query is obtained.

[0043] Step two: the trigger word of the set target entity is matched with the abstract trigger word of the event class, the personalized features are extracted based on the matched target entity, and the user's preference is predicted;

[0044] In the embodiment of the present application, the trigger word of the output target entity is set, the trigger word is matched with the abstract trigger word of the event class, and only when the matching result is consistent, the object element and the environment element of the target entity can be matched. If the matching is not matched, it is considered that the target entity is not consistent with the event class, the next event class in the event knowledge graph is searched and matched, the event object element matching and the event environment element matching are the same as above, when the trigger word, the object element and the environment element of the event are consistent with a certain event class in the event knowledge graph, the event class is stored, the next event class in the event knowledge graph is continuously matched, if there are multiple event classes in the finally matched event class set and the event classes have a classification relationship, the most bottom sub-event class is found and is taken as the event class to which the event instance belongs, if the finally matched event class set is empty, it is judged that the target entity cannot find the corresponding event class in the event knowledge graph.

[0045] An exemplary, the target entity data and search intent data matched output is preprocessed, since the recommended predicted item can correspond to the entity of the knowledge graph, different items can describe the same target entity, therefore the interaction feature matrix is formed based on the interaction of items and entities, each interaction matrix is combined with the associated entity, the interaction matrix corresponding to the target entity based on the associated semantics and the knowledge graph are obtained, the interaction matrix includes the historical click data of the user and the user, the knowledge graph includes the user node corresponding to the historical click data, the associated data information of the user node on the knowledge graph is fused to obtain the personalized feature of the user, the associated data information of the interaction node in the knowledge graph is aggregated to obtain the personalized feature of the user associated interaction data, the user personalized feature and the user associated interaction data feature are extracted alternately to obtain the user feature vector and the user associated interaction feature vector, the user personalized feature and the user associated interaction feature vector are iterated multiple times to make the waveform of the user personalized feature and the waveform of the user associated interaction personalized feature interact and interfere, the user feature vector and the user associated interaction feature vector are more deeply mined, the user feature vector and the user associated interaction feature vector are subjected to inner product operation, and the preference of the user is predicted through a prediction function.

[0046] Step three: each vertical field carries out multi-angle recommendation in the field based on the knowledge graph architecture;

[0047] In the embodiment of the application, the final feature vector representation of the user target entity and the event item is extracted from the knowledge graph, the preference is predicted by doing inner product operation on the user feature vector and the event item feature vector, each value in the multi-dimensional vector can be regarded as a type of evaluation value, that is, the first value in the vector represents a certain event item corresponding to the user entity target, the larger the value, the higher the preference of the user to the related event item, and the larger the value of the event item feature vector, the higher the recommendation value of the event item, therefore, after calculating the inner product, the larger the value of the user feature vector and the event item feature vector, the larger the inner product result, that is, the greater the interest of the user in the corresponding event item recommended based on the knowledge graph according to the searched target entity, and vice versa, the interest is smaller, the user representation vector and the event item representation vector are subjected to inner product operation, the predicted interest value of the event item corresponding to the user target entity is obtained through the inner product operation, the predicted value at this time is the value in the entire recommended range, the output value is limited between 0 and 1 by using an activation function on the range of the output value, the predicted interest value obtained at this time is the click probability of the user to the event item, the larger the click probability, the more likely the user to interact with the event item, therefore, the user feature vector and the event item feature vector extracted from the knowledge graph are used to generate the interest prediction value of the user to the event item recommendation.

[0048] Exemplary, the knowledge graph architecture model outputs the interaction value of the user multi-target entity and the multi-event item by inputting the user target entity queried and the path set of the matched event item, forms a recommendation scheme according to the interaction value, extracts the path between the target entity and the event item from the knowledge graph, projects the target entity and the entity type in the path into two embedding vectors, since there are different relationships between entities, even if the first and last entities are the same, the interaction relationship may also be different, therefore, the relationship between the two entities before and after the connection is also projected into an independent embedding vector, on the basis of the embedding vector sequence representing the path, the neural network is used to sequence the data information according to the long-term dependence of the user search input, the entire path representation about the target entity in the knowledge graph is obtained by calculating the embedded vector, after obtaining the path representation, the interaction score between the user target entity and the event item is calculated according to the path, for each path in the path set between the user target entity and the event item, the embedded vector of the path is calculated respectively, and the score set of the interaction between the user target entity and the event item is obtained, the interaction scores are collected as total scores, based on the fact that the higher the score is, the greater the weight is when weighting, by setting the weight hyperparameter, the recommendation reason is provided according to the path information between the user target entity and the event item with the highest interaction score while the related event item recommendation result is provided to the user according to the target entity searched by the user, and different path types correspond to different recommendation reasons.

[0049] Step four: establish application rules of multiple vertical fields, and apply the knowledge graph of the target entity obtained from the query and the matched event class item to multiple vertical fields.

[0050] In the embodiment of the application, the knowledge graph application rules of multiple vertical fields are constructed, the target entity data of different fields is acquired based on the data information formed by the operation of the user on the webpage query end in the corresponding vertical field, the collected data set of the field is sequenced, and the data is further cleaned and pretreated, the knowledge triplets conforming to the knowledge link in the knowledge graph of the field are constructed, the target json data is obtained after feature extraction, and the json data is transmitted to the front and rear ends for logical interaction, and the visualization for the knowledge graph application is formed according to the interaction and matching prediction.

[0051] Exemplarily, the user uses the Web service intelligent search to realize single-keyword search, multi-keyword search and demand-based search functions. The single-keyword search function refers to returning Web services containing the input keyword of the user when the user inputs a Web service keyword. In the multi-keyword search function, the user selects relevant attribute information such as the function, provider and type of the Web service through multiple tabs, so as to obtain Web services meeting the conditions. The demand search is that when the user inputs the target demand in the search box, the front-end interface transmits the demand to the back-end, the back-end analyzes the user demand, obtains the Web service meeting the user demand by using a Web service recommendation algorithm and returns the front-end, and the user clicks the related Web service to enter the Web service knowledge graph visualization interface after obtaining the Web service meeting the target demand through the search interface. In the knowledge graph visualization interface, the user can realize functions such as exploring the relationship between Web services and filtering the type of Web services. The interface directly displays the relationship between multiple entities in the form of a knowledge graph.

[0052] Embodiment two: the embodiment two of the present application provides a knowledge graph application system based on big data, Figure 2 A module composition schematic diagram of the knowledge graph application system based on big data provided by the embodiment two of the present application is shown in the figure, and the system comprises: Figure 2

[0053] An entity semantic association module is configured to identify semantic associations between a target entity and corresponding event items that a user wishes to find from a query submitted by the user.

[0054] A knowledge graph analysis matching module is configured to analyze and match a trigger word of the target entity and an abstract trigger word of an event class by using a knowledge graph.

[0055] A knowledge graph application module is configured to apply a knowledge graph of the target entity and the matched event class items obtained from the query to multiple vertical fields.

[0056] In some embodiments of the present application, the entity semantic association module comprises:

[0057] A target entity query module is configured to find a target entity and related content thereof in an existing knowledge graph by using a set query statement.

[0058] A search intent acquisition module is configured to acquire a corresponding search intent according to a target entity and related content thereof accessed by a user.

[0059] A keyword association module is configured to query semantic association information data between keywords.

[0060] In some embodiments of the present application, the knowledge graph analysis matching module comprises: ​

[0061] An event class matching module is configured to match the trigger word of the target entity with the abstract trigger word of the event class;

[0062] A feature extraction module is configured to alternately extract user personalized features and user associated interactive data features;

[0063] A user preference prediction module is configured to predict the user's preference through a prediction function;

[0064] A multi-angle recommendation module is configured to perform multi-angle recommendation in the field based on the knowledge graph architecture in each vertical field.

[0065] In some embodiments of the present application, the knowledge graph application module comprises:

[0066] An application rule establishment module is configured to establish application rules in multiple vertical fields;

[0067] A multi-field application module is configured to perform knowledge graph recommendation application based on the application rules in multiple fields;

[0068] A visual display module is configured to form a visual display for the knowledge graph application according to the interaction and matching prediction.

[0069] It should be noted that, in this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0070] Finally, it should be noted that the above description is only for the preferred embodiments of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent replacements to some technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A big data based knowledge graph application method, characterized in that: The method comprises the following steps: Step 1: identifying the target entity and search intention that the user wants to find from the query submitted by the user; Step 2: matching the trigger word of the set target entity with the abstract trigger word of the event class, extracting personalized features based on the matched target entity, and predicting the user's preference; Step 3: each vertical field makes multi-angle recommendations in the field based on the knowledge graph architecture; Step 4: establish application rules for multiple vertical fields, and apply the knowledge graph of the target entity obtained from the query and the matched event class items to multiple vertical fields; The step of matching the trigger word of the set target entity with the abstract trigger word of the event class comprises: Setting the trigger word of the output target entity, matching the trigger word with the abstract trigger word of the event class; When the matching result is consistent, continue to match the object elements and environmental elements of the target entity; When the matching result is not consistent, continue to search and match the next event class in the event knowledge graph; When the trigger word, object element and environmental element of the event can all match a certain event class in the event knowledge graph, store the event class and continue to match the next event class in the event knowledge graph; When there are multiple event classes in the final matched event class set and there is a classification relationship between the event classes, find the most bottom-level sub-event class and take it as the event class to which the event instance belongs; When the final matched event class set is empty, it is judged that the target entity cannot find the corresponding event class in the event knowledge graph; The step of extracting personalized features based on the matched target entity and predicting the user's preference comprises: Preprocessing the target entity data and search intention data matched and outputted; Forming an interaction feature matrix based on the interaction of items and entities, combining each interaction matrix with the associated entity to obtain an interaction matrix based on associated semantics and a knowledge graph corresponding to the target entity; Fusing the associated data information of the user node on the knowledge graph to obtain the personalized features of the user; Aggregating the associated data information of the interaction node in the knowledge graph to obtain the personalized features of the user associated interaction data; Extracting the user personalized features and the user associated interaction data features alternately to obtain the user feature vector and the user associated interaction feature vector respectively; The interaction matrix includes the user and the historical click data of the user, the knowledge graph includes the user node corresponding to the historical click data, and the inner product operation is performed on the user feature vector and the user associated interaction feature vector to predict the user's preference through a prediction function. 2.The big data based knowledge graph application method according to claim 1, characterized in that: The step of identifying the target entity and search intention that the user wants to find from the query submitted by the user comprises: Obtaining the query data of the user under the search page of the system, jumping to the search main interface of the user through analyzing the webpage and taking the user query keywords as the medium, defining the concept entity and the relationship between entities according to the application field, describing the attribute information of the entity itself by using the data attribute, and finding out the target entity and its related content in the existing knowledge graph through the set query statement; According to the target entity and its related content accessed by the user, a corresponding information class intention and a transaction class intention are acquired; The step of multi-angle recommendation of each vertical field based on the knowledge graph architecture includes: The final feature vector representation of the user target entity and the event item is extracted from the knowledge graph, and the preference is predicted by performing inner product operation on the user feature vector and the event item feature vector. Each value in the multi-dimensional vector is regarded as an evaluation value of a type, that is, the first value in the vector represents a certain event item corresponding to the user entity target. The greater the value, the higher the preference of the user to the related event item. The greater the value of the event item feature vector, the higher the recommendation value of the event item. 3.The knowledge graph application method based on big data according to claim 2, characterized in that: The step of applying the knowledge graph of the target entity and the matched event class item acquired from the query to the multi-vertical field includes: The knowledge graph application rules of the multi-vertical field are constructed, and the target entity data of different fields are acquired based on the data information formed by the user's operation on the web query end in the corresponding vertical field. The collected data set of the field is serialized, and the data is further cleaned and preprocessed to construct the knowledge triplets conforming to the knowledge link in the knowledge graph of the field. After feature extraction, the target json data is converted, and the json data is transmitted to the front and back ends for logical interaction. According to the interaction and matching prediction, the visualization for the knowledge graph application is formed. 4.The knowledge graph application system for performing the knowledge graph application method based on big data according to claim 3, characterized in that: The system includes: An entity semantic association module for identifying semantic association between a target entity and a corresponding event item that the user wants to find from a query submitted by the user; A knowledge graph analysis and matching module for analyzing and matching a trigger word of the target entity and an abstract trigger word of an event class by using a knowledge graph; A knowledge graph application module for applying the knowledge graph of the target entity and the matched event class item acquired from the query to a multi-vertical field.

5. The knowledge graph application system of claim 4, wherein: The entity semantic association module includes: A target entity query module for finding a target entity and its related content in an existing knowledge graph through a set query statement; A search intention acquisition module for acquiring a corresponding search intention according to a target entity and its related content accessed by the user; A keyword association module for querying semantic association information data between keywords.

6. The knowledge graph application system of claim 5, wherein: The knowledge graph analysis and matching module includes: An event class matching module for matching a trigger word of a target entity and an abstract trigger word of an event class; A feature extraction module for alternately extracting user personalized features and user associated interaction data features; A user preference prediction module for predicting user preferences by using a prediction function; A multi-angle recommendation module for multi-angle recommendation of each vertical field based on a knowledge graph architecture.

Images