Data management method and device based on knowledge graph

Abstract

The present application provides a data management method and device based on a knowledge graph, which relates to the field of information processing technology, including: obtaining a target news report of a news event and a pre-constructed news event knowledge graph; generating a knowledge graph event node based on the event elements of the news event extracted from the target news report; generating a location entity and an object entity based on the location and event object of the news event; updating the event entity knowledge graph, the location entity knowledge graph, and the object entity knowledge graph based on the knowledge graph event node, the location entity, and the object entity; obtaining an updated news event knowledge graph based on the updated event entity knowledge graph, the updated location entity knowledge graph, and the updated object entity knowledge graph. The present application obtains a comprehensive and reliable updated knowledge graph by uniformly updating the person entities and location entities of the news event and the relationships between different entities.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and more specifically, to a data management method and device based on a knowledge graph. Background Art

[0002] Online media platforms are an important way for people to obtain news data; various news media and portal websites will publish and display news on the Internet. However, news reports released by multiple news media on the same news event are often the same, and portal websites may lack content verification when displaying news reports released by different news media, resulting in the content displayed on the news page being thin, messy and repetitive.

[0003] At the same time, if users want to search for places or people in news events that they are interested in, they can only enter keywords and browse them one by one to find them sporadically. This is not only inefficient and the search content is not comprehensive, but the search content cannot intuitively show the relationship between different news events, nor can it determine whether the people or places involved in a news event are related to other news events or other people or places.

[0004] There is no existing method for data management based on knowledge graphs. Summary of the invention

[0005] The purpose of the embodiments of the present application is to provide a data management method and device based on a knowledge graph to solve the above-mentioned problems existing in the prior art. Data can be managed based on the knowledge graph, and the news event knowledge graph can be updated based on news events, so as to comprehensively and intuitively display to users the relationship between news events and between the places where news events occur and the people involved.

[0006] In a first aspect, a data management method based on a knowledge graph is provided, which may include:

[0007] Obtaining a target news report of a news event and a pre-constructed news event knowledge graph; wherein the news event knowledge graph includes an event entity knowledge graph, a location entity knowledge graph, and an object entity knowledge graph;

[0008] Extracting event elements of the news event from the target news report; wherein the event elements include the occurrence location and the event object;

[0009] Based on the event elements, generate a knowledge graph event node;

[0010] Adding the knowledge graph event node to the event entity knowledge graph to obtain an updated event entity knowledge graph;

[0011] From the place entity knowledge graph, match the place entity that is the same as the occurrence place to obtain a target place entity;

[0012] Establishing an association relationship between the knowledge graph event node and the target location entity to obtain an updated location entity knowledge graph;

[0013] From the object entity knowledge graph, match the object entity identical to the event object to obtain a target object entity;

[0014] Establishing an association relationship between the knowledge graph event node and the target object entity to obtain an updated object entity knowledge graph;

[0015] Based on the updated event entity knowledge graph, the updated place entity knowledge graph and the updated object entity knowledge graph, an updated news event knowledge graph is obtained.

[0016] In an optional implementation, the method for acquiring the target news report includes:

[0017] Acquire all news reports related to the news event; wherein the news reports include news reports without news sources marked and news reports with news sources marked; any news report includes news content and release time;

[0018] Extract multiple news sources from all news reports with marked news sources; remove duplicates from the obtained multiple news sources to obtain duplicated news sources;

[0019] Extract the news reports published by each deduplicated news source to obtain multiple first news reports;

[0020] Sorting the first news reports according to the order of release time to obtain a plurality of sorted first news reports;

[0021] The first news report that is released first is regarded as the first target news report;

[0022] After deleting the first news reports having the same news content as the first target news report from the sorted plurality of first news reports, a filtered first news report is obtained;

[0023] From all news reports without marked news sources, delete news reports with the same news content as the first target news report and published later than the first target news report, to obtain filtered news reports without marked news sources;

[0024] A target news report is obtained based on the first target news report, the filtered first news report, and the filtered news reports without marked news sources.

[0025] In an optional implementation, the news report further includes: a news ID that published the news report;

[0026] Obtaining a target news report based on the first target news report, the filtered first news report, and the filtered news reports without marked news sources includes:

[0027] Using the news IDs of the first target news report, the first news report after screening, and the news reports without news source marked after screening as target news IDs;

[0028] Matching the evaluation value corresponding to the target news ID from a pre-constructed news ID and evaluation value comparison table;

[0029] Obtaining a second news report based on the first target news report, the filtered first news report, and the filtered news report without a marked news source;

[0030] After deduplication of the obtained second news report, a deduplicated second news report is obtained;

[0031] From the second news reports after deduplication, delete the second news reports whose evaluation values ​​of the corresponding news IDs are less than a preset evaluation value threshold, to obtain a third news report;

[0032] Analyzing the news content of the third news report to obtain a content analysis result of the third news report;

[0033] Based on the content analysis result of the third news reports, marking the different parts of the content between the third news reports to obtain marked third news reports;

[0034] The marked third news reports, news IDs corresponding to different third news reports, and release times are sorted in chronological order of release time to obtain target news reports.

[0035] In an optional implementation, before extracting the event elements of the news event from the target news report, the method further includes:

[0036] Obtaining the event type of the news event;

[0037] Matching the event element extraction model corresponding to the event type from a pre-constructed comparison table of event types and event element extraction models to obtain a target event element extraction model;

[0038] Extracting event elements of the news event from the target news report includes:

[0039] The news content of the target news report is input into the target event element extraction model to obtain the event elements of the news event.

[0040] In an optional implementation, the event elements further include: news summary;

[0041] Before generating a knowledge graph event node based on the event element, the method further includes:

[0042] Preprocessing the news content of the target news report to obtain preprocessed news content;

[0043] Performing word segmentation processing on the preprocessed news content to obtain a plurality of extracted words;

[0044] For any extracted word, counting the frequency of occurrence of the extracted word in the pre-processed news content to obtain frequency data of the extracted word;

[0045] Perform cluster analysis on each extracted word to obtain cluster analysis results;

[0046] Determining keywords in the extracted words based on the frequency data of different extracted words and the cluster analysis results;

[0047] Extracting sentences and paragraphs containing the keywords from the preprocessed news content;

[0048] Performing semantic analysis on sentences and paragraphs containing the keywords to obtain semantic analysis results;

[0049] Based on the keywords and the semantic analysis result, a news summary of the target news report is generated.

[0050] In an optional implementation, before generating a knowledge graph event node based on the event element, the method further includes:

[0051] For any event object, based on the event object, a target news report is analyzed to determine the type of association between the event object and the news event;

[0052] Based on the association type between the event object and the news event, determining the association relationship between the event object and other event objects;

[0053] Based on the association type between the event object and the news event, the association type between the event object and the location where the news event occurs is obtained.

[0054] In an optional implementation, generating a knowledge graph event node based on the event element includes:

[0055] The place of occurrence is used as the place entity of the knowledge graph event node;

[0056] Take each event object as the object entity of the knowledge graph event node;

[0057] Based on the association relationship between the event objects, establishing the relationship between the location entity and different object entities;

[0058] Establishing relationships between different object entities based on the association types between different event objects and the news events and the association types between different event objects and the locations where the news events occurred;

[0059] Using the news summary as an attribute of a knowledge graph event node;

[0060] The knowledge graph event node is obtained based on the place entity, each object entity, the attributes, the relationship between the place entity and different object entities, and the relationship between different object entities.

[0061] In a second aspect, a data management device based on a knowledge graph is provided, which may include:

[0062] An acquisition unit, used to acquire a target news report of a news event and a pre-constructed news event knowledge graph; wherein the news event knowledge graph includes an event entity knowledge graph, a location entity knowledge graph, and an object entity knowledge graph;

[0063] An extraction unit, used to extract event elements of the news event from the target news report; wherein the event elements include the occurrence location and the event object;

[0064] A generating unit, used for generating a knowledge graph event node based on the event elements;

[0065] An updating unit is used to add the knowledge graph event node to the event entity knowledge graph to obtain an updated event entity knowledge graph; match the place entity that is the same as the occurrence place from the place entity knowledge graph to obtain a target place entity; establish an association relationship between the knowledge graph event node and the target place entity to obtain an updated place entity knowledge graph; match the object entity that is the same as the event object from the object entity knowledge graph to obtain a target object entity; establish an association relationship between the knowledge graph event node and the target object entity to obtain an updated object entity knowledge graph; and obtain an updated news event knowledge graph based on the updated event entity knowledge graph, the updated place entity knowledge graph and the updated object entity knowledge graph.

[0066] In a third aspect, an electronic device is provided, the electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0067] Memory, used to store computer programs;

[0068] The processor is used to implement any method step described in the first aspect when executing the program stored in the memory.

[0069] In a fourth aspect, a computer-readable storage medium is provided, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, any method step described in the first aspect is implemented.

[0070] This application organizes and analyzes the data contained in news reports based on news events, thereby achieving data management in order to present comprehensive and accurate data to users.

[0071] By checking for duplicates, eliminating similarities and retaining differences, and integrating news reports released by different news media, the present application can systematically display different news reports released by different news media on the same news event at one time, thus avoiding a poor user experience caused by repeatedly viewing the same news report; by verifying news reports with different contents, the present application can screen out false news and ensure the authenticity and reliability of the news; by marking the different contents contained in different news reports, the present application can not only intuitively present the differences to users, but also present news reports on the same news event in a more comprehensive and rich manner.

[0072] The present application constructs a news event knowledge graph, creates event nodes for news events, creates entities for places and people in news events, and establishes relationships between event nodes, places and people based on analyzing the associations between events, places and people. At the same time, based on big data search technology, it retrieves people, places and other data related to the news event, and adds attributes to news events, places and people. When presenting a news event, it does not simply present the report content of the news event, but automatically associates and analyzes the places, people and other news event data related to the news event, thereby efficiently and comprehensively displaying all relevant content of the news event without the need for users to search separately. At the same time, based on the relationship between different entities, the present application can intuitively display the association between different people or different places or different news events.

[0073] This application generates an event node for each news event, generates different entities based on the people and places in the news event, and updates the knowledge graph entities based on the generated different entities; then, by analyzing the associations between people, places and events, and establishing relationships between different entities, it realizes the construction and update of the news event knowledge graph. At the same time, the update of this application does not simply update the event nodes, but uniformly updates the relationships between person entities, place entities and different entities. The update is comprehensive, and the updated knowledge graph is comprehensive and reliable, which can intuitively display the associations between entities and the associations between entities and event nodes. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0075] Figure 1 A flow chart of a data management method based on knowledge graph provided in an embodiment of the present application;

[0076] Figure 2 A schematic diagram of the structure of a data management device based on a knowledge graph provided in an embodiment of the present application;

[0077] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0078] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0079] The data management method based on knowledge graph provided in the embodiment of the present application can be applied in a server or in a terminal with strong computing power, and the method can be applied to a database management system. The server can be a physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and basic cloud computing services such as big data and artificial intelligence platforms. The terminal can be a user equipment (User Equipment, UE) such as a mobile phone, a smart phone, a laptop, a digital broadcast receiver, a personal digital assistant (PDA), a tablet computer (PAD), a handheld device, a vehicle-mounted device, a wearable device, a computing device or other processing equipment connected to a wireless modem, a mobile station (MobileStation, MS), a mobile terminal (Mobile Terminal), etc. The terminal and the server can be directly or indirectly connected by wired or wireless communication, which is not limited in this application.

[0080] The preferred embodiments of the present application are described below in conjunction with the drawings in the specification. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application. In addition, the embodiments and features in the embodiments of the present application may be combined with each other if there is no conflict.

[0081] Figure 1 A flow chart of a data management method based on a knowledge graph provided in an embodiment of the present application. Figure 1 As shown, the method may include:

[0082] Step S110, obtaining a target news report of a news event and a pre-constructed news event knowledge graph; extracting event elements of the news event from the target news report; and generating a knowledge graph event node based on the event elements.

[0083] Among them, the news event knowledge graph includes the event entity knowledge graph, the location entity knowledge graph and the object entity knowledge graph; the event elements of a news event include the location of occurrence and the event object.

[0084] In an embodiment of the present application, when any news event occurs, crawler technology or other big data technology is used to collect news reports about the news event published by various news websites or different news media on their public accounts, and by processing all the collected news reports about the news event, a target news report of the news event is obtained.

[0085] In an embodiment of the present application, news reports include news reports without marked news sources and news reports with marked news sources; specifically, the news reports can be determined as news reports with marked news sources and news reports without marked news sources by extracting whether "source" or "news source" or other keywords exist in the news reports.

[0086] In the embodiment of the present application, it is also possible to determine whether the news report is a news report with a marked news source or a news report without a marked news source by judging whether the news ID contained in the news report is consistent with the news ID that publishes the news report; for example, Newspaper One publishes a news report with Title A, but the news content of Title A contains "Newspaper Two, a certain place, May 14" or other similar descriptions, then it means that the news report with Title A is a news report with a marked news source; and its news source is Newspaper Two.

[0087] In the embodiment of the present application, any news report includes a news title, news content, release time and a news ID that releases the news report; the news ID can be the full name, abbreviation or other identifier that can uniquely identify the news media.

[0088] In an embodiment of the present application, a method for acquiring a target news report includes:

[0089] Acquire all news reports related to the news event; extract multiple news sources from all news reports with marked news sources; remove duplicates from the obtained multiple news sources to obtain duplicated news sources; extract news reports published by each duplicated news source to obtain multiple first news reports;

[0090] The first news reports are sorted in the order of their release time to obtain a plurality of sorted first news reports; the first news report that is released first is used as the first target news report; from the sorted plurality of first news reports, first news reports having news contents identical to those of the first target news report are deleted to obtain screened first news reports; from all news reports without marked news sources, news reports having news contents identical to those of the first target news report and released later than the first target news report are deleted to obtain screened news reports without marked news sources; based on the first target news report, the screened first news reports and the screened news reports without marked news sources, a target news report is obtained.

[0091] In the embodiment of the present application, since multiple news media may cite news reports published by the same media, the news sources extracted from all news reports marked with news sources may be repeated, so it is necessary to first remove the duplicates to obtain multiple different news sources; at the same time, the reports marked with news sources are exactly the same as the news reports published by the news media (or news ID) that is their news source, so it is only necessary to retain the news reports published by the news media marked as the news source, and obtain multiple first news reports from different news sources.

[0092] For example, Newspaper One publishes news report B with the news source marked as XX Newspaper; News report B published by Newspaper One is exactly the same as News report B published by XX Newspaper, so only News report B published by XX Newspaper needs to be retained.

[0093] In the embodiment of the present application, since the first news reports from different news sources may have the same content, it is necessary to deduplicate multiple first news reports, take the first news report with the earliest release time as the first target news report, and delete the news reports that are the same as the first target news report from all the first news reports, so that only the one with the earliest release time is retained among the multiple news reports with the same content, and obtain the filtered first news report.

[0094] Similarly, among the news reports whose news sources are not marked, there may also be news reports that are the same as the first target news report. Therefore, it is also necessary to delete the news reports whose news sources are not marked and are the same as the first target news report and are published later than the first target news report, so as to retain multiple news reports with different contents, that is, the news reports without news sources that have been screened.

[0095] For example, there are 8 news reports, namely A, B, C, D, E, F, G and H; among them, A, B, C and D are news reports with marked news sources, and their news sources are XX newspaper, XX newspaper, newspaper three and newspaper four respectively, while E, F, G and H are news reports without marked news sources; the news report released by XX newspaper is M, the news report released by newspaper three is F, and the news report released by newspaper four is G; the release time of M is earlier than the release time of F; the release time of F is earlier than the release time of G; the contents of M, F and E are exactly the same.

[0096] Extract news sources to obtain XX Newspaper, Newspaper Three and Newspaper Four; extract news reports released by XX Newspaper and Newspaper Three as M, F and G respectively; take M as the first target news report, delete F which has exactly the same content as M, and obtain the first news report G after screening; from E, F, G and H, delete F and E which have exactly the same content as M and are released later than M, and obtain the screened news reports G and H without marked news sources.

[0097] In the embodiment of the present application, the target news report is obtained based on the first target news report, the filtered first news report and the filtered news reports without marked news sources, including:

[0098] The news IDs of the first target news report, the first news report after screening, and the news report without news source after screening are used as the target news ID; the evaluation value corresponding to the target news ID is matched from the pre-constructed news ID and evaluation value comparison table; the second news report is obtained based on the first target news report, the first news report after screening, and the news report without news source after screening; after deduplication of the obtained second news report, a deduplicated second news report is obtained; from the deduplicated second news report, the second news report whose evaluation value of the corresponding news ID is less than a preset evaluation value threshold is deleted to obtain a third news report;

[0099] The news content of the third news reports is analyzed to obtain content analysis results of the third news reports; based on the content analysis results of the third news reports, different parts of the content between the third news reports are marked to obtain marked third news reports; the marked third news reports, news IDs corresponding to different third news reports, and release times are sorted in chronological order of release time to obtain target news reports.

[0100] In the embodiment of the present application, the evaluation value corresponding to the target news ID represents the authority of the source.

[0101] In the embodiment of the present application, the calculation formula of the evaluation value corresponding to the target news ID includes:

[0102] ;

[0103] Among them, A, E, F, M, and S represent reporting accuracy, expert citation rate, error correction frequency, false information record, and reader satisfaction survey, respectively; 、 w 、 wxya 、 wM and wS represent the weights of reporting accuracy, expert citation rate, error correction frequency, false information record and reader satisfaction survey respectively; among them, error correction frequency and false information record are penalty items, and negative weights are used to reflect the penalty effect.

[0104] In the embodiment of the present application, the evaluation indicators include: reporting accuracy, expert citation rate (the proportion of citing expert opinions), error correction frequency (the frequency of the media correcting erroneous reports), false information record (the media's past record of false information), and reader satisfaction survey (obtaining readers' satisfaction scores through questionnaires, etc.).

[0105] In other embodiments of the present application, the evaluation value corresponding to the target news ID may also be calculated based on other factors.

[0106] In an embodiment of the present application, a method for determining the weight of an evaluation index includes:

[0107] For any target news ID, obtain the historical indicator data of the news ID, including user feedback, expert opinions and the results of historical data analysis; use machine learning algorithms (such as linear regression, decision trees, neural networks, etc.) to train the weight optimization model, and optimize the model's hyperparameters through cross-validation and other methods; use the trained weight optimization model to output the weights of each evaluation indicator; as new data is continuously added, regularly update the weight optimization model to ensure the timeliness and accuracy of the weights.

[0108] For example, based on the news IDs of the first target news report (M), the first news report after screening (G), and the news reports without marked news sources after screening (G and H), the target news IDs are: XX Newspaper, Newspaper Four, and Newspaper Five (H); the evaluation values ​​of matching the three news IDs are 95, 92, and 20 respectively; M, G, G, and H are taken as the second news report, and M, G, and H are obtained after deduplication; assuming that the evaluation value threshold is 80, H is deleted, and the third news report is M and G; the third news report M and G is analyzed, and assuming that the only difference between the two is the location, the location information is labeled to obtain the labeled third news report; the labeled M and G, the news IDs corresponding to M and G, and the release time are sorted in the order of release time to obtain the target news report.

[0109] This application eliminates duplicate news reports through content deduplication to prevent users from seeing multiple duplicate reports. At the same time, it uses the release time as the deletion standard, that is, if the content is the same, the first published news report is retained, so as to facilitate the determination of the news source and the earliest news report published by the source; and for news reports that are different from the news reports published by the source, it is necessary to use evaluation values ​​to filter them to improve credibility; this application verifies the evaluation value in order to eliminate false information or unverified information mixed therein.

[0110] When different information is released by a news ID whose evaluation value exceeds the threshold, only the information will be marked to highlight the different information so that users can view it intuitively; the number of objects involved in the event at different times may change, so due to the different release times, the data contained in the news will be different. Therefore, as long as the evaluation value of the news ID exceeds the preset threshold of this application, this application will not filter different news reports, thereby retaining complete news reports and avoiding untimely and incomplete content. Instead, it can display continuous news reports in chronological order.

[0111] In another embodiment of the present application, the credibility of the different annotated contents may also be analyzed and judged, thereby further improving the reliability and accuracy of the target news report.

[0112] In an embodiment of the present application, news events can be divided according to event types. For example, news events of different event types such as sports news, entertainment news, financial news, etc. require different event elements; for example, entertainment news events require time, place, people, event name, etc.

[0113] However, using a general event element extraction model or only using keyword extraction may not be able to accurately and specifically extract the event elements required for this type of news event. Therefore, this application pre-constructs and trains a variety of event element extraction models using news reports of different event types, and matches the corresponding event element extraction model based on the event type of the news event, so that the corresponding event elements can be obtained more accurately and specifically.

[0114] In an embodiment of the present application, before extracting event elements of a news event from a target news report, the method also includes: obtaining the event type of the news event; matching the event element extraction model corresponding to the event type from a pre-constructed comparison table of event types and event element extraction models to obtain a target event element extraction model.

[0115] In an embodiment of the present application, event elements of a news event are extracted from a target news report, including: inputting the news content of the target news report into a target event element extraction model to obtain the event elements of the news event.

[0116] In another embodiment of the present application, a place extraction model, a time extraction model, and a person extraction model may be constructed separately, so as to extract the place, time, and person in the target news report respectively.

[0117] The training method of the event element extraction model of the present application includes:

[0118] Obtain multiple types of public news reports, and add annotations to each news report, where the annotation content is event elements; use the news reports and the corresponding annotated data as training data sets to train a pre-built event element extraction model for news events, and obtain a trained event element extraction model for news events.

[0119] In an embodiment of the present application, before generating a knowledge graph event node based on the event element, the method further includes:

[0120] Preprocess the news content of the target news report to obtain the preprocessed news content; perform word segmentation on the preprocessed news content to obtain multiple extracted words; for any extracted word, count the frequency of the extracted word in the preprocessed news content to obtain the frequency data of the extracted word; perform cluster analysis on each extracted word to obtain the cluster analysis result; determine the keywords in the extracted word based on the frequency data of different extracted words and the cluster analysis results; extract sentences and paragraphs containing keywords from the preprocessed news content; perform semantic analysis on the sentences and paragraphs containing keywords to obtain the semantic analysis result; generate a news summary of the target news report based on the keywords and the semantic analysis result.

[0121] In actual applications, news reports of some news events can be generated based on event elements, such as news reports of sports news events, where someone participated in a certain game and what results were obtained; however, the event elements of some news events cannot summarize and express the news summary of the news events. Therefore, this application also performs word segmentation, clustering analysis and semantic analysis on the target news reports to generate news summaries.

[0122] In another embodiment of the present application, it can be determined based on the event type of the news event whether the news summary of the target news report of the news event is generated only by event elements or needs to be generated through processing and analysis.

[0123] In the embodiment of the present application, the news content of the target news report is preprocessed to obtain the preprocessed news content, including: removing modal particles and conjunctions in the news content.

[0124] In an embodiment of the present application, before generating a knowledge graph event node based on the event element, the method further includes:

[0125] For any event object, based on the event object, the target news report is analyzed to determine the association type between the event object and the news event; based on the association type between the event object and the news event, the association relationship between the event object and other event objects is determined; based on the association type between the event object and the news event, the association type between the event object and the location where the news event occurs is obtained.

[0126] In an embodiment of the present application, event objects can be divided into multiple object types such as parties, witnesses, photographers, reporters and persons in charge; parties can be further subdivided; and different types of event objects have different association types with news events; the object type can be used as the association type with the news event; that is, the association type between photographer A and event B is photographer; and after determining the association type between each event object and the news event, the relationship between different event objects can be determined; further, based on the association type between each event object and the news event, the relationship with the location where the news event occurred is determined.

[0127] In one embodiment of the present application, the classification of object types can be achieved through a variety of methods such as keyword extraction and semantic analysis.

[0128] In the embodiment of the present application, based on the event elements, a knowledge graph event node is generated, including:

[0129] The place of occurrence is taken as the place entity of the knowledge graph event node; each event object is taken as the object entity of the knowledge graph event node; based on the association relationship between each event object, the relationship between the place entity and different object entities is established; based on the association type between different event objects and news events and the association type between different event objects and the place of news events, the relationship between different object entities is established; the news summary is taken as the attribute of the knowledge graph event node; based on the place entity, each object entity, attributes, the relationship between the place entity and different object entities and the relationship between different object entities, the knowledge graph event node is obtained.

[0130] In an embodiment of the present application, a news event will not only generate an event node (or event entity), but will also generate corresponding entities based on the occurrence area and each event object in the news event, and based on the previously determined association type or relationship as the relationship between different entities (i.e., the edge in the knowledge graph); based on the entities and relationships (i.e., nodes and edges), the corresponding news event knowledge graph is obtained.

[0131] In an embodiment of the present application, the generated news summary is used as an attribute of the knowledge graph event node.

[0132] In another embodiment of the present application, information such as articles and behaviors publicly released (or published) by each event object may be obtained as an attribute of the object entity of the event object.

[0133] Step S120: add the knowledge graph event node to the event entity knowledge graph to obtain an updated event entity knowledge graph; from the place entity knowledge graph, match the place entity that is the same as the occurrence place to obtain the target place entity.

[0134] In an embodiment of the present application, before adding a knowledge graph event node to an event entity knowledge graph, it is necessary to compare and analyze the knowledge graph event node with each knowledge graph event node in the event entity knowledge graph. If there are no identical nodes, the knowledge graph event node is added to the event entity knowledge graph; if there are identical nodes, the knowledge graph event node in the event entity knowledge graph is directly updated.

[0135] In an embodiment of the present application, if the place entity knowledge graph does not match the place entity that is the same as the place where the event occurred, a new target place entity is created in the place entity knowledge graph; at the same time, a relationship between the target place entity and the city or province to which it belongs is established.

[0136] Step S130: Establish an association relationship between the knowledge graph event node and the target location entity to obtain an updated location entity knowledge graph; match the object entity that is the same as the event object from the object entity knowledge graph to obtain the target object entity.

[0137] In an embodiment of the present application, an association relationship between a knowledge graph event node and a target location entity is established, that is, an edge is generated between the knowledge graph event node and the target location entity, and is marked as the place where the knowledge graph event node occurs.

[0138] In an embodiment of the present application, if an object entity identical to the object is not matched in the object entity knowledge graph, a new target object entity is created in the object entity knowledge graph.

[0139] In the embodiment of the present application, since the object cannot be uniquely identified based on the object name, gender, position, place of origin and other public information are added to the object as a label of the object; the object is identified as uniquely as possible through the label and the object name.

[0140] In the matching and generation of object entities, the name plus label matching method is adopted; at the same time, based on the association type between the event object and the news event, when updating or generating a new object entity, one or more new labels are generated for the event object.

[0141] Step S140, establish an association relationship between the knowledge graph event node and the target object entity to obtain an updated object entity knowledge graph; based on the updated event entity knowledge graph, the updated location entity knowledge graph and the updated object entity knowledge graph, obtain an updated news event knowledge graph.

[0142] In an embodiment of the present application, the news event knowledge graph not only includes the event entity knowledge graph, but also includes the location entity knowledge graph and the object entity knowledge graph; therefore, the present application achieves comprehensive and accurate updating of the news event knowledge graph by updating them separately.

[0143] In another embodiment of the present application, after obtaining the event entity knowledge graph of the news event, which also includes the location entity knowledge graph and the object entity knowledge graph, the time attribute of the news event can be further analyzed to determine the time attribute type of the news event. The time attribute type is determined according to the duration and occurrence time of the event, that is, whether the news event is a short-term news event, a short-term news event, etc.

[0144] In an embodiment of the present application, after determining the time attribute type of the news event, a status label can also be generated for the news event; for example, not started, ongoing, or ended; at the same time, based on the time attribute type of the news event, it can be determined whether to track or monitor the subsequent development of the news event and the corresponding tracking time threshold; for example, the tracking time threshold of a short-term news event can be set to 2 days; the tracking time threshold of a short-term news event is the duration of the news event plus 10 days; the tracking time threshold is the tracking time after the event ends.

[0145] In another embodiment of the present application, tracking end conditions for different types of news events can be set, and tracking continues until the tracking end conditions are met.

[0146] In an embodiment of the present application, multiple entities (objects and locations) of the news event are continuously tracked based on a preset tracking end condition or tracking duration, and the tracking results are used as attribute data of the object entity, event entity, and location entity.

[0147] In an embodiment of the present application, tracking of object entities may include obtaining articles publicly released (or published) by each event object, etc.; tracking of location entities may include obtaining other subsequent events at the corresponding location or whether the damage to the location caused by the news event has been repaired or the subsequent impact of the news event on the location.

[0148] In another embodiment of the present application, tracking of the object entity also includes: obtaining the followed objects of the object entity's social media account, analyzing the public comment data and published content of the followed objects and the object entity, and determining the relationship between the object entity and the followed object (for example, classmates); and then generating an object entity for the followed object, and establishing an association relationship between the object entity and the object entity of the followed object.

[0149] In an embodiment of the present application, the tracking of the location entity can be obtained by collecting and analyzing the notices issued by the account of the location or the account of the relevant department involved in the event.

[0150] In an embodiment of the present application, tracking of event entities may include analyzing whether there will be a triggering event for the event type; if there is a triggering event, continuously tracking the completion of the triggering event and the impact of the news event; if there is no triggering event, only tracking the subsequent impact caused by the news event; the acquisition of such impact data includes: determining through data published by relevant persons or subject accounts of the area where the news event occurred (such as a shopping mall’s official account), or persons closely related to the event object of the news event (determined through intimacy analysis) and other relevant accounts, or determining based on big data collection and analysis technology.

[0151] In another embodiment of the present application, footprints of the area and event where the news event occurred are established for each event object of any news event, and spatial or temporal intersection analysis can be performed on different event objects based on the established footprints; for example, party A participated in an event at a certain time in a certain place in city X; another party B participated in another event at a certain time in a shopping mall in the same city, then footprints containing time and place are generated for the two parties respectively, and intersection analysis shows that the two parties were at the same place in city X at the same time, less than 200 meters apart.

[0152] In another embodiment of the present application, after each update of the news event knowledge graph, a label can be generated for the news event, and correlation analysis and cluster analysis can be performed on the news event and other events, places and objects in the event knowledge graph based on the place entity, object entity and label, and the frequency and characteristics of occurrence of a certain region or a certain type of event can be counted or extracted.

[0153] In an embodiment of the present application, the data management method based on the knowledge graph also includes: providing a query interface, obtaining keywords input by the user through the query interface, matching the keywords to object entities, place entities or event nodes through keyword matching technology, and calculating the degree of association between the keywords and the matching content, and displaying the content that the user may expect to query in descending order of the degree of association; obtaining the user's click-to-view operation, and presenting other place entities, object entities and event entities related to the user's clicked object as the associative content of the user's clicked object in a preset area of ​​the display device.

[0154] In an embodiment of the present application, a user can perform association analysis and group analysis on places, objects or events with a certain label by inputting labels and keywords; for example, inputting two names at the same time to analyze whether the object entities corresponding to the two names are associated, and if so, what event is the cause of the association.

[0155] In another embodiment of the present application, a display component may be added to the display device so that the displayed content can be displayed in full screen, zoomed in and out, automatically laid out, manually refreshed, and downloaded.

[0156] In another embodiment of the present application, the data management method based on the knowledge graph can be applied to a variety of scenarios. Taking the enterprise scenario as an example, news events are limited to enterprise news events. Based on news events occurring in the enterprise (including the establishment of cooperative relationships), etc., the enterprise employees or partners involved in the enterprise news events are determined, and the enterprise employees or partners are taken as object entities. Based on the event elements of the enterprise news events, corresponding news events are generated; the place of occurrence (such as the headquarters location or the location of a branch company) is taken as a place entity, thereby generating or updating the corresponding news event knowledge graph.

[0157] By searching the news event knowledge graph for employees, locations or events, it is possible to clarify the employee's experience in the company, the relationship with partner companies, or the company's footprint in a certain location and what happened in the company over a certain period of time.

[0158] Corresponding to the above method, the embodiment of the present application also provides a data management device based on knowledge graph, such as Figure 2 As shown, the data management device based on the knowledge graph includes:

[0159] An acquisition unit 210 is used to acquire a target news report of a news event and a pre-constructed news event knowledge graph; wherein the news event knowledge graph includes an event entity knowledge graph, a location entity knowledge graph, and an object entity knowledge graph;

[0160] The extraction unit 220 is used to extract event elements of the news event from the target news report; wherein the event elements include the occurrence location and the event object;

[0161] A generating unit 230, configured to generate a knowledge graph event node based on event elements;

[0162] The updating unit 240 is used to add the knowledge graph event node to the event entity knowledge graph to obtain an updated event entity knowledge graph; match the place entity that is the same as the place of occurrence from the place entity knowledge graph to obtain the target place entity; establish an association relationship between the knowledge graph event node and the target place entity to obtain an updated place entity knowledge graph; match the object entity that is the same as the event object from the object entity knowledge graph to obtain the target object entity; establish an association relationship between the knowledge graph event node and the target object entity to obtain an updated object entity knowledge graph; based on the updated event entity knowledge graph, the updated place entity knowledge graph and the updated object entity knowledge graph, an updated news event knowledge graph is obtained.

[0163] The functions of each functional unit of the knowledge graph-based data management device provided in the above-mentioned embodiments of the present application can be realized through the above-mentioned method steps. Therefore, the specific working process and beneficial effects of each unit in the knowledge graph-based data management device provided in the embodiments of the present application will not be repeated here.

[0164] The present application also provides an electronic device, such as Figure 3 As shown, it includes a processor 310 , a communication interface 320 , a memory 330 and a communication bus 340 , wherein the processor 310 , the communication interface 320 , and the memory 330 communicate with each other via the communication bus 340 .

[0165] Memory 330, for storing computer programs;

[0166] The processor 310 is used to execute the program stored in the memory 330 to implement the following steps:

[0167] Obtaining target news reports of news events and a pre-constructed news event knowledge graph; wherein the news event knowledge graph includes an event entity knowledge graph, a location entity knowledge graph, and an object entity knowledge graph;

[0168] Extracting event elements of news events from target news reports; wherein event elements include occurrence location and event object;

[0169] Generate knowledge graph event nodes based on event elements;

[0170] Add the knowledge graph event node to the event entity knowledge graph to obtain an updated event entity knowledge graph; from the location entity knowledge graph, match the location entity that is the same as the occurrence location to obtain the target location entity; establish an association relationship between the knowledge graph event node and the target location entity to obtain an updated location entity knowledge graph; from the object entity knowledge graph, match the object entity that is the same as the event object to obtain the target object entity; establish an association relationship between the knowledge graph event node and the target object entity to obtain an updated object entity knowledge graph; based on the updated event entity knowledge graph, the updated location entity knowledge graph and the updated object entity knowledge graph, obtain an updated news event knowledge graph.

[0171] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0172] The communication interface is used for communication between the above electronic device and other devices.

[0173] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0174] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0175] The implementation methods and beneficial effects of the components of the electronic device in the above embodiments to solve the problems can be seen in Figure 1 The various steps in the illustrated embodiment are implemented, therefore, the specific working process and beneficial effects of the electronic device provided by the embodiment of the present application are not repeated here.

[0176] In another embodiment provided in the present application, a computer-readable storage medium is also provided, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes any of the knowledge graph-based data management methods in the above embodiments.

[0177] In another embodiment provided in the present application, a computer program product comprising instructions is also provided, which, when executed on a computer, enables the computer to execute any of the knowledge graph-based data management methods in the above-mentioned embodiments.

[0178] Those skilled in the art will appreciate that the embodiments in the present application can be provided as methods, systems, or computer program products. Therefore, the embodiments in the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments in the present application can be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0179] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0180] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0181] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0182] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present application.

[0183] Obviously, those skilled in the art can make various changes and modifications to the embodiments in the present application without departing from the spirit and scope of the embodiments in the present application. Thus, if these modifications and variations of the embodiments in the present application fall within the scope of the claims and their equivalents in the embodiments of the present application, the embodiments of the present application are also intended to include these modifications and variations.

Claims

1. A data management method based on knowledge graph, It is characterized in that The method comprises: Obtaining a target news report of a news event and a pre-constructed news event knowledge graph; wherein the news event knowledge graph includes an event entity knowledge graph, a location entity knowledge graph, and an object entity knowledge graph; Extracting event elements of the news event from the target news report; wherein the event elements include the occurrence location and the event object; Based on the event elements, generate a knowledge graph event node; Adding the knowledge graph event node to the event entity knowledge graph to obtain an updated event entity knowledge graph; From the place entity knowledge graph, match the place entity that is the same as the occurrence place to obtain a target place entity; Establishing an association relationship between the knowledge graph event node and the target location entity to obtain an updated location entity knowledge graph; From the object entity knowledge graph, match the object entity identical to the event object to obtain a target object entity; Establishing an association relationship between the knowledge graph event node and the target object entity to obtain an updated object entity knowledge graph; Based on the updated event entity knowledge graph, the updated place entity knowledge graph and the updated object entity knowledge graph, an updated news event knowledge graph is obtained.

2. The method according to claim 1, It is characterized in that The method for obtaining the target news report includes: Acquire all news reports related to the news event; wherein the news reports include news reports without news sources marked and news reports with news sources marked; any news report includes news content and release time; Extract multiple news sources from all news reports with marked news sources; remove duplicates from the obtained multiple news sources to obtain duplicated news sources; Extract the news reports published by each deduplicated news source to obtain multiple first news reports; Sorting the first news reports according to the order of release time to obtain a plurality of sorted first news reports; The first news report that is released first is regarded as the first target news report; After deleting the first news reports having the same news content as the first target news report from the sorted plurality of first news reports, a filtered first news report is obtained; From all news reports without marked news sources, delete news reports with the same news content as the first target news report and published later than the first target news report, to obtain filtered news reports without marked news sources; A target news report is obtained based on the first target news report, the filtered first news report, and the filtered news reports without marked news sources.

3. The method according to claim 2, It is characterized in that The news report also includes: a news ID that publishes the news report; wherein the news ID is a unique identifier of the news media that publishes the news report; Obtaining a target news report based on the first target news report, the filtered first news report, and the filtered news reports without marked news sources includes: Using the news IDs of the first target news report, the first news report after screening, and the news reports without news source marked after screening as target news IDs; Matching the evaluation value corresponding to the target news ID from a pre-constructed news ID and evaluation value comparison table; Obtaining a second news report based on the first target news report, the filtered first news report, and the filtered news report without a marked news source; After deduplication of the obtained second news report, a deduplicated second news report is obtained; From the second news reports after deduplication, delete the second news reports whose evaluation values ​​of the corresponding news IDs are less than a preset evaluation value threshold, to obtain a third news report; Analyzing the news content of the third news report to obtain a content analysis result of the third news report; Based on the content analysis result of the third news reports, marking the different parts of the content between the third news reports to obtain marked third news reports; The marked third news reports, news IDs corresponding to different third news reports, and release times are sorted in chronological order of release time to obtain target news reports.

4. The method according to claim 1, It is characterized in that Before extracting the event elements of the news event from the target news report, the method further includes: Obtaining the event type of the news event; Matching the event element extraction model corresponding to the event type from a pre-constructed comparison table of event types and event element extraction models to obtain a target event element extraction model; Extracting event elements of the news event from the target news report includes: The news content of the target news report is input into the target event element extraction model to obtain the event elements of the news event.

5. The method according to claim 4, It is characterized in that The event elements also include: news summary; Before generating a knowledge graph event node based on the event element, the method further includes: Preprocessing the news content of the target news report to obtain preprocessed news content; Performing word segmentation processing on the preprocessed news content to obtain a plurality of extracted words; For any extracted word, counting the frequency of occurrence of the extracted word in the pre-processed news content to obtain frequency data of the extracted word; Perform cluster analysis on each extracted word to obtain cluster analysis results; Determining keywords in the extracted words based on the frequency data of different extracted words and the cluster analysis results; Extracting sentences and paragraphs containing the keywords from the preprocessed news content; Performing semantic analysis on sentences and paragraphs containing the keywords to obtain semantic analysis results; Based on the keywords and the semantic analysis result, a news summary of the target news report is generated.

6. The method according to claim 5, It is characterized in that Before generating a knowledge graph event node based on the event element, the method further includes: For any event object, based on the event object, a target news report is analyzed to determine the type of association between the event object and the news event; Based on the association type between the event object and the news event, determining the association relationship between the event object and other event objects; Based on the association type between the event object and the news event, the association type between the event object and the location where the news event occurs is obtained.

7. The method according to claim 6, It is characterized in that Based on the event elements, a knowledge graph event node is generated, including: The place of occurrence is used as the place entity of the knowledge graph event node; Take each event object as the object entity of the knowledge graph event node; Based on the association relationship between the event objects, establishing the relationship between the location entity and different object entities; Establishing relationships between different object entities based on the association types between different event objects and the news events and the association types between different event objects and the locations where the news events occurred; Using the news summary as an attribute of a knowledge graph event node; The knowledge graph event node is obtained based on the place entity, each object entity, the attributes, the relationship between the place entity and different object entities, and the relationship between different object entities.

8. A data management device based on knowledge graph, It is characterized in that The device comprises: An acquisition unit, used to acquire a target news report of a news event and a pre-constructed news event knowledge graph; wherein the news event knowledge graph includes an event entity knowledge graph, a location entity knowledge graph, and an object entity knowledge graph; An extraction unit, used to extract event elements of the news event from the target news report; wherein the event elements include the occurrence location and the event object; A generating unit, used for generating a knowledge graph event node based on the event elements; An updating unit is used to add the knowledge graph event node to the event entity knowledge graph to obtain an updated event entity knowledge graph; match the place entity that is the same as the occurrence place from the place entity knowledge graph to obtain a target place entity; establish an association relationship between the knowledge graph event node and the target place entity to obtain an updated place entity knowledge graph; match the object entity that is the same as the event object from the object entity knowledge graph to obtain a target object entity; establish an association relationship between the knowledge graph event node and the target object entity to obtain an updated object entity knowledge graph; and obtain an updated news event knowledge graph based on the updated event entity knowledge graph, the updated place entity knowledge graph and the updated object entity knowledge graph.

9. An electronic device, It is characterized in that The electronic device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory, used to store computer programs; A processor, for implementing any of the methods described in claims 1-7 when executing a program stored in a memory.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images