Event context mining and pushing method and system based on directional tracking and program product

Abstract

The invention belongs to the technical field of information pushing, and particularly discloses an event venation mining and pushing method and system based on directional tracking and a program product. Through comprehensive event information extraction, deep understanding of the event is realized, the core content of the event can be accurately grasped, and the development context and change of the event can be effectively reflected in combination with construction and continuous updating of the event context graph, so that the detailed situation of subsequent development of the event can be intuitively known. Moreover, relevant information retrieval is performed for each event, and reliable and valuable update contents are screened out based on filtering conditions, so that the search information is ensured to be closely related to the events and have timeliness, and sufficient and accurate data support is provided for timely update of event venation. Meanwhile, by constructing the interest portrait of the user, the personalized interest of the user is accurately mastered, and then targeted related event update content is pushed, so that the attention demand of the user on the progress of the interested event can be met, and the user experience is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of information push, and in particular relates to an event context mining and push method, system and program product based on directional tracking. Background Art

[0002] With the rapid development of artificial intelligence technology, the field of news communication has entered the era of "intelligent media". Against this background, personalized news recommendation systems have developed rapidly. Relying on computer technology, statistical knowledge, and combining data algorithms, human-computer interaction, etc., they have established a personalized association mechanism between users and resources. Such systems improve the efficiency of news distribution through algorithm control, solve the problem of manual editing, and are user-oriented, prompting a return to the audience-oriented approach.

[0003] However, existing personalized news recommendation systems mainly push information based on association recommendations, that is, recommending related news based on news articles that users click on or are interested in. Although this method can provide a certain degree of personalized experience, it has obvious flaws: it fails to deeply understand the user's attention needs for the subsequent progress of the same event or related events. Users tend to be more concerned about the overall context and in-depth development of an event, rather than just isolated news points. Therefore, existing news recommendation systems are still insufficient in meeting users' in-depth information needs. Summary of the invention

[0004] The purpose of the present invention is to provide an event context mining and pushing method, system and program product based on directional tracking to solve the above-mentioned problems existing in the prior art.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] First, a method for event context mining and pushing based on directional tracking is provided, including:

[0007] For the pre-built event context map, extract the event attribute information corresponding to each event node;

[0008] Based on the event attribute information corresponding to each event node, search for the update attribute information of the corresponding event, and form an event update list based on the update attribute information of each event;

[0009] Using the event update list, the event attribute information corresponding to each event node in the event context map is updated to obtain an updated event context map;

[0010] Based on the event click behavior of the target user and the content information of the clicked event, a user profile of the target user is constructed, and the event of interest to the target user is determined according to the user profile of the target user;

[0011] Determine the corresponding event node and its updated attribute information in the updated event context map according to the event of interest and / or the clicked event of the target user;

[0012] The update event content is extracted from the update attribute information of the corresponding event node, and the update event content is pushed to the target user end.

[0013] In a possible design, before extracting event attribute information corresponding to each event node in the pre-constructed event context map, the method further includes:

[0014] Collect event information manuscripts corresponding to target events clicked and / or followed by users;

[0015] Extracting the text content of the target event from the event information manuscript, and extracting event attribute information from the text content using a preset large model based on the set first large model prompt word, to obtain event basic information of the target event, wherein the event basic information includes event content, event type, event elements, event status, keywords, entity words, summary, and ID of users who have clicked and / or followed;

[0016] Based on the set second largest model prompt word, the preset large model is used to remove duplication and conduct correlation analysis on the event basic information corresponding to each target event, so as to determine the correlation relationship between the target events corresponding to each event basic information;

[0017] According to each target event, the basic event information corresponding to each target event and the association relationship between each target event, an event context map is constructed in the graph database with each target event as an event node, the association relationship between the target events as an edge, and the basic event information as the event attribute information of the corresponding event node.

[0018] In a possible design, extracting the text content of the target event from the event information manuscript includes:

[0019] Extracting initial text content, audio content and video content from event information manuscripts;

[0020] Use corresponding speech recognition tools to perform language recognition on audio content and video content to obtain corresponding supplementary text content;

[0021] The initial text content is combined with the supplementary text content to obtain the text content of the target event.

[0022] In a possible design, searching for updated attribute information of a corresponding event based on the event attribute information corresponding to each event node includes:

[0023] Extract keywords, entity words and abstracts from the event attribute information corresponding to each event node, and use the keywords, entity words and abstracts to search for relevant information on the information search engine to obtain the corresponding event search text content;

[0024] Conditionally filtering the search text content based on the event attribute information of each target event to obtain filtered event search text content;

[0025] Based on the set first large model prompt word, the preset large model is used to extract event attribute information from the filtered event search text content to obtain the corresponding new event basic information;

[0026] Based on the set second largest model prompt word, the preset large model is used to perform deduplication and correlation analysis on the basic information of each newly added event, and the correlation relationship between the events corresponding to the basic information of each newly added event is determined;

[0027] The target event corresponding to each newly added event basic information is determined based on the association relationship between the events corresponding to each newly added event basic information, and each newly added event basic information is used as the updated attribute information of the corresponding target event.

[0028] In a possible design, the event update list is composed based on the updated attribute information of each event, including: entering each updated attribute information and the corresponding target event into a template table in pairs to obtain the event update list.

[0029] In a possible design, the user portrait of the target user is constructed based on the event click behavior of the target user and the content information of the clicked event, including:

[0030] Based on the target user's event click behavior and the content information of the clicked event, a collaborative filtering method is used to determine similar users of the target user and the events that the similar users have clicked and / or followed;

[0031] The events that similar users have clicked and / or followed are regarded as the events of interest to the target user, and a user profile of the target user is constructed based on the events of interest to the target user.

[0032] In one possible design, the method further includes:

[0033] The event nodes in the updated event context map that are associated with the event node of the event clicked by the target user are used as associated event nodes;

[0034] The update attribute information corresponding to the associated event node is determined, and the update event content is extracted from the update attribute information of the associated event node and pushed to the target user end.

[0035] In the second aspect, an event context mining and pushing system based on directional tracking is provided, including a manuscript collection unit, an information extraction unit, an information analysis unit, a graph construction unit, an information search unit, a graph update unit, a portrait construction unit, a context search unit and a content push unit, wherein:

[0036] A manuscript collection unit, used to collect event information manuscripts corresponding to target events clicked and / or followed by users;

[0037] An information extraction unit is used to extract the text content of the target event from the event information manuscript, and extract event attribute information from the text content using a preset large model based on the set first large model prompt word to obtain event basic information of the target event, wherein the event basic information includes event content, event type, event elements, event status, keywords, entity words, summary, and ID of users who have clicked and / or followed;

[0038] An information analysis unit, configured to perform deduplication and correlation analysis on the event basic information corresponding to each target event using a preset large model based on the set second large model prompt word, and determine the correlation relationship between the target events corresponding to each event basic information;

[0039] A graph construction unit is used to construct an event context graph in a graph database based on each target event, event basic information corresponding to each target event, and associations between each target event, with each target event as an event node, associations between target events as edges, and event basic information as event attribute information of corresponding event nodes;

[0040] An information search unit, used to search for update attribute information of corresponding events according to event attribute information corresponding to each event node in the event context map, and form an event update list based on the update attribute information of each event;

[0041] A graph updating unit, used to update the event attribute information corresponding to each event node in the event context graph using the event update list to obtain an updated event context graph;

[0042] A portrait construction unit, used to construct a user portrait of the target user based on the event clicking behavior of the target user and the content information of the clicked event, and determine the event of interest to the target user according to the user portrait of the target user;

[0043] A context search unit, used to determine the corresponding event node and its updated attribute information in the updated event context map according to the event of interest and / or the clicked event of the target user;

[0044] The content push unit is used to extract the update event content from the update attribute information of the corresponding event node, and push the update event content to the target user end.

[0045] Thirdly, we provide an event context mining and push system based on directional tracking, including:

[0046] A memory for storing instructions;

[0047] A processor is used to read the instructions stored in the memory and execute any one of the methods described in the first aspect according to the instructions.

[0048] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored on the computer-readable storage medium, and when the instructions are executed on a computer, the computer executes any one of the methods described in the first aspect. In addition, a computer program product is provided, and when the computer program product is executed on a computer, the computer executes any one of the methods described in the first aspect.

[0049] Beneficial effects:

[0050] 1. The present invention can accurately grasp the core content of the event through in-depth understanding of the target event information and comprehensive event information extraction. At the same time, combined with the construction and continuous updating of the event context map, it can clearly outline the development context of the event, so as to intuitively understand the details of the subsequent development of the event, avoiding the problems of information fragmentation and unclear context under the traditional association recommendation method.

[0051] 2. The present invention intelligently locates event-centered dynamic information sources, can regularly retrieve relevant information for each event, and screen out reliable and valuable update content based on filtering conditions, ensuring that the search information is closely related to the event and is timely, providing sufficient and accurate data support for the timely update of event context, and solving the problem that the prior art is difficult to accurately lock relevant information sources.

[0052] 3. The present invention performs event information deduplication and correlation relationship analysis based on an advanced big model, which can timely record events with status updates and specific update content, and accurately integrate them into the event context map, so that the event context map always keeps the latest status and reflects the real-time progress of the event, so that users can grasp the latest developments of the event at the first time.

[0053] 4. The present invention constructs a user's interest profile through collaborative filtering, and accurately pushes updated content of related events based on the user's personalized interest profile, as well as pushes updated content of events and related events that the user has clicked on. This personalized push mechanism can provide users with more accurate and personalized information recommendation services, meet users' needs for in-depth tracking of specific events and acquisition of relevant information, greatly improves user experience, ensures that users will not miss important progress in events of interest to them, and overcomes the defect of the prior art that there is a lack of personalized targeted push mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 Schematic diagram of the steps of the method in Example 1 of the present invention;

[0056] Figure 2 This is a schematic diagram of the structure of the system in Example 2 of the present invention;

[0057] Figure 3 Schematic diagram of the system structure in Example 3 of the present invention. DETAILED DESCRIPTION

[0058] It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention. The specific structures and functional details disclosed herein are only used to describe the exemplary embodiments of the present invention. However, the present invention can be embodied in many alternative forms, and it should not be understood that the present invention is limited to the embodiments set forth herein.

[0059] It should be understood that unless otherwise clearly specified and limited, the corresponding terms should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection, an indirect connection through an intermediate medium, or the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the embodiments can be understood according to specific circumstances.

[0060] In the following description, certain details are provided to facilitate a complete understanding of the example embodiments. However, it will be appreciated by those of ordinary skill in the art that the example embodiments may be implemented without these certain details. For example, devices may be shown in block diagrams to avoid obscuring the examples with unnecessary details. In other embodiments, well-known processes, structures, and techniques may not be shown in unnecessary detail to avoid obscuring the embodiments.

[0061] Embodiment 1:

[0062] This embodiment provides an event context mining and pushing method based on directional tracking, which can be applied to corresponding intelligent media platforms, such as Figure 1 As shown, the method comprises the following steps:

[0063] S1. Collect event information manuscripts corresponding to target events clicked and / or followed by users.

[0064] In specific implementation, the platform can obtain event information manuscripts corresponding to target events clicked and / or followed by users on the platform through big data collection methods. The event information manuscripts may contain corresponding text content, audio content, video content, picture content, etc.

[0065] S2. Extract the text content of the target event from the event information manuscript, and extract event attribute information from the text content using a preset large model based on the set first large model prompt word to obtain the event basic information of the target event, wherein the event basic information includes event content, event type, event elements, event status, keywords, entity words, summary, and ID of users who have clicked and / or followed.

[0066] In specific implementation, the initial text content, audio content and video content can be extracted from the event information manuscript; then the corresponding speech recognition tools, such as PaddleSpeech, etc., are used to perform language recognition on the audio content and video content to obtain the corresponding supplementary text content; the initial text content is combined with the supplementary text content to obtain the text content of the target event. After obtaining the text content of the target event, the preset large model, such as Qwen-14B, etc., can be used based on the set first large model prompt word (i.e. prompt) to extract event attribute information from the text content to obtain the event basic information of the target event. After the event attribute information is extracted through the large model, the extraction results can be verified to ensure that the extracted event attribute information conforms to the predefined pattern. The event basic information may include event content, event type, event elements, event status, keywords, entity words, summaries, and IDs of users who have clicked and / or followed, etc.

[0067] S3. Based on the set second largest model prompt word, a preset large model is used to perform deduplication and correlation analysis on the event basic information corresponding to each target event, and determine the correlation relationship between the target events corresponding to each event basic information.

[0068] In specific implementation, based on the set second largest model prompt word (i.e. prompt), a preset large model, such as Qwen-14B, can be used to deduplicate and perform correlation analysis on the basic event information corresponding to each target event, so as to explore the temporal sequence, causal relationship, parallel relationship and other correlation relationships between the basic event information and the target events.

[0069] S4. Based on each target event, the basic event information corresponding to each target event, and the association relationship between each target event, an event context map is constructed in a graph database with each target event as an event node, the association relationship between the target events as an edge, and the basic event information as the event attribute information of the corresponding event node.

[0070] In specific implementation, after determining the correlation between each target event, an event context map can be constructed in a graph database (such as a neo4j graph database) based on each target event, the event basic information corresponding to each target event, and the correlation between each target event, with each target event as an event node, the correlation between the target events as an edge, and the event basic information as the event attribute information of the corresponding event node. After the event context map is initialized, when there are new events, the related events in the map can be retrieved, and based on the same method as above, the relationship between the new events and the events in the event context map can be identified, thereby updating the event context map and constructing a continuously updated event context map.

[0071] S5. According to the event attribute information corresponding to each event node in the event context map, search for the update attribute information of the corresponding event, and form an event update list based on the update attribute information of each event.

[0072] In specific implementation, the platform can realize intelligent positioning of dynamic information sources centered on events based on the constructed event context map. The process includes: extracting keywords, entity words and abstracts from the event attribute information corresponding to each event node, and using keywords, entity words and abstracts to search for relevant information on information search engines (such as Quark Search) to obtain corresponding event search text content. Conditional filtering is performed on the search text content based on the event attribute information of each target event. For example, based on the event attribute information of each target event, the content in the trusted source list issued by the authoritative network management agency that is later than the existing event status of the target event is filtered to obtain the filtered event search text content. Then, based on the set first large model prompt word, a preset large model is used to extract event attribute information from the filtered event search text content to obtain the corresponding new event basic information; based on the set second large model prompt word, a preset large model is used to deduplicate and perform correlation analysis on each new event basic information to determine the correlation relationship between the events corresponding to each new event basic information. Then, based on the correlation relationship between the events corresponding to each newly added event basic information, the target event corresponding to each newly added event basic information is determined, and each newly added event basic information is used as the updated attribute information of the corresponding target event. Each updated attribute information and the corresponding target event are entered into the template table in pairs to obtain an event update list.

[0073] S6. Use the event update list to update the event attribute information corresponding to each event node in the event context map to obtain an updated event context map.

[0074] In specific implementation, after obtaining the event update list, the event attribute information corresponding to each event node in the event context map can be updated using the event update list to obtain an updated event context map.

[0075] S7. Based on the event clicking behavior of the target user and the content information of the clicked event, a user profile of the target user is constructed, and the event of interest to the target user is determined according to the user profile of the target user.

[0076] In specific implementation, based on the target user's event click behavior and the content information of the clicked event, the collaborative filtering method can be used to determine the target user's similar users and the events that the similar users have clicked and / or paid attention to. Then, the events that the similar users have clicked and / or paid attention to are regarded as the target user's events of interest, and the target user's user profile is constructed based on the target user's events of interest. When information is subsequently pushed to the target user, the target user's events of interest can be determined based on the target user's user profile, so as to push targeted event information.

[0077] S8. Determine the corresponding event node and its updated attribute information in the updated event context map according to the event of interest and / or the clicked event of the target user.

[0078] In specific implementation, the corresponding event nodes can be determined in the updated event context map according to the event of interest and / or the clicked event of the target user, and the updated attribute information of the corresponding event can be extracted.

[0079] S9. Extract the update event content from the update attribute information of the corresponding event node, and push the update event content to the target user end.

[0080] In specific implementation, the updated event content can be extracted from the updated attribute information of the corresponding event node, and the updated event content can be pushed to the target user end. At the same time, the event node in the updated event context map that has a correlation relationship with the event node of the event clicked by the target user can also be used as the associated event node, and then the updated attribute information corresponding to the associated event node is determined, and the updated event content is extracted from the updated attribute information of the associated event node and pushed to the target user end.

[0081] The method of this embodiment integrates and extracts the text content of the event manuscript, extracts and verifies the event information comprehensively based on the big model, realizes the deep understanding of the target event, can accurately grasp the core content of the event, and at the same time, combined with the construction and continuous updating of the event context map, can effectively reflect the development context and changes of the event, so as to intuitively understand the details of the subsequent development of the event, and avoid the problem of information fragmentation and unclear context under the traditional association recommendation method. The intelligent positioning of the dynamic information source centered on the event can retrieve relevant information for each event at a regular time, and filter out reliable and valuable updated content based on the filtering conditions, ensuring that the search information is closely related to the event and has timeliness, providing sufficient and accurate data support for the timely update of the event context, and solving the problem that the existing technology is difficult to accurately lock the relevant information source. Through collaborative filtering, the user's interest portrait is constructed, based on the user's click behavior and the relevant information of the clicked event, the accurate grasp of the user's personalized interests is realized, and then the personalized push of the update status and specific content of the relevant event is carried out according to the user's interest portrait, meeting the user's attention to the progress of the event of interest and improving the user experience.

[0082] Embodiment 2:

[0083] This embodiment provides an event context mining and push system based on directional tracking, such as Figure 2 As shown, it includes a manuscript collection unit, an information extraction unit, an information analysis unit, a graph construction unit, an information search unit, a graph update unit, a portrait construction unit, a context search unit and a content push unit, wherein:

[0084] A manuscript collection unit, used to collect event information manuscripts corresponding to target events clicked and / or followed by users;

[0085] An information extraction unit is used to extract the text content of the target event from the event information manuscript, and extract event attribute information from the text content using a preset large model based on the set first large model prompt word to obtain event basic information of the target event, wherein the event basic information includes event content, event type, event elements, event status, keywords, entity words, summary, and ID of users who have clicked and / or followed;

[0086] An information analysis unit, configured to perform deduplication and correlation analysis on the event basic information corresponding to each target event using a preset large model based on the set second large model prompt word, and determine the correlation relationship between the target events corresponding to each event basic information;

[0087] A graph construction unit is used to construct an event context graph in a graph database based on each target event, event basic information corresponding to each target event, and associations between each target event, with each target event as an event node, associations between target events as edges, and event basic information as event attribute information of corresponding event nodes;

[0088] An information search unit, used to search for update attribute information of corresponding events according to event attribute information corresponding to each event node in the event context map, and form an event update list based on the update attribute information of each event;

[0089] A graph updating unit, used to update the event attribute information corresponding to each event node in the event context graph using the event update list to obtain an updated event context graph;

[0090] A portrait construction unit, used to construct a user portrait of the target user based on the event clicking behavior of the target user and the content information of the clicked event, and determine the event of interest to the target user according to the user portrait of the target user;

[0091] A context search unit, used to determine the corresponding event node and its updated attribute information in the updated event context map according to the event of interest and / or the clicked event of the target user;

[0092] The content push unit is used to extract the update event content from the update attribute information of the corresponding event node, and push the update event content to the target user end.

[0093] Embodiment 3:

[0094] This embodiment provides an event context mining and push system based on directional tracking, such as Figure 3 As shown, at the hardware level, it includes:

[0095] Data interface, used to establish data connection between the processor and the user end;

[0096] A memory for storing instructions;

[0097] The processor is used to read the instructions stored in the memory and execute the event context mining and pushing method based on directional tracing in Example 1 according to the instructions.

[0098] Optionally, the system further includes an internal bus, through which the processor, the memory and the data interface can be interconnected, and the internal bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0099] The memory may include, but is not limited to, random access memory (RAM), read only memory (ROM), flash memory, first input first output (FIFO) and / or first in last out (FILO) memory, etc. The processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may 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.

[0100] Embodiment 4:

[0101] This embodiment provides a computer-readable storage medium, on which instructions are stored, and when the instructions are executed on a computer, the computer executes the event context mining and pushing method based on directional tracing in Embodiment 1. The computer-readable storage medium refers to a carrier for storing data, which may include but is not limited to a floppy disk, an optical disk, a hard disk, a flash memory, a USB flash drive, and / or a memory stick, etc., and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0102] This embodiment also provides a computer program product, which, when running on a computer, executes the event context mining and pushing method based on directional tracing in Embodiment 1. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device.

[0103] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An event context mining and pushing method based on directional tracking, characterized in that: include: For the pre-built event context map, extract the event attribute information corresponding to each event node; Based on the event attribute information corresponding to each event node, search for the update attribute information of the corresponding event, and form an event update list based on the update attribute information of each event; Using the event update list, the event attribute information corresponding to each event node in the event context map is updated to obtain an updated event context map; Based on the event click behavior of the target user and the content information of the clicked event, a user profile of the target user is constructed, and the event of interest to the target user is determined according to the user profile of the target user; Determine the corresponding event node and its updated attribute information in the updated event context map according to the event of interest and / or the clicked event of the target user; The update event content is extracted from the update attribute information of the corresponding event node, and the update event content is pushed to the target user end.

2. The event context mining and pushing method based on directional tracking according to claim 1 is characterized in that: Before extracting event attribute information corresponding to each event node in the pre-constructed event context map, the method further includes: Collect event information manuscripts corresponding to target events clicked and / or followed by users; Extracting the text content of the target event from the event information manuscript, and extracting event attribute information from the text content using a preset large model based on the set first large model prompt word, to obtain event basic information of the target event, wherein the event basic information includes event content, event type, event elements, event status, keywords, entity words, summary, and ID of users who have clicked and / or followed; Based on the set second largest model prompt word, the preset large model is used to remove duplication and conduct correlation analysis on the event basic information corresponding to each target event, so as to determine the correlation relationship between the target events corresponding to each event basic information; According to each target event, the basic event information corresponding to each target event and the association relationship between each target event, an event context map is constructed in the graph database with each target event as an event node, the association relationship between the target events as an edge, and the basic event information as the event attribute information of the corresponding event node.

3. The event context mining and pushing method based on directional tracking according to claim 2 is characterized in that: The step of extracting the text content of the target event from the event information manuscript includes: Extracting initial text content, audio content and video content from event information manuscripts; Use corresponding speech recognition tools to perform language recognition on audio content and video content to obtain corresponding supplementary text content; The initial text content is combined with the supplementary text content to obtain the text content of the target event.

4. The event context mining and pushing method based on directional tracking according to claim 2 is characterized in that: The step of searching for updated attribute information of a corresponding event based on the event attribute information corresponding to each event node includes: Extract keywords, entity words and abstracts from the event attribute information corresponding to each event node, and use the keywords, entity words and abstracts to search for relevant information on the information search engine to obtain the corresponding event search text content; Conditionally filtering the search text content based on the event attribute information of each target event to obtain filtered event search text content; Based on the set first large model prompt word, the preset large model is used to extract event attribute information from the filtered event search text content to obtain the corresponding new event basic information; Based on the set second largest model prompt word, the preset large model is used to perform deduplication and correlation analysis on the basic information of each newly added event, and the correlation relationship between the events corresponding to the basic information of each newly added event is determined; The target event corresponding to each newly added event basic information is determined based on the association relationship between the events corresponding to each newly added event basic information, and each newly added event basic information is used as the updated attribute information of the corresponding target event.

5. The event context mining and pushing method based on directional tracking according to claim 4 is characterized in that: The event update list is formed based on the updated attribute information of each event, including: entering each updated attribute information and the corresponding target event into a template table in pairs to obtain the event update list.

6. The event context mining and pushing method based on directional tracking according to claim 1 is characterized in that: The step of constructing a user profile of the target user based on the event click behavior of the target user and the content information of the clicked event includes: Based on the target user's event click behavior and the content information of the clicked event, a collaborative filtering method is used to determine similar users of the target user and the events that the similar users have clicked and / or followed; The events that similar users have clicked and / or followed are regarded as the events of interest to the target user, and a user profile of the target user is constructed based on the events of interest to the target user.

7. The event context mining and pushing method based on directional tracking according to claim 1 is characterized in that: The method further comprises: The event nodes in the updated event context map that are associated with the event node of the event clicked by the target user are used as associated event nodes; The update attribute information corresponding to the associated event node is determined, and the update event content is extracted from the update attribute information of the associated event node and pushed to the target user end.

8. The event context mining and push system based on directional tracking is characterized by: It includes a manuscript collection unit, an information extraction unit, an information analysis unit, a graph construction unit, an information search unit, a graph update unit, a portrait construction unit, a context search unit and a content push unit, wherein: A manuscript collection unit, used to collect event information manuscripts corresponding to target events clicked and / or followed by users; An information extraction unit is used to extract the text content of the target event from the event information manuscript, and extract event attribute information from the text content using a preset large model based on the set first large model prompt word to obtain event basic information of the target event, wherein the event basic information includes event content, event type, event elements, event status, keywords, entity words, summary, and ID of users who have clicked and / or followed; An information analysis unit, configured to perform deduplication and correlation analysis on the event basic information corresponding to each target event using a preset large model based on the set second large model prompt word, and determine the correlation relationship between the target events corresponding to each event basic information; A graph construction unit is used to construct an event context graph in a graph database based on each target event, event basic information corresponding to each target event, and associations between each target event, with each target event as an event node, associations between target events as edges, and event basic information as event attribute information of corresponding event nodes; An information search unit, used to search for update attribute information of corresponding events according to event attribute information corresponding to each event node in the event context map, and form an event update list based on the update attribute information of each event; A graph updating unit, used to update the event attribute information corresponding to each event node in the event context graph using the event update list to obtain an updated event context graph; A portrait construction unit, used to construct a user portrait of the target user based on the event clicking behavior of the target user and the content information of the clicked event, and determine the event of interest to the target user according to the user portrait of the target user; A context search unit, used to determine the corresponding event node and its updated attribute information in the updated event context map according to the event of interest and / or the clicked event of the target user; The content push unit is used to extract the update event content from the update attribute information of the corresponding event node, and push the update event content to the target user end.

9. The event context mining and push system based on directional tracking is characterized by: include: A memory for storing instructions; A processor is used to read the instructions stored in the memory and execute the event context mining and pushing method based on directional tracing as described in any one of claims 1 to 7 according to the instructions.

10. A computer program product, characterized in that When the computer program product is run on a computer, the event context mining and pushing method based on directional tracing described in any one of claims 1 to 7 is executed.

Images