Hot event venation tracking method and device, computer equipment and storage medium

Abstract

The invention relates to a hotspot event venation tracking method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring an event message set of a target hotspot event; constructing a complete event message graph corresponding to the event message set; obtaining a minimum event message tree of the complete event message graph, splitting the minimum event message graph into a plurality of message clusters by removing a target edge to obtain a message cluster set, obtaining core event messages of each message cluster in the message cluster set, and extracting a semantic relationship between every two core event messages; the semantic relationship comprises a logic relationship and a time sequence; and determining serial words and sentences and a serial sequence according to the logical relationship and the time sequence among the core event messages, and connecting the core event messages according to the serial words and sentences and the serial sequence to obtain a development process venation diagram of the target hotspot event. By adopting the method, the complete development context of the hot event can be accurately presented, and powerful support is provided for public opinion monitoring, risk assessment and decision making.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a hot event context tracking method and device, computer equipment and storage medium. BACKGROUND

[0002] After a hot event occurs, its related information is massive and complex, and has a profound influence on the decision-making of enterprises and various organizations. The limitations of existing technologies in sorting out the context of hot events make the obtained context unable to accurately reflect the overall picture of the event, making it difficult to meet the comprehensive and accurate requirements of public opinion monitoring on event analysis, and there is an urgent need for more effective methods to sort out the development context of events to assist public opinion monitoring.

[0003] When a hot event occurs, the hot event will continue to develop and change within a certain time range, and a large number of event messages about the hot event will appear. The more complex and variable the development of the hot event is, the more event messages about the hot event there are, and the more difficult it is to clarify the hot event. The existing technology for clarifying the development context of the hot event is only to link the previously determined important event messages in chronological order. In fact, in the development and change of events, some previously unimportant event messages have played a key role in subsequent critical event messages, so ignoring such event messages when clarifying the development context of the hot event will cause the event development context to change abruptly. In addition, in many cases, event messages are not simply in chronological order, and directly linking them in chronological order will cause the relationship between event messages in the event development context to be chaotic. SUMMARY

[0004] Therefore, it is necessary to provide a hot event context tracking method, device, computer equipment and storage medium to solve the above technical problems.

[0005] A hot event context tracking method, the method comprising:

[0006] obtaining an event message set of a target hot event; the event message set comprises a plurality of event messages;

[0007] constructing a complete event message graph corresponding to the event message set; the complete event message graph comprises nodes and edges, wherein the nodes are event messages in the event message set, and the edges are feature distances between the nodes; the feature distance is calculated by the embedding features between two event messages;

[0008] obtaining a minimum event message tree of the complete event message graph, splitting the minimum event message graph into a plurality of message clusters by removing target edges to obtain a message cluster set, obtaining core event messages of each message cluster in the message cluster set, and extracting semantic relationships between every two core event messages; the semantic relationship includes a logical relationship and a time sequence;

[0009] According to the logical relationship and time sequence between the core event messages, concatenated sentences and concatenated sequences are determined, and the core event messages are connected according to the concatenated sentences and the concatenated sequences to obtain the development process context diagram of the target hot event.

[0010] In an embodiment, further comprising: determining a target edge in the minimum event message tree, removing the target edge, and placing the two message clusters obtained by splitting into the message cluster set; if there is a to-be-split message cluster in the message cluster set, determining and removing a target edge in the to-be-split message cluster to obtain two new message clusters, and updating the message cluster set; the to-be-split message cluster is a message cluster with a node quantity greater than a preset quantity; iteratively updating the message cluster set until there is no to-be-split message cluster in the message cluster set, stopping iteration, and outputting the current message cluster set.

[0011] In an embodiment, further comprising: traversing each edge in the minimum event message tree, removing the current edge to split the minimum event message tree to obtain two message clusters; and determining the current edge as the target edge of the minimum event message tree when the sum of the inverses of the total number of edges in the two message clusters is greater than the inverse of the total number of edges in the minimum event message tree.

[0012] In an embodiment, further comprising: traversing each edge in the to-be-split message cluster, removing the current edge to split the to-be-split message cluster to obtain two new message clusters; and determining the current edge as the target edge of the to-be-split message cluster when the sum of the inverses of the total number of edges in the two new message clusters is greater than the inverse of the total number of edges in the to-be-split message cluster.

[0013] In an embodiment, further comprising: obtaining the average value of the embedding features of the event messages corresponding to each message cluster in the message cluster set to obtain the average vector of each message cluster; and obtaining the core event message of each message cluster according to the event message corresponding to the embedding feature most similar to the average vector in the message cluster.

[0014] In an embodiment, further comprising: traversing the semantic relationships between the core event messages, determining whether there is a logical relationship in the semantic relationships, if there is, determining concatenated sentences between the core event messages according to the logical relationship, and connecting the corresponding core event messages according to the concatenated sentences; determining whether there is a time sequence in the semantic relationships, if there is, determining the concatenated sequences between the core event messages according to the time sequence, and connecting the corresponding core event messages according to the concatenated sequences; until the end of the traversal, obtaining the development process context diagram of the target hot event.

[0015] In an embodiment, further comprising: obtaining the development process context diagram of other hot events; and determining the development relationship between the target hot event and the other hot events according to the development process context diagram of the target hot event and the development process context diagram of the other hot events.

[0016] A hotspot event context tracking device, the device comprising:

[0017] An event message acquisition module for acquiring an event message set of a target hotspot event; the event message set comprising a plurality of event messages;

[0018] A complete event message graph construction module for constructing a complete event message graph corresponding to the event message set; the complete event message graph comprising nodes and edges, wherein the nodes are event messages in the event message set, and the edges are feature distances between the nodes; the feature distances are calculated through embedded features between two event messages;

[0019] A semantic relationship extraction module for acquiring a minimum event message tree of the complete event message graph, splitting the minimum event message graph into a plurality of message clusters by removing target edges to obtain a message cluster set, acquiring core event messages of each message cluster in the message cluster set, and extracting semantic relationships between every two core event messages; the semantic relationships comprising logical relationships and time sequences;

[0020] A result output module for determining concatenated sentences and concatenated sequences according to the logical relationships and time sequences between the core event messages, connecting the core event messages according to the concatenated sentences and concatenated sequences, and obtaining a development process context graph of the target hotspot event.

[0021] A computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0022] Acquiring an event message set of a target hotspot event; the event message set comprising a plurality of event messages;

[0023] Constructing a complete event message graph corresponding to the event message set; the complete event message graph comprising nodes and edges, wherein the nodes are event messages in the event message set, and the edges are feature distances between the nodes; the feature distances are calculated through embedded features between two event messages;

[0024] Acquiring a minimum event message tree of the complete event message graph, splitting the minimum event message graph into a plurality of message clusters by removing target edges to obtain a message cluster set, acquiring core event messages of each message cluster in the message cluster set, and extracting semantic relationships between every two core event messages; the semantic relationships comprising logical relationships and time sequences;

[0025] Determining concatenated sentences and concatenated sequences according to the logical relationships and time sequences between the core event messages, connecting the core event messages according to the concatenated sentences and concatenated sequences, and obtaining a development process context graph of the target hotspot event.

[0026] A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the following steps:

[0027] Obtaining an event message set of a target hot event; the event message set comprises a plurality of event messages;

[0028] Constructing a complete event message graph corresponding to the event message set; the complete event message graph comprises nodes and edges, wherein the nodes are event messages in the event message set, and the edges are feature distances between the nodes; the feature distances are calculated by embedding features between two event messages;

[0029] Obtaining a minimum event message tree of the complete event message graph, splitting the minimum event message graph into a plurality of message clusters by removing target edges to obtain a message cluster set, obtaining core event messages of each message cluster in the message cluster set, and extracting semantic relationships between every two core event messages; the semantic relationships comprise logical relationships and time sequences;

[0030] Determining concatenated sentences and concatenated sequences according to the logical relationships and the time sequences between the core event messages, connecting the core event messages according to the concatenated sentences and the concatenated sequences to obtain a development process context graph of the target hot event.

[0031] The hot event context tracking method, device, computer device and storage medium described above, by constructing a complete event message graph, realize comprehensive correlation analysis of each event message in a hot event, ensure the integrity of the event development context, calculate the feature distances between event messages based on embedding features, quantify the semantic similarity between event messages, and avoid context confusion caused by relying only on time sequence. Further, the minimum event message tree is obtained and the message clusters are split by removing target edges, the sub-events in the hot event with different development directions are reasonably divided, the event context is clearer, then the core event messages in each message cluster are extracted, the key nodes are screened out, the core context of event development is highlighted, information redundancy is reduced, the logical relationships and the time sequences between the core event messages are analyzed, the event development path is reasonably constructed, and it is ensured that the event context not only conforms to the time logic, but also reflects the cause-and-effect relationship between events. Finally, based on the logical relationships and the time sequences, concatenated sentences and concatenated sequences are generated, the event development context is expressed more naturally and smoothly, the readability and decision support capability of public opinion monitoring are improved. The embodiments of the present application can accurately present the complete development context of a hot event, and provide strong support for public opinion monitoring, risk assessment and decision making. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 It is a flowchart of the hot event context tracking method in one embodiment;

[0033] Figure 2 FIG. 1 is a structural block diagram of a hotspot event thread tracking device according to an embodiment;

[0034] Figure 3 FIG. 2 is an internal structural diagram of a computer device according to an embodiment. DETAILED DESCRIPTION

[0035] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0036] In one embodiment, as shown in FIG. 1, a hotspot event thread tracking method is provided, including the following steps: Figure 1

[0037] Step 102, obtaining an event message set of a target hotspot event.

[0038] The event messages of the target hotspot event can be from news, articles, blogs, comments, etc. The event message set includes multiple event messages.

[0039] Step 104, constructing a complete event message graph corresponding to the event message set.

[0040] The complete event message graph includes nodes and edges, wherein the nodes are the event messages in the event message set, and the edges are the feature distances between the nodes. The feature distance is calculated by the embedding features between two event messages. Specifically, the embedding features of each event message are extracted, and based on the embedding features of any two event messages, the feature distance between the two event messages is calculated. Each event message is taken as a node, and the feature distance between any two event messages is taken as the edge between the two event messages, to construct the complete event message graph corresponding to the event message set.

[0041] Step 106, obtaining a minimum event message tree of the complete event message graph, splitting the minimum event message graph into multiple message clusters by removing target edges to obtain a message cluster set, obtaining core event messages of each message cluster in the message cluster set, and extracting semantic relationships between every two core event messages.

[0042] ​The plurality of event messages are clustered to divide the plurality of event messages into a plurality of message clusters, and one message cluster can be the development of one aspect or one stage of the target hot event. For example, the target hot event is a large-scale public opinion event on a social platform, and thus a plurality of event messages are generated. Because the evolution of a general public opinion event has complexity and multi-factor driving, there can be multiple reasons and new reasons in the public opinion fermentation process, and different reasons can dominate different small events (one small event corresponds to one event message). According to the reasons for the initiation or spread of public opinion, the plurality of event messages can be divided into a plurality of message clusters. For example, cause factors, secondary propagation factors, emotional fermentation factors, and involved party response factors, core event messages of each message cluster are determined, and the core event message of one message cluster can be one or more.

[0043] Then, semantic relationships between each two core event messages are extracted, and the semantic relationships include logical relationships and time sequences.

[0044] In step 108, the concatenation words and the concatenation order are determined according to the logical relationships and the time sequences between the core event messages, the core event messages are connected according to the concatenation words and the concatenation order, and a development process context diagram of the target hot event is obtained.

[0045] According to the semantic relationships between any two core event messages, the core event messages are concatenated to obtain the development process context diagram of the target hot event.

[0046] Through the context diagram, it can be accurately tracked how the public opinion evolves from the initial event into a large-scale hot event, and which events play a role in the process. The core events that affect the trend of public opinion, such as specific statements, KOL (key opinion leader) driving, policy intervention, etc., are found to adjust the public opinion response strategy in time, and the change trend of public emotion with the development of the event can also be analyzed to predict whether it will further ferment or cool down. It is helpful to identify early signals that may trigger a public relations crisis and intervene in guiding public opinion in advance.

[0047] When recommending news, the development process context diagram of the target hot event corresponding to a piece of news can be recommended to a specific user, and through the recommended event context diagram, the user can more clearly see the cause, key nodes, and latest progress of the hot event to which the news belongs, avoiding focusing on fragmented information. It is helpful for enterprises to track public opinion hotspots related to themselves and identify potential crises or marketing opportunities in advance.

[0048] In the hot event context tracking method, the complete event message graph is constructed to realize comprehensive correlation analysis of each event message in the hot event, ensure the integrity of the event development context, calculate the feature distance between event messages based on embedded features, and quantify the semantic similarity between event messages to avoid context confusion caused by relying only on time sequence. Further, the minimum event message tree is obtained and the target edge is removed to split the message cluster, the sub-events in the hot event with different development directions are reasonably divided to make the event context clearer, then the core event messages in each message cluster are extracted to screen out the key nodes, highlight the core context of event development, reduce information redundancy, and reasonably construct the event development path by combining the logical relationship and time sequence analysis between the core event messages to ensure that the event context not only conforms to the time logic but also reflects the causal relationship between events. Finally, the series of words and the series order are generated based on the logical relationship and time sequence to make the expression of the event development context more natural and smooth, and improve the readability and decision support capability of public opinion monitoring. The embodiment of the present application can accurately present the complete development context of the hot event, and provide strong support for public opinion monitoring, risk assessment and decision making.

[0049] In one embodiment, obtaining the minimum event message tree of the complete event message graph includes: obtaining the minimum event message tree according to a connection mode with the smallest total sum of edges in the complete event message graph.

[0050] In this embodiment, the minimum event message tree is extracted from the complete event message graph, wherein the minimum event message tree is a connection mode containing all nodes in the complete event message graph, and the total sum of edges in the minimum event message tree is the smallest among all connection modes. The minimum event message tree is split into multiple message clusters by gradually removing the target edge in the minimum event message tree.

[0051] Specifically, the event message set includes 100 event messages, the 100 event messages are taken as 100 nodes, and the feature distance between any two event messages is taken as the edge between the two event messages to construct the complete event message graph corresponding to the event message set. Next, the redundant edges in the complete event message graph are removed as long as all nodes in the complete event message graph are included, and multiple connection modes can be obtained. In the multiple connection modes, the connection mode with the smallest total sum of edges is selected as the minimum event message tree. Finally, the target edge in the minimum event message tree is gradually removed to split the minimum event message tree into multiple message clusters. The feature distance d between two event messages can be calculated according to the following formula:

[0052]

[0053] Wherein, xi is an embedded feature of an event message, yi is an embedded feature of another event message, i is from 1 to n, and the embedded features of the two event messages are n-dimensional feature vectors.

[0054] In one embodiment, the minimum event message tree is split into a plurality of message clusters by removing a target edge to obtain a message cluster set, including: determining a target edge in the minimum event message tree, removing the target edge, and putting the two message clusters obtained by splitting into the message cluster set; if there is a message cluster to be split in the message cluster set, determining and removing a target edge in the message cluster to be split, obtaining two new message clusters, and updating the message cluster set; the message cluster to be split is a message cluster with a node number greater than a preset number; iteratively updating the message cluster set until there is no message cluster to be split in the message cluster set, stopping iteration, and outputting the current message cluster set.

[0055] In this embodiment, because the sum of the edges in the minimum event message tree is the smallest among all connection modes, the minimum event message tree is already in the simplest connection mode, so the minimum event message tree has no redundant edges, and the minimum event message tree can be split into two message clusters as long as one edge is removed. A target edge in the minimum event message tree is determined, and the target edge in the minimum event message tree is removed, at this time the minimum event message tree is split into two message clusters. The two message clusters obtained by splitting are put into the message cluster set, and the following loop is executed: the target edge of the message cluster in the message cluster set is continuously determined, and the target edge of the message cluster in the message cluster set is removed, until the number of nodes in all message clusters in the message cluster set is less than the preset number, and the loop is ended. It should be noted that in one loop round, if the number of nodes in a message cluster is already less than the preset number, the message cluster will not participate in splitting in subsequent loop rounds.

[0056] In one embodiment, determining the target edge in the minimum event message tree includes: traversing each edge in the minimum event message tree, removing the current edge to split the minimum event message tree to obtain two message clusters; and determining the current edge as the target edge of the minimum event message tree when the sum of the reciprocals of the sums of the edges in the two message clusters is greater than the reciprocal of the sum of the edges in the minimum event message tree.

[0057] In this embodiment, the sum of the edges in the minimum event message tree is determined, denoted as the first sum. Each edge in the minimum event message tree is traversed to determine the sum of the edges in the two message clusters obtained by splitting according to the edge, denoted as the second sum and the third sum, respectively. When the sum of the reciprocals of the second sum and the third sum is greater than the reciprocal of the first sum, the edge is determined as the target edge. If the sum of the reciprocals of the second sum and the third sum is not greater than the reciprocal of the first sum, it is determined whether the next edge is the target edge.

[0058] In one embodiment, determining the target edge in the message cluster to be split comprises: traversing each edge in the message cluster to be split, removing the current edge to split the message cluster to be split to obtain two new message clusters; and determining the current edge as the target edge of the message cluster to be split when the sum of the inverses of the total number of edges in the two new message clusters is greater than the inverse of the total number of edges in the message cluster to be split.

[0059] In the present embodiment, the total number of edges in the message cluster to be split is determined as a fourth total number. For any edge in the message cluster to be split, the total number of edges in the two new message clusters obtained by splitting according to the edge is determined as a fifth total number and a sixth total number, respectively; and the edge is determined as the target edge when the sum of the inverses of the fifth total number and the sixth total number is greater than the inverse of the fourth total number. If the sum of the inverses of the fifth total number and the sixth total number is not greater than the inverse of the fourth total number, it is determined whether the next edge is the target edge.

[0060] The target edge can be determined by judging whether the following formula is satisfied:

[0061]

[0062] wherein a and b are the total number of edges in the two new message clusters obtained by splitting, and c is the total number of edges in the message cluster to be split or the minimum event message tree.

[0063] In one embodiment, obtaining the core event message of each message cluster in the message cluster set comprises: obtaining the average value of the embedding features of the event messages corresponding to each message cluster in the message cluster set to obtain the average vector of each message cluster; and obtaining the core event message of each message cluster according to the event message corresponding to the embedding feature most similar to the average vector in each message cluster.

[0064] In the present embodiment, the embedding features of all event messages in each message cluster are extracted, the average value of all embedding features in each message cluster is calculated to obtain the average vector of each message cluster, and the event message corresponding to the embedding feature most similar to the average vector of each message cluster in the message cluster is taken as the core event message of the message cluster. The corresponding similarity W between the average vector A of each message cluster and the embedding feature B of any event message in the message cluster can be calculated according to the following formula: AB

[0065]

[0066] wherein X i represents the average vector A, Y i represents the embedding feature B, i takes a value from 1 to n, and A and B are both n-dimensional feature vectors.

[0067] ​The large language model can also synthesize multiple event messages in each message cluster to obtain a core event message of the message cluster.

[0068] In one embodiment, the series connection words and the series connection order are determined according to the logical relationship and the time sequence between the core event messages, the core event messages are connected according to the series connection words and the series connection order, and the development process context diagram of the target hot event is obtained. The method comprises the following steps: traversing the semantic relationship between the core event messages, determining whether there is a logical relationship in the semantic relationship, if there is, determining the series connection words between the core event messages according to the logical relationship, and connecting the corresponding core event messages according to the series connection words; determining whether there is a time sequence in the semantic relationship, if there is, determining the series connection order between the core event messages according to the time sequence, and connecting the corresponding core event messages according to the series connection order; until the end of the traversal, the development process context diagram of the target hot event is obtained.

[0069] In this embodiment, for any two core events, the two core event messages are subjected to word segmentation processing, stop word removal processing, part-of-speech tagging processing and named entity recognition processing to obtain respective processing results of the two core event messages, and the semantic relationship between the two core event messages is determined according to the respective processing results of the two core event messages. The semantic relationship between the two core event messages includes whether there is a logical relationship and a time sequence between the two core event messages, and the logical relationship includes a cause-and-effect relationship, a transition relationship, a parallel relationship and a progressive relationship.

[0070] Word segmentation processing is to divide a core event message into words or phrases. Stop word removal is to remove common words in a core event message, such as “of”, “is”, “in” and the like, which do not contribute much to the semantics. Part-of-speech tagging is to identify the part of speech (noun, verb, adjective, etc.) of each word in a core event message. Named entity recognition is to identify characters, places and other nouns in a core event message.

[0071] The large language model can be used to determine the semantic relationship between the two core event messages according to the respective processing results of the two core event messages.

[0072] The development process context diagram of the target hot event is obtained by concatenating each core event message according to the semantic relationship between any two core event messages, including: if the semantic relationship between any two core event messages is that there is no logical relationship and no time sequence, the two core event messages are not concatenated; if the semantic relationship between any two core event messages is that there is only a logical relationship, concatenation words and sentences between the two core event messages are determined according to the logical relationship, and the two core event messages are concatenated based on the concatenation words and sentences between the two core event messages; if the semantic relationship between any two core event messages is that there is only a time sequence, a concatenation order between the two core event messages is determined according to the time sequence, and the two core event messages are concatenated based on the concatenation order between the two core event messages; if the semantic relationship between any two core event messages is that there is a logical relationship and a time sequence, concatenation words and sentences between the two core event messages are determined according to the logical relationship, a concatenation order between the two core event messages is determined according to the time sequence, and the two core event messages are concatenated based on the concatenation words and sentences and the concatenation order between the two core event messages.

[0073] The two core event messages are concatenated according to the semantic relationship between the two core event messages. For example, if the contents of the two core event messages are parallel, the two core event messages can be connected together using concatenation words and sentences such as “and”, “and”, etc.; if the contents of the two core event messages are transitional, the two core event messages can be connected together using concatenation words and sentences such as “but”, “but”, etc.; if the contents of the two core event messages have a cause-and-effect relationship, the two core event messages can be connected together using concatenation words and sentences such as “because”, “because”, etc.; if the contents of the two core event messages are progressive, the two core event messages can be connected together using concatenation words and sentences such as “further”, etc.

[0074] If the semantic relationship between any two core event messages is that there is a time sequence, the core event message in the time sequence before concatenates the core event message in the time sequence after.

[0075] In one embodiment, the method further comprises: obtaining a development process context diagram of another hot event; and determining a development contact between the target hot event and the other hot event according to the development process context diagram of the target hot event and the development process context diagram of the other hot event.

[0076] In the embodiment, the same or similar event messages in the development process context diagram of the target hot event and the development process context diagram of other hot events can be determined as the connection between the target hot event and other hot events in development. When a hot event occurs, if it is found that it has the same key node or development mode as similar events in history, the possible evolution direction can be predicted, and the response plan can be prepared in advance to avoid the spread of crisis.

[0077] In addition, by analyzing the public opinion transmission context of multiple events, it can also be found out which information sources have strong transmission power, and the core factors affecting the development of public opinion can be found out according to historical public opinion events, so as to develop accurate guidance measures to prevent public opinion from getting out of control.

[0078] It should be understood that, although Figure 1 The steps in the flowchart of the method are displayed in sequence according to the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps in the method can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily sequential, but can be executed in rotation or alternation with other steps or at least part of the sub-steps or stages of other steps.

[0079] In one embodiment, as shown in Figure 2 A hot event context tracking device is provided, comprising:

[0080] The event message acquisition module 202 is configured to acquire a set of event messages of a target hot event; the set of event messages comprises a plurality of event messages;

[0081] The complete event message graph construction module 204 is configured to construct a complete event message graph corresponding to the set of event messages; the complete event message graph comprises nodes and edges, wherein the nodes are the event messages in the set of event messages, and the edges are feature distances between the nodes; the feature distance is calculated by the embedded features between two event messages;

[0082] The semantic relationship extraction module 206 is configured to acquire a minimum event message tree of the complete event message graph, split the minimum event message graph into a plurality of message clusters by removing target edges to obtain a set of message clusters, acquire core event messages of each message cluster in the set of message clusters, and extract semantic relationships between every two core event messages; the semantic relationships include logical relationships and time sequences;

[0083] The result output module 208 is configured to determine the concatenated sentence and the concatenated order according to the logical relationship and the time sequence among the core event messages, connect the core event messages according to the concatenated sentence and the concatenated order, and obtain the development process context diagram of the target hot event.

[0084] In one embodiment, the processor is further configured to determine a target edge in the minimum event message tree, remove the target edge, and place the two message clusters obtained by splitting into the message cluster set; determine and remove a target edge in a to-be-split message cluster in the message cluster set to obtain two new message clusters and update the message cluster set; the to-be-split message cluster is a message cluster with a node quantity greater than a preset quantity; iteratively update the message cluster set until there is no to-be-split message cluster in the message cluster set, stop iteration, and output the current message cluster set.

[0085] In one embodiment, the processor is further configured to traverse each edge in the minimum event message tree, remove the current edge to split the minimum event message tree to obtain two message clusters; and determine the current edge as the target edge of the minimum event message tree when the sum of the inverses of the total sum of the edges in the two message clusters is greater than the inverse of the total sum of the edges in the minimum event message tree.

[0086] In one embodiment, the processor is further configured to traverse each edge in the to-be-split message cluster, remove the current edge to split the to-be-split message cluster to obtain two new message clusters; and determine the current edge as the target edge of the to-be-split message cluster when the sum of the inverses of the total sum of the edges in the two new message clusters is greater than the inverse of the total sum of the edges in the to-be-split message cluster.

[0087] In one embodiment, the processor is further configured to obtain the average value of the embedding features of the event messages corresponding to each message cluster in the message cluster set to obtain the average vector of each message cluster; and obtain the core event message of each message cluster according to the event message corresponding to the embedding feature most similar to the average vector in the message cluster.

[0088] In one embodiment, the processor is further configured to traverse the semantic relationship among the core event messages, determine whether there is a logical relationship in the semantic relationship, and if so, determine the concatenated sentence between the core event messages according to the logical relationship and connect the corresponding core event messages according to the concatenated sentence; determine whether there is a time sequence in the semantic relationship, and if so, determine the concatenated order between the core event messages according to the time sequence and connect the corresponding core event messages according to the concatenated order; until the end of the traversal, obtain the development process context diagram of the target hot event.

[0089] In one embodiment, the processor is further configured to obtain the development process context diagram of other hot events; and determine the development relationship between the target hot event and the other hot events according to the development process context diagram of the target hot event and the development process context diagram of the other hot events.

[0090] The specific limitation of the hotspot event thread tracking device can refer to the limitation of the hotspot event thread tracking method in the foregoing, and will not be described here. Each module in the above hotspot event thread tracking device can be realized by software, hardware and a combination thereof in whole or in part. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to call and execute the operations corresponding to each module by the processor.

[0091] In one embodiment, a computer device, which can be a terminal, has an internal structure diagram as shown in Figure 3 The computer device includes a processor, a memory, a network interface, a display screen and an input device connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with external terminals through network connection. The computer program is executed by the processor to implement a hotspot event thread tracking method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0092] Those skilled in the art can understand that Figure 3 The structure shown in the foregoing is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the diagram, or combine certain components, or have a different component arrangement.

[0093] In one embodiment, a computer device is provided, which includes a memory and a processor. The memory stores a computer program. The processor executes the computer program to implement the steps of the method in the above embodiments.

[0094] In one embodiment, a computer readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement the steps of the method in the above embodiments.

[0095] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, databases, or other media in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0096] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0097] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A hot event context tracking method, characterized in that: The method comprises: Obtain an event message set of a target hot event; the event message set includes multiple event messages; Constructing a complete event message graph corresponding to the event message set; the complete event message graph includes nodes and edges, wherein a node is an event message in the event message set and an edge is a characteristic distance between nodes; the characteristic distance is calculated by using the embedded features between two event messages; Obtain a minimum event message tree of the complete event message graph, split the minimum event message graph into multiple message clusters by removing target edges to obtain a message cluster set, obtain core event messages of each message cluster in the message cluster set, and extract the semantic relationship between each two core event messages; the semantic relationship includes a logical relationship and a time sequence; The concatenated words and sequence are determined according to the logical relationship and time sequence between the core event messages, and the core event messages are connected according to the concatenated words and sequence to obtain a development process diagram of the target hot event.

2. The method according to claim 1, characterized in that The minimum event message graph is split into multiple message clusters by removing the target edge, and the obtained message cluster set includes: Determine the target edge in the minimum event message tree, remove the target edge, and put the two message clusters obtained by splitting into the message cluster set; If there is a message cluster to be split in the message cluster set, determine and remove the target edge in the message cluster to be split, obtain two new message clusters, and update the message cluster set; the message cluster to be split is a message cluster with a node number greater than a preset number; Iterate and update the message cluster set until there is no message cluster to be split in the message cluster set, then stop iterating and output the current message cluster set.

3. The method according to claim 2, characterized in that Determine the target edge in the minimum event message tree including: Traverse each edge in the minimum event message tree, remove the current edge to split the minimum event message tree, and obtain two message clusters; When the sum of the reciprocals of the sums of the edges in the two message clusters is greater than the reciprocal of the sum of the edges in the minimum event message tree, the current edge is determined as the target edge of the minimum event message tree.

4. The method according to claim 2, characterized in that Determining the target edge in the message cluster to be split includes: Traverse each edge in the message cluster to be split, remove the current edge to split the message cluster to be split, and obtain two new message clusters; When the sum of the reciprocals of the sum of the edges in the two new message clusters is greater than the reciprocal of the sum of the edges in the message cluster to be split, the current edge is determined as the target edge of the message cluster to be split.

5. The method according to claim 1, wherein The core event messages of each message cluster in the message cluster set are obtained: Obtain the average value of the embedded features of the event messages corresponding to each message cluster in the message cluster set to obtain the average vector of each message cluster; According to the event message corresponding to the embedded feature that is most similar to the average vector in each message cluster, the core event message of each message cluster is obtained.

6. The method according to claim 1, characterized in that Determining the concatenated words and sentences and the concatenation order based on the logical relationship and time sequence between the core event messages, and connecting the core event messages based on the concatenated words and sentences and the concatenation order to obtain the development process diagram of the target hot event includes: Traversing the semantic relationships between the core event messages, determining whether there is a logical relationship in the semantic relationship, and if so, determining the concatenated phrases between the core event messages based on the logical relationship, and connecting the corresponding core event messages based on the concatenated phrases; Determine whether there is a time sequence in the semantic relationship; if so, determine a concatenation order between the core event messages according to the time sequence, and connect the corresponding core event messages according to the concatenation order; When the traversal is completed, a development process diagram of the target hot event is obtained.

7. The method according to claim 1, characterized in that The method further comprises: Obtain the development process of other hot events; Determine the development relationship between the target hot event and other hot events based on the development process pulse diagram of the target hot event and the development process pulse diagram of other hot events.

8. A hot event context tracking device, characterized in that: The device comprises: An event message acquisition module is used to acquire an event message set of a target hot event; the event message set includes multiple event messages; A complete event message graph construction module is used to construct a complete event message graph corresponding to the event message set; the complete event message graph includes nodes and edges, wherein a node is an event message in the event message set and an edge is a characteristic distance between nodes; the characteristic distance is calculated by using the embedded features between two event messages; A semantic relationship extraction module is configured to obtain a minimum event message tree of the complete event message graph, split the minimum event message graph into multiple message clusters by removing target edges to obtain a message cluster set, obtain the core event messages of each message cluster in the message cluster set, and extract the semantic relationship between each two core event messages; the semantic relationship includes a logical relationship and a time sequence; The result output module is used to determine the serial words and serial order according to the logical relationship and time sequence between the core event messages, connect the core event messages according to the serial words and serial order, and obtain the development process context diagram of the target hot event.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images