Event context generation method, device, equipment and storage medium

Abstract

The present invention discloses an event context generation method, device, equipment and storage medium. The method comprises: obtaining event data to be combed, clustering the event data to be combed to obtain multiple clusters; determining keyword information in the clusters, and determining weight information using the keyword information and a semantic similarity algorithm; generating an event subgraph using the weight information, and outputting event context information using the event subgraph, wherein the event context information at least includes an event context tree diagram. The technical solution of the embodiment of the present invention accurately clusters the event data to be combed, integrates the keyword information of each event in the cluster into the event feature word weight and the event association relationship weight, enhances the scientific nature of the association analysis, improves the accuracy of the analysis, accurately judges the association between news, and then uses the weight information to output the event context tree diagram, so as to intuitively and accurately present the event evolution process.

Description

Technical Field

[0001] The present invention relates to the field of information technology application technology, and in particular to an event context generation method, device, equipment and storage medium. Background Art

[0002] In the wave of the information age, news media, as an important window for the public to obtain information, has seen an explosive growth in the amount of information. However, how to quickly and accurately extract the context of events from this massive amount of news data has become an urgent problem to be solved in the field of information processing.

[0003] Traditional event analysis methods are limited by manual operations, are not only inefficient, but also easily affected by subjective factors, and are difficult to adapt to the rapid changes in the information age. In recent years, with the rapid development of machine learning and natural language processing technologies, news analysis methods have ushered in new breakthroughs. Using text mining technology to extract key information, these methods have improved the automation level of event analysis to a certain extent.

[0004] However, existing automated event context analysis methods are inadequate when dealing with multi-dimensional information, complex event relationships, and large-scale data sets. Traditional event context analysis usually presents results in the form of text lists and cannot provide intuitive and accurate event context information. Summary of the invention

[0005] The present invention provides an event context generation method, device, equipment and storage medium to solve the problem that event context generation effect is poor and unintuitive.

[0006] In a first aspect, the present invention provides an event context generation method, comprising:

[0007] Acquire event data to be sorted out, and cluster the event data to be sorted out to obtain a plurality of clusters, wherein the clusters contain data of a plurality of events to be sorted out;

[0008] Determine the keyword information in the cluster, and determine the weight information using the keyword information and a semantic similarity algorithm, wherein the keyword information includes keywords and corresponding keyword weights, and the weight information includes event feature word weights and event association relationship weights;

[0009] An event subgraph is generated using the weight information, and event context information is output using the event subgraph, wherein the event subgraph includes nodes corresponding to the event feature word weights and node edges corresponding to the event association relationship weights, and the event context information at least includes an event context tree diagram.

[0010] In a second aspect, the present invention provides an event context generating device, comprising:

[0011] A cluster generation module, used for acquiring event data to be sorted out, clustering the event data to be sorted out to obtain a plurality of clusters, wherein the clusters contain data of a plurality of events to be sorted out;

[0012] A weight information determination module, used to determine the keyword information in the cluster, and determine the weight information using the keyword information and a semantic similarity algorithm, wherein the keyword information includes keywords and corresponding keyword weights, and the weight information includes event feature word weights and event association relationship weights;

[0013] An event context information output module is used to generate an event subgraph using the weight information, and output event context information using the event subgraph, wherein the event subgraph includes nodes corresponding to the event feature word weights and node edges corresponding to the event association relationship weights, and the event context information at least includes an event context tree diagram.

[0014] In a third aspect, the present invention provides an electronic device, the electronic device comprising:

[0015] at least one processor;

[0016] and a memory communicatively coupled to the at least one processor;

[0017] The memory stores a computer program that can be executed by at least one processor, and the computer program is executed by at least one processor so that the at least one processor can execute the event context generation method of the first aspect mentioned above.

[0018] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer instructions, which are used to enable a processor to implement the event context generation method of the first aspect when executed.

[0019] The event context generation scheme provided by the present invention obtains event data to be sorted, clusters the event data to be sorted to obtain multiple clusters, wherein the clusters contain data of multiple events to be sorted, determines keyword information in the clusters, and uses the keyword information and a semantic similarity algorithm to determine weight information, wherein the keyword information includes keywords and corresponding keyword weights, and the weight information includes event feature word weights and event association relationship weights, generates an event subgraph using the weight information, and outputs event context information using the event subgraph, wherein the event subgraph includes nodes corresponding to the event feature word weights and node edges corresponding to the event association relationship weights, and the event context information at least includes an event context tree diagram. By adopting the above technical solution, the event data to be sorted out is accurately clustered, and the keyword information in the clustering cluster and the semantic similarity algorithm are used to determine the weight information, so as to integrate the keyword information of each event into the event feature word weight and the event association relationship weight. This not only enhances the scientific nature of the correlation analysis, but also improves the accuracy of the analysis, accurately judges the correlation between news, and then uses the weight information to output the event context tree diagram, which can intuitively and accurately present the event evolution process.

[0020] It should be understood that the content described in this section is not intended to identify the key or important features of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 creative work.

[0022] Figure 1 is a flow chart of an event context generation method provided according to Embodiment 1 of the present invention;

[0023] Figure 2 is a flow chart of an event context generation method provided according to Embodiment 2 of the present invention;

[0024] Figure 3 An event context tree diagram provided according to the second embodiment of the present invention;

[0025] Figure 4 is an event context axis diagram provided according to the second embodiment of the present invention;

[0026] Figure 5 is an event flow graph provided according to the second embodiment of the present invention;

[0027] Figure 6 is a schematic diagram of the structure of an event context generating device provided according to Embodiment 3 of the present invention;

[0028] Figure 7 It is a schematic diagram of the structure of an electronic device provided according to Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In the description of the present invention, unless otherwise specified, "multiple" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are an "or" relationship. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] Embodiment 1

[0032] Figure 1 A flowchart of an event context generating method is provided for the first embodiment of the present invention. This embodiment is applicable to the case of generating event context. The method can be executed by an event context generating device. The event context generating device can be implemented in the form of hardware and / or software. The event context generating device can be configured in an electronic device. The electronic device can be composed of two or more physical entities or one physical entity.

[0033] like Figure 1As shown, the event context generation method provided in the first embodiment of the present invention specifically includes the following steps:

[0034] S101 , obtaining event data to be sorted out, and clustering the event data to be sorted out to obtain a plurality of clusters, wherein the clusters contain data of a plurality of events to be sorted out.

[0035] In this embodiment, a preset clustering algorithm, such as a Single-Pass algorithm, can be used to cluster the acquired event data to be sorted, thereby obtaining multiple clusters. The event data to be sorted can be news data to be sorted, and the news data contains data of multiple events to be sorted.

[0036] S102, determining keyword information in the cluster, and determining weight information using the keyword information and a semantic similarity algorithm, wherein the keyword information includes keywords and corresponding keyword weights, and the weight information includes event feature word weights and event association relationship weights.

[0037] In this embodiment, for each cluster, a preset information extraction model, such as a KeyBERT model, can be used to extract keywords from the cluster, and then a preset weighting algorithm, such as a TF-IDF algorithm, can be used to weight the keywords, thereby obtaining keyword information. Then, by performing a preset operation on the keyword weights corresponding to the keywords, such as weighted summation, the data event feature word weights of each event to be combed are obtained, and the semantic similarity algorithm is used to determine the event association relationship weights of the data of each event to be combed in the cluster.

[0038] S103. Generate an event subgraph using the weight information, and output event context information using the event subgraph, wherein the event subgraph includes nodes corresponding to the event feature word weights and node edges corresponding to the event association relationship weights, and the event context information at least includes an event context tree diagram.

[0039] In the present embodiment, the event feature word weight can be abstracted as a node, and the event association relationship weight can be abstracted as the edge of the node, so as to generate an event subgraph comprising the above-mentioned edge and the above-mentioned node. For example, the event feature word weight is 1 (abstracted as node 1) and the event feature word weight is 2 (abstracted as node 2) in the data of event a, the event feature word weight is 3 (abstracted as node 3) in the data of event b, and the event association relationship weight of event a and event b is 4 (abstracted as edge), then in the generated event subgraph, node 1 and node 2 respectively have edges with node 3, and the attribute value of the node is the event feature word weight. According to the event occurrence date and the release time of the event data to be combed, the integrated event subgraph can generate an event context tree diagram. Wherein, the event context tree of the event context tree diagram can include nodes and edges in the event subgraph.

[0040] The event context generation method provided by the embodiment of the present invention obtains event data to be sorted, clusters the event data to be sorted to obtain multiple clusters, wherein the clusters contain data of multiple events to be sorted, determines keyword information in the clusters, and uses the keyword information and a semantic similarity algorithm to determine weight information, wherein the keyword information includes keywords and corresponding keyword weights, and the weight information includes event feature word weights and event association relationship weights, generates an event subgraph using the weight information, and outputs event context information using the event subgraph, wherein the event subgraph includes nodes corresponding to the event feature word weights and node edges corresponding to the event association relationship weights, and the event context information at least includes an event context tree diagram. The technical solution of the embodiment of the present invention accurately clusters the event data to be sorted, and uses the keyword information in the cluster and the semantic similarity algorithm to determine the weight information, so as to integrate the keyword information of each event into the event feature word weight and the event association relationship weight. This not only enhances the scientific nature of the correlation analysis, but also improves the accuracy of the analysis, accurately judges the correlation between news, and then uses the weight information to output the event context tree diagram, which can intuitively and accurately present the event evolution process.

[0041] Optionally, before clustering the event data to be sorted to obtain a plurality of clustering clusters, the process includes: filtering the event data to be sorted according to the sensitivity rating of the source of the event data to be sorted to obtain filtered data; wherein, clustering the event data to be sorted to obtain a plurality of clustering clusters includes: vectorizing the filtered data using a vector model to obtain event vectors to be sorted; performing a first clustering of the event vectors to be sorted using a clustering algorithm HDBSCAN to obtain an initial vector cluster, and performing a second clustering of the initial vector cluster using a clustering algorithm Single-Pass to obtain a plurality of target vector clusters; and for each target vector cluster, determining the event data to be sorted corresponding to the current target vector cluster as a clustering cluster.

[0042] Specifically, the generation step of clustering clusters may include the following four steps:

[0043] Step 1: Call the news filtering module to filter the event data to be sorted. The specific steps include:

[0044] ①: Filter the event data to be sorted that lacks information such as title, content, and release time;

[0045] ②: De-duplicate the event data to be sorted after removing missing information;

[0046] ③: Obtain the source website list of the event data to be sorted after deduplication;

[0047] ④: In the source website list, the preset specific sensitive websites are graded, and the deduplicated event data to be sorted whose grades exceed the preset values ​​are deleted to obtain filtered data.

[0048] Step 2: Call the vector model BGE to vectorize the filtered data to obtain the event vector to be sorted.

[0049] Step 3: Based on the spatial distance between the event vectors to be sorted, the clustering algorithm HDBSCAN is used to perform the first clustering on the event vectors to be sorted, and the abnormal event vectors to be sorted are eliminated to obtain the initial vector cluster.

[0050] Step 4: Use the clustering algorithm Single-Pass to perform a second clustering on the initial vector cluster to obtain multiple target vector clusters, and for each target vector cluster, determine the event data to be sorted in the current target vector cluster as a clustering cluster. The clustering cluster can be understood as a data set of events to be sorted in different dimensions.

[0051] Optionally, the method of determining the keyword information in the clustering cluster includes: using the KeyBERT model and the NER model to determine the target words from the clustering cluster, and filtering the target words in the target words to generate a keyword group, wherein the target words include content words and initial keywords, and the target words include content words with overlapping elements and shorter content words or initial keywords in the initial keywords; segmenting the clustering cluster to obtain a segmentation result, and using the TF-IDF algorithm to determine the keyword information in the clustering cluster according to the segmentation result and the keyword group associated with the segmentation result.

[0052] Specifically, the process of determining the keyword information in each cluster includes the following six steps:

[0053] 1) Call the keyword extraction model (i.e., the Chinese-based KeyBERT model), determine the initial keywords and their weights from the clusters, and take the top 10 initial keywords based on the ranking of the weights.

[0054] 2) Call the entity extraction model (i.e., the NER model fine-tuned based on Roberta+GlobalPointer) to extract content words such as time, place, person, organization, and position from the clusters.

[0055] 3) Summarize the results of 2) and 3) to obtain an initial keyword group.

[0056] 4) Filter the initial keyword group, filter the target words in the target words, that is, retain the content words with overlapping elements in the initial keyword group and the longer content words or initial keywords in the initial keywords to generate keyword groups. For example, if the content word is ab and the initial keyword is abc, abc is retained. The overlapping elements can be overlapping characters, words, and phrases.

[0057] 5) Segment the current cluster to obtain the segmentation result.

[0058] 6) The word segmentation result and the keyword group associated with the word segmentation result (ie, the keyword group in the current cluster) are input into the TF-IDF algorithm to obtain the keyword information of the current cluster.

[0059] Embodiment 2

[0060] Figure 2 This is a flow chart of a method for generating an event context provided in the second embodiment of the present invention. The technical solution of the embodiment of the present invention is further optimized on the basis of the above-mentioned optional technical solutions, and a specific method for generating an event context is provided.

[0061] Optionally, the method of determining weight information by using the keyword information and the semantic similarity algorithm includes: for each cluster, combining the events to be sorted in the current cluster in pairs to obtain event pairs to be sorted; determining event feature word weights by using the keyword information, and determining similarity weights of the event pairs to be sorted by using the semantic similarity algorithm; and determining event association weights by using the similarity weights. The advantage of such a setting is that the event feature word weights and event association weights are accurately determined.

[0062] Optionally, the output of event context information using the event subgraph includes: using the PageRank algorithm to update the event feature word weights of the nodes in the event subgraph to obtain the important relevance weights, and normalizing and filtering the important relevance weights to obtain the target subgraph; aggregating the target subgraph according to the event date, and sorting the event data to be sorted on the same date in the aggregation result according to the release time to obtain the sorting result; outputting the event context tree diagram according to the dependency probability of the nodes of the target subgraph in the sorting result. The advantage of such a setting is that it combines the relevance between events, combines the event feature word weights and the semantic weights between events with the PageRank algorithm for sorting, thereby more accurately judging the relevance between events, making the output event context tree diagram more accurate.

[0063] Optionally, the event context information also includes an event context axis diagram and an event flow diagram; after generating the event context tree diagram according to the dependency probability of the target subgraph node in the sorting result, it also includes: using the hierarchical analysis method to determine the multiple dimensional weights of the event data to be sorted in the clustering cluster, and performing a multivariate linear regression operation on the multiple dimensional weights and the actual data corresponding to each dimensional weight to obtain the event attention; for the event data to be sorted with the maximum daily event attention in the clustering cluster, respectively determine the earliest released event data, and input the earliest released event data into a preset large language model to obtain target information and corresponding event progress labels, wherein the target information includes the event data to be sorted of the same event progress; using the target information and the corresponding event progress label to output the event context axis diagram, and output the daily event flow diagram according to the event attention and the clustering cluster. The advantage of this setting is that it introduces the hierarchical analysis method and multivariate linear regression algorithm to determine the attention of events, more comprehensively evaluates the importance of events, provides readers with a more accurate reference, and uses the small sample learning ability of the large language model to automatically judge the progress of events, greatly reducing the need for human participation and improving the degree of automation of event context analysis.

[0064] like Figure 2 As shown, a method for generating an event context provided by Embodiment 2 of the present invention specifically includes the following steps:

[0065] S201. Acquire event data to be sorted, and filter the event data to be sorted according to the sensitivity rating of the source of the event data to be sorted to obtain filtered data; vectorize the filtered data using a vector model to obtain an event vector to be sorted; perform a first clustering on the event vector to be sorted using a clustering algorithm HDBSCAN to obtain an initial vector cluster, and perform a second clustering on the initial vector cluster using a clustering algorithm Single-Pass to obtain multiple target vector clusters; for each target vector cluster, determine the event data to be sorted corresponding to the current target vector cluster as a clustering cluster.

[0066] S202, using the KeyBERT model and the NER model, determining the target words from the clusters, and filtering the target words in the target words to generate a keyword group, wherein the target words include content words and initial keywords, and the target words include content words with overlapping elements and shorter content words or initial keywords in the initial keywords; performing word segmentation on the clusters to obtain word segmentation results, and using the TF-IDF algorithm to determine the keyword information in the clusters according to the word segmentation results and the keyword group associated with the word segmentation results.

[0067] S203 . For each of the clusters, combine the events to be combed in the current cluster in pairs to obtain pairs of events to be combed.

[0068] Specifically, the data of multiple events to be sorted out in the clusters can be combined in pairs. The events to be sorted out in the clusters can be distinguished according to the sources of the event data to be sorted out.

[0069] S204: Determine the event feature word weights using the keyword information, and determine the similarity weights of the event pairs to be sorted using a semantic similarity algorithm.

[0070] Exemplarily, the sum of all weight values ​​in the keyword information of the cluster can be determined as the initial NewRank value, ie, the event feature word weight. The similarity weights of the data of the two events to be combed in the event pair to be combed are determined using a preset semantic similarity algorithm.

[0071] S205: Determine the event association relationship weight using the similarity weight.

[0072] Specifically, the event association relationship weights of the two events to be combed in the event pair to be combed may be determined by setting mathematical operations on the similarity weights of the event pair to be combed.

[0073] Furthermore, the use of the similarity weight to determine the event association relationship weight includes: weighted summing up the keyword weights corresponding to the overlapping keywords in the event pairs to be sorted out to obtain an intersection weight value; and weighted summing up the intersection weight value and the similarity weight to obtain the event association relationship weight.

[0074] Specifically, for each pair of events to be sorted, the keyword weights corresponding to the overlapping keywords contained in the data of the two events to be sorted in the current pair of events to be sorted can be weighted and summed to obtain the intersection weight value, and then the intersection weight value and the similarity weight can be weighted and summed to obtain the association relationship weight of the two events to be sorted in the current pair of events to be sorted.

[0075] S206, generating an event subgraph using the weight information; updating the event feature word weights of the nodes in the event subgraph using the PageRank algorithm to obtain important relevance weights, and normalizing and filtering the important relevance weights to obtain a target subgraph.

[0076] Specifically, the PageRank algorithm can be used to update the event feature word weights of the nodes in the event subgraph, and the updated event feature word weights are the target NewRank values, that is, the important relevance weights. The important relevance weights can be standardized by Z-score transformation. The standardized important relevance weights will obey the normal distribution, and then be filtered by the 3Sigma principle to update the nodes in the event subgraph and obtain the target subgraph. The advantage of this setting is that the normal distribution and the 3sigma principle are introduced to eliminate abnormal event data, ensure the rationality of event data filtering, and effectively reduce the impact of event data with lower importance on the output event context.

[0077] S207 , aggregating the target subgraph according to the event date, and sorting the event data to be sorted on the same date in the aggregation result according to the release time to obtain a sorting result.

[0078] Specifically, the target subgraphs can be aggregated according to the event date to obtain a subgraph set, i.e., an aggregation result, wherein each subgraph set can contain multiple target subgraphs with the same event date. Then, for each aggregation result, the event data to be sorted with the same date in the current aggregation result are sorted according to the release time to obtain a sorting result.

[0079] S208: Output an event context tree diagram according to the dependency probability of the nodes of the target subgraph in the sorting result.

[0080] Specifically, the method for determining the event context tree diagram includes the following three steps:

[0081] 1) In the target subgraph with the earliest date in the aggregation result, the node corresponding to the largest important relevance weight NewRank is used as the root node Tree of the event context tree root (that is, the parent node of the next node), and determine the remaining nodes as child nodes of the root node:

[0082] Tree root =argmaxNewRank n

[0083] Among them, n is the number of nodes in the target subgraph, and argmax represents the node corresponding to the largest important relevance weight NewRank.

[0084] 2) Determine the target subgraph for the next date based on the sorting results. Determine the dependency probability between the target subgraph and its parent node, determine the node with the maximum dependency probability as the child node of the parent node (that is, the parent node of the next node), and determine the remaining nodes as the child nodes of the parent node:

[0085]

[0086] Among them, p(x|y) is the dependent probability function.

[0087] 3) Repeat step 2) to obtain the tree corresponding to the current cluster, and obtain the event context tree diagram by integrating these trees.

[0088] The advantage of this setting is that the above steps determine the optimal event context for the day, and finally construct an event context tree diagram, which can present the event evolution process more intuitively.

[0089] For example, Figure 3 is an event context tree diagram. Figure 3 The trees in the green line, the purple line, and the orange line are the trees corresponding to different clusters.

[0090] S209, using the hierarchical analysis method to determine the weights of multiple dimensions of the event data to be sorted out in the cluster, and performing a multivariate linear regression operation on the multiple dimension weights and the actual data corresponding to each dimension weight to obtain the event attention.

[0091] Specifically, the process of using the hierarchical analysis method to determine the weights of multiple dimensions of the event data to be sorted out in the cluster includes the following five steps:

[0092] ① Establish a hierarchical model;

[0093] ②Construct the judgment matrix of indicator layers in each dimension;

[0094] ③ Calculate the weight of indicators in each dimension;

[0095] ④Consistency test;

[0096] ⑤ Obtain the final weights of each dimension.

[0097] For example, the event attention Y newAtt Methods of determination include:

[0098] Y newAtt =W 网站等级 X 网站等级 +W 点赞量 X 点赞量 +W 转发量 X 转发量 +W 评论量 X 评论量 +W 浏览量 X 浏览量

[0099] Among them, the dimension weights include: website level weight W 网站等级 、Like weight W 点赞量 , forwarding weight W 转发量 , Comment weight W 评论量 And the pageview weight W 浏览量 .X 网站等级 is the actual level of the website, W 点赞量 is the actual number of likes, W 转发量 is the actual forwarding amount, W 评论量 is the actual forwarding amount, W 浏览量 The actual number of page views.

[0100] S210: for the event data to be sorted out with the maximum daily event attention in the cluster, respectively determine the earliest released event data, and input the earliest released event data into a preset large language model to obtain target information and corresponding event progress labels.

[0101] The target information includes the event data to be sorted out regarding the progress of the same event.

[0102] Specifically, for the event data to be sorted out with the largest event attention in the clusters every day, the earliest released event data are determined respectively. The earliest released event data are input into the preset large language model in pairs, and the small sample learning ability of the preset large language model is used to determine whether the events corresponding to the earliest released event data are the same progress. If so, the earliest released event data is labeled, such as important progress, and the data and its label are output, that is, the target information and the corresponding event progress label are output.

[0103] S211. Output an event context axis diagram using the target information and the corresponding event progress labels, and output a daily event flow diagram based on the event attention and the clustering clusters.

[0104] The event context information also includes an event context axis diagram and an event flow diagram.

[0105] Specifically, the event context axis diagram may include refined information of target information and corresponding event progress labels, as well as refined information of other daily event data with the highest event attention to be sorted out. Figure 4 It is an event context axis diagram. Figure 4 "Important progress" is the event progress label. The data in the box it refers to is the refined information of the target information, and the data in the remaining boxes are the refined information of the other daily maximum event attention events to be sorted out. The cumulative value of the daily event attention and the sum of the total number of events can be determined as the daily event news attention. Use the date as the horizontal axis and the daily event news attention as the vertical axis to draw an event flow curve. Obtain the event focus topic in the preset curve peak area from the cluster cluster, and display the topic in the event flow curve to obtain the daily event flow graph. Figure 5 is an event flow graph, Figure 5 The horizontal axis is the date, the vertical axis is the daily event news attention, and the data in the box is the topic of event attention.

[0106] The event context generation method provided by the embodiment of the present invention integrates keyword extraction, content word recognition, vectorization, LLM deep understanding and graph analysis technology, realizing the specialization, automation and efficiency of event context analysis. This method can not only accurately construct event trees, event axes and event flows, but also comprehensively reveal the multiple paths, main clues and heat trends of event development, significantly improve the efficiency and accuracy of event analysis, and improve the comprehensiveness and depth of analysis. This method enables readers to quickly understand the development context and heat fluctuations of events by drawing event context tree diagrams, event context axis diagrams and event flow diagrams, providing strong support for relevant decision-making.

[0107] Embodiment 3

[0108] Figure 6 This is a schematic diagram of the structure of an event context generation device provided in Embodiment 3 of the present invention. Figure 6 As shown, the device includes: a cluster generation module 301, a weight information determination module 302 and an event context information output module 303, wherein:

[0109] A cluster generation module, used for acquiring event data to be sorted out, clustering the event data to be sorted out to obtain a plurality of clusters, wherein the clusters contain data of a plurality of events to be sorted out;

[0110] A weight information determination module, used to determine the keyword information in the cluster, and determine the weight information using the keyword information and a semantic similarity algorithm, wherein the keyword information includes keywords and corresponding keyword weights, and the weight information includes event feature word weights and event association relationship weights;

[0111] An event context information output module is used to generate an event subgraph using the weight information, and output event context information using the event subgraph, wherein the event subgraph includes nodes corresponding to the event feature word weights and node edges corresponding to the event association relationship weights, and the event context information at least includes an event context tree diagram.

[0112] The event context generating device provided by the embodiment of the present invention accurately clusters the event data to be sorted, and uses the keyword information in the clustering cluster and the semantic similarity algorithm to determine the weight information, so as to integrate the keyword information of each event into the event feature word weight and the event association relationship weight. This not only enhances the scientific nature of the correlation analysis, but also improves the accuracy of the analysis, accurately judges the correlation between news, and then uses the weight information to output the event context tree diagram, which can intuitively and accurately present the event evolution process.

[0113] Optionally, the method of determining the keyword information in the clustering cluster includes: using the KeyBERT model and the NER model to determine the target words from the clustering cluster, and filtering the target words in the target words to generate a keyword group, wherein the target words include content words and initial keywords, and the target words include content words with overlapping elements and shorter content words or initial keywords in the initial keywords; segmenting the clustering cluster to obtain a segmentation result, and using the TF-IDF algorithm to determine the keyword information in the clustering cluster according to the segmentation result and the keyword group associated with the segmentation result.

[0114] Optionally, the weight information determination module includes:

[0115] An event pair determination unit, for each of the clusters, combines the events to be combed in the current cluster in pairs to obtain event pairs to be combed;

[0116] A similarity weight determination unit, used to determine the event feature word weight using the keyword information, and determine the similarity weight of the event pair to be sorted using a semantic similarity algorithm;

[0117] The association relationship weight determination unit is used to determine the event association relationship weight using the similarity weight.

[0118] Optionally, using the similarity weight to determine the event association relationship weight includes: weighted summing up the keyword weights corresponding to the overlapping keywords in the event pairs to be sorted out to obtain an intersection weight value; and weighted summing up the intersection weight value and the similarity weight to obtain the event association relationship weight.

[0119] Optionally, the event context information output module includes:

[0120] A target subgraph determination unit, configured to update the event feature word weights of the nodes in the event subgraph using the PageRank algorithm to obtain important relevance weights, and to perform standardization and filtering on the important relevance weights to obtain a target subgraph;

[0121] An event sorting unit, used to aggregate the target subgraph according to the event date, and sort the event data to be sorted on the same date in the aggregation result according to the release time, so as to obtain a sorting result;

[0122] An event context tree diagram output unit is used to output an event context tree diagram according to the dependency probability of the nodes of the target subgraph in the sorting result.

[0123] Optionally, the event context information also includes an event context axis diagram and an event flow diagram.

[0124] Optionally, the device further comprises:

[0125] An event attention determination module is used to determine the weights of multiple dimensions of the event data to be sorted in the cluster clusters by using a hierarchical analysis method after generating an event context tree diagram according to the dependency probability of the target subgraph node in the sorting result, and perform a multivariate linear regression operation on the multiple dimensional weights and the actual data corresponding to each dimensional weight to obtain the event attention;

[0126] A model processing module is used to determine the earliest published event data for the event data to be sorted out with the largest daily event attention in the cluster, and input the earliest published event data into a preset large language model to obtain target information and corresponding event progress labels, wherein the target information includes the event data to be sorted out of the same event progress;

[0127] The image output module is used to output the event context axis diagram using the target information and the corresponding event progress label, and output the daily event flow diagram according to the event attention and the clustering cluster.

[0128] Optionally, the device further comprises:

[0129] The data filtering module is used to filter the event data to be sorted out according to the sensitivity rating of the source of the event data to be sorted out to obtain filtered data before clustering the event data to be sorted out to obtain multiple clusters.

[0130] Optionally, the cluster generation module includes:

[0131] A vectorization unit, used for vectorizing the filtered data using a vector model to obtain an event vector to be sorted;

[0132] A target vector cluster generating unit, configured to perform a first clustering on the event vectors to be sorted using a clustering algorithm HDBSCAN to obtain an initial vector cluster, and perform a second clustering on the initial vector cluster using a clustering algorithm Single-Pass to obtain a plurality of target vector clusters;

[0133] The target vector cluster generating unit is used to determine the to-be-sorted event data corresponding to the current target vector cluster as a clustering cluster for each of the target vector clusters.

[0134] The event context generation device provided in the embodiment of the present invention can execute the event context generation method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0135] Embodiment 4

[0136] Figure 7 A schematic diagram of the structure of an electronic device 40 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0137] like Figure 7As shown, the electronic device 40 includes at least one processor 41, and a memory connected to the at least one processor 41, such as a read-only memory (ROM) 42, a random access memory (RAM) 43, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 41 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 42 or the computer program loaded from the storage unit 48 to the random access memory (RAM) 43. In the RAM 43, various programs and data required for the operation of the electronic device 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other through a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0138] A number of components in the electronic device 40 are connected to the I / O interface 45, including: an input unit 46, such as a keyboard, a mouse, etc.; an output unit 47, such as various types of displays, speakers, etc.; a storage unit 48, such as a disk, an optical disk, etc.; and a communication unit 49, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0139] The processor 41 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 41 executes the various methods and processes described above, such as the event context generation method.

[0140] In some embodiments, the event context generation method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 48. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 40 via the ROM 42 and / or the communication unit 49. When the computer program is loaded into the RAM 43 and executed by the processor 41, one or more steps of the event context generation method described above may be performed. Alternatively, in other embodiments, the processor 41 may be configured to execute the event context generation method in any other appropriate manner (e.g., by means of firmware).

[0141] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0142] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0143] The computer device provided above can be used to execute the event context generation method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0144] Embodiment 5

[0145] In the context of the present invention, a computer-readable storage medium may be a tangible medium, wherein the computer-executable instructions, when executed by a computer processor, are used to perform an event context generation method, the method comprising:

[0146] Acquire event data to be sorted out, and cluster the event data to be sorted out to obtain a plurality of clusters, wherein the clusters contain data of a plurality of events to be sorted out;

[0147] Determine the keyword information in the cluster, and determine the weight information using the keyword information and a semantic similarity algorithm, wherein the keyword information includes keywords and corresponding keyword weights, and the weight information includes event feature word weights and event association relationship weights;

[0148] An event subgraph is generated using the weight information, and event context information is output using the event subgraph, wherein the event subgraph includes nodes corresponding to the event feature word weights and node edges corresponding to the event association relationship weights, and the event context information at least includes an event context tree diagram.

[0149] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by an instruction execution system, device or equipment or used with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0150] The computer device provided above can be used to execute the event context generation method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0151] It is worth noting that in the embodiment of the above-mentioned event context generating device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0152] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for generating an event context, characterized in that: include: Acquire event data to be sorted out, and cluster the event data to be sorted out to obtain a plurality of clusters; Determine the keyword information in the cluster, and determine the weight information using the keyword information and a semantic similarity algorithm, wherein the keyword information includes keywords and corresponding keyword weights, and the weight information includes event feature word weights and event association relationship weights; Generate an event subgraph using the weight information, and output event context information using the event subgraph, wherein the event subgraph includes nodes corresponding to the event feature word weights and node edges corresponding to the event association relationship weights, and the event context information at least includes an event context tree diagram; Wherein, the determining weight information by using the keyword information and the semantic similarity algorithm includes: For each of the clusters, the events to be combed in the current cluster are combined in pairs to obtain pairs of events to be combed; Determine the event feature word weights using the keyword information, and determine the similarity weights of the event pairs to be sorted using a semantic similarity algorithm; Determining the event association relationship weight using the similarity weight; The determining the event association relationship weight by using the similarity weight includes: Performing weighted summation on the keyword weights corresponding to the overlapping keywords in the event pairs to be sorted out to obtain an intersection weight value; The intersection weight value and the similarity weight are weighted and summed to obtain the event association relationship weight.

2. The method according to claim 1, characterized in that Before clustering the event data to be sorted out to obtain a plurality of clusters, the method includes: Filtering the event data to be sorted out according to the sensitivity rating of the source of the event data to be sorted out to obtain filtered data; The clustering of the event data to be sorted out to obtain a plurality of clusters includes: Vectorizing the filtered data using a vector model to obtain an event vector to be sorted; Performing a first clustering on the event vectors to be sorted using a clustering algorithm HDBSCAN to obtain an initial vector cluster, and performing a second clustering on the initial vector cluster using a clustering algorithm Single-Pass to obtain multiple target vector clusters; For each of the target vector clusters, the to-be-sorted event data corresponding to the current target vector cluster is determined as a clustering cluster.

3. The method according to claim 2, characterized in that The method of determining the keyword information in the cluster includes: Using the KeyBERT model and the NER model, a target word is determined from the cluster, and the target word in the target word is filtered to generate a keyword group, wherein the target word includes a content word and an initial keyword, and the target word includes a content word with overlapping elements and a shorter content word or initial keyword in the initial keyword; The clusters are segmented to obtain segmentation results, and the keyword information in the clusters is determined according to the segmentation results and keyword groups associated with the segmentation results using a TF-IDF algorithm.

4. The method according to claim 1, characterized in that: The step of outputting event context information by using the event subgraph includes: Using the PageRank algorithm to update the event feature word weights of the nodes in the event subgraph to obtain important relevance weights, and performing standardization and filtering on the important relevance weights to obtain a target subgraph; Aggregating the target subgraph according to the event date, and sorting the event data to be sorted on the same date in the aggregation result according to the release time to obtain a sorting result; Output an event context tree diagram according to the dependency probability of the nodes of the target subgraph in the sorting result.

5. The method according to claim 4, characterized in that The event context information also includes an event context axis diagram and an event flow diagram; after generating an event context tree diagram according to the dependency probability of the target subgraph node in the sorting result, it also includes: Determine multiple dimensional weights of the event data to be sorted out in the cluster using the hierarchical analysis method, and perform a multivariate linear regression operation on the multiple dimensional weights and the actual data corresponding to each dimensional weight to obtain the event attention; For the event data to be sorted out with the largest event attention in the clusters every day, respectively determine the earliest released event data, and input the earliest released event data into a preset large language model to obtain target information and corresponding event progress labels, wherein the target information includes the event data to be sorted out of the same event progress; The event context axis diagram is outputted using the target information and the corresponding event progress label, and the daily event flow diagram is outputted according to the event attention and the clustering cluster.

6. An event context generating device, characterized in that: include: A cluster generation module, used for acquiring event data to be sorted out, clustering the event data to be sorted out to obtain a plurality of clusters, wherein the clusters contain data of a plurality of events to be sorted out; A weight information determination module, used to determine the keyword information in the cluster, and determine the weight information using the keyword information and a semantic similarity algorithm, wherein the keyword information includes keywords and corresponding keyword weights, and the weight information includes event feature word weights and event association relationship weights; An event context information output module, used to generate an event subgraph using the weight information, and output event context information using the event subgraph, wherein the event subgraph includes nodes corresponding to the event feature word weights and node edges corresponding to the event association relationship weights, and the event context information at least includes an event context tree diagram; Wherein, the weight information determination module includes: An event pair determination unit, for each of the clusters, combines the events to be combed in the current cluster in pairs to obtain event pairs to be combed; A similarity weight determination unit, used to determine the event feature word weight using the keyword information, and determine the similarity weight of the event pair to be sorted using a semantic similarity algorithm; An association relationship weight determination unit, used to determine the event association relationship weight using the similarity weight; The method of using the similarity weight to determine the event association relationship weight includes: performing weighted summation on the keyword weights corresponding to the overlapping keywords in the event pairs to be sorted out to obtain an intersection weight value; and performing weighted summation on the intersection weight value and the similarity weight to obtain the event association relationship weight.

7. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the event context generation method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the event context generation method according to any one of claims 1 to 5 when executed.

Images