Event context generation method and device and computer readable medium

Abstract

The embodiment of the invention provides an event context generation method and device and a computer readable medium, and the method comprises the steps: firstly carrying out the extraction of summary information of a sample during cluster merging, and effectively reducing the interference of noise data on a merging process, moreover, the preset condition judgment during the event venation output contains the recognition of multiple events and the judgment of topic association, so that samples which contain multiple event topics and are irrelevant to topic events can be effectively excluded, the finally output event venation is more coherent and has higher logicality, and the accuracy of the event venation output is improved. Therefore, the accuracy of the finally output event venation is higher.

Description

Technical Field

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

[0002] With the development of Internet technology, online media is also making continuous progress. As a new communication medium, the Internet provides people with a convenient and fast way to obtain news and other media information. As a result, more and more users choose to obtain news and other media information on the Internet. Constructing event context for media data on the Internet has important application value.

[0003] The current event context generation scheme mainly relies on the clustering algorithm based on text similarity. The clustering algorithm is used to summarize and generalize all samples, return a certain number of clusters, and then filter and select the number of clusters by text similarity, and finally output the event context of the event according to the release time. However, in this type of scheme, due to the large noise of the received media data and different expressions, text similarity alone cannot distinguish the similarity between media data well. The interference of a large amount of noise data on the clustering algorithm will also cause serious logical problems in the output event context, which is seriously deviated from reality. Therefore, there is currently no event context generation scheme that can generate events with high accuracy. Summary of the invention

[0004] One object of the present application is to provide an event context generation method, device and computer-readable medium to solve the problem of insufficient generation accuracy in existing solutions.

[0005] To achieve the above purpose, an embodiment of the present application provides an event context generation method, the method comprising: Performing text clustering on text data related to the event topic to obtain multiple clusters, each cluster including at least one piece of text data; Calculating a first similarity between each cluster and the event subject, and filtering the clusters according to the first similarity; Extracting summary information of each cluster, calculating a second similarity between the summary information of each cluster, and merging the clusters according to the second similarity; sorting the clusters according to their release time; After confirming that the cluster meets the preset conditions, an event context about the event topic is generated, wherein the preset conditions include that the cluster center of the cluster contains a single event topic and that the cluster center of the cluster is associated with the event topic, and the event context includes event information sorted by release time.

[0006] Furthermore, before performing text clustering on media data related to the event theme to obtain multiple clusters, the following steps are also included: Perform data cleaning on media data related to event topics.

[0007] Furthermore, text clustering is performed on the media data related to the event theme to obtain multiple clusters, including: The SinglePass text clustering algorithm is used to perform text clustering on media data related to event topics and obtain multiple clusters.

[0008] Further, calculating a first similarity between each cluster and the event subject, and filtering the clusters according to the first similarity, includes: Calculating a first similarity between a cluster center of each cluster and an event topic, and comparing the first similarity with a first threshold; If the first similarity is less than a first threshold, deleting the cluster corresponding to the first similarity; If the first similarity is greater than or equal to a first threshold, the cluster corresponding to the first similarity is retained.

[0009] Furthermore, the summary information of each cluster is extracted, including: The cluster center of each cluster is input into the large model to obtain summary information of each cluster.

[0010] Further, calculating a second similarity between summary information of each cluster, and merging the clusters according to the second similarity; Calculating a second similarity between summary information of cluster centers of each cluster, and comparing the second similarity with a second threshold; If the second similarity is less than a second threshold, retaining two clusters corresponding to the second threshold; If the second similarity is greater than or equal to a second threshold, two clusters corresponding to the second threshold are merged.

[0011] Further, after confirming that the cluster meets the preset conditions, an event context about the event theme is generated, including: Determine whether each cluster meets the preset conditions; If the preset conditions are met, the cluster is retained, and an event context about the event topic is generated according to the retained cluster.

[0012] Furthermore, the method further comprises: If the preset condition is not met, the cluster is deleted.

[0013] Some embodiments of the present application also provide an event context generation device, wherein the device includes a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the aforementioned event context generation method.

[0014] Other embodiments of the present application further provide a computer-readable medium having computer program instructions stored thereon, wherein the computer program instructions can be executed by a processor to implement the event context generation method.

[0015] Compared with the prior art, in an event context generation scheme provided by an embodiment of the present application, text clustering is first performed on text data related to the event theme to obtain multiple clusters, and then the first similarity between each cluster and the event subject is calculated, and the clusters are filtered according to the first similarity. After extracting the summary information of each cluster, the second similarity between the summary information of each cluster is calculated, and the clusters are merged according to the second similarity, and the clusters are sorted according to their release time, and after confirming that the clusters meet the preset conditions, the event context about the event theme is generated. Since the scheme first extracts the summary information of the sample when merging the clusters, the interference of noise data on the merging process can be effectively reduced, and since the preset condition judgment when outputting the event context includes the identification of multiple events and the judgment of the theme association, it can effectively exclude samples containing multiple event themes and irrelevant to the theme event, so that the final output event context is more coherent and has stronger logic, so the accuracy of the final output event context has higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 A schematic diagram of a flow chart of an event context generation method provided in an embodiment of the present application; Figure 2 A schematic diagram of the process of extracting summary information and merging clusters in an embodiment of the present application; Figure 3 A schematic diagram of a multi-event recognition model in an embodiment of the present application; Figure 4 A schematic diagram of a subject association judgment model in an embodiment of the present application; The same or similar reference numerals in the drawings represent the same or similar components. DETAILED DESCRIPTION

[0017] The present application is described in further detail below in conjunction with the accompanying drawings.

[0018] In a typical configuration of the present application, the terminal and the equipment of the service network each include one or more processors (CPU), input / output interface, network interface and memory.

[0019] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0020] Computer readable media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. Information can be computer program instructions, data structures, program devices or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

[0021] The embodiment of the present application provides a method for generating an event context, which performs text clustering on text data related to an event theme to obtain multiple clusters, then calculates a first similarity between each cluster and the event subject, and filters the clusters according to the first similarity, and after extracting summary information of each cluster, calculates a second similarity between the summary information of each cluster, and merges the clusters according to the second similarity, and after confirming that the clusters meet preset conditions, generates an event context about the event theme. Since the present scheme first extracts summary information of the sample when merging clusters, it can effectively reduce the interference of noise data on the merging process, and since the preset condition judgment when outputting the event context includes the identification of multiple events and the judgment of theme association, it can effectively exclude samples containing multiple event themes and samples that are not related to the theme event, so that the final output event context is more coherent and has stronger logic, so the accuracy of the final output event context has higher accuracy.

[0022] In actual scenarios, the execution subject of the method can be a user device, a network device, or a device formed by integrating a user device and a network device through a network, or it can also be an application running on the above device. The user device includes but is not limited to various terminal devices such as computers, mobile phones, and tablet computers; the network device includes but is not limited to network hosts, single network servers, multiple network server sets, or cloud computing-based computer sets. Here, the cloud is composed of a large number of hosts or network servers based on cloud computing, where cloud computing is a type of distributed computing, a virtual computer composed of a group of loosely coupled computer sets.

[0023] Figure 1 The following is a processing flow of an event context generation method provided by an embodiment of the present application. The method includes at least the following steps: Step S101 , performing text clustering on text data related to the event theme to obtain multiple clusters.

[0024] The event theme refers to the subject content corresponding to the event context that needs to be generated this time. For example, if the user wants to know the relevant context of the earthquake in City A in YY, the event theme can be "YY A event earthquake". If the user wants to know the relevant context of a dispute event that recently occurred in Company B, the event theme can be "Company B xx event". In this embodiment, the event theme can be recorded as "title".

[0025] The text data related to the event theme refers to the sample data related to the event theme that is initially obtained. For example, in an actual scenario, relevant news texts can be crawled from the network using keywords corresponding to the event theme by means of crawlers, etc., as the text data for this processing. In this step, a text clustering algorithm can be used to perform a preliminary text clustering on the text data, thereby obtaining multiple clusters, each of which can include at least one text data. In order to improve the processing speed, the SinglePass text clustering algorithm can be used in this embodiment to perform text clustering on the media data related to the event theme to obtain multiple clusters. The SinglePass text clustering algorithm, also known as the single-channel method or the single-pass method, processes each text data in turn, processes one sample data each time, and divides the sample data into an existing cluster or creates a new cluster based on the matching degree between the currently processed sample data and the existing cluster. The algorithm has a high time efficiency, so it can achieve efficient and fast clustering results.

[0026] In order to facilitate subsequent data processing and improve the accuracy of the processing results, the media data related to the event theme can be cleaned before text clustering is performed on the media data related to the event theme. The specific cleaning method can be set according to the needs of the actual application scenario. For example, the data cleaning in this embodiment may include removing stop words, deleting some texts containing all numbers or letters, deleting the same or similar texts published by the same media, etc.

[0027] For example, in this embodiment, the text data related to the event theme initially obtained is [text1, text2, text3, text4, …]. After data cleaning of each text data, each text data is processed using the SinglePass algorithm to obtain the following specific results: cluster c1[text1, text5, text10 ,…], cluster c2[text2, text3, text32 ,…], cluster c3[text7, text15, text100 , …], ….

[0028] Step S102, calculating a first similarity between each cluster and the event subject, and filtering the clusters according to the first similarity.

[0029] For each cluster, a text data that can represent the central idea of ​​all the text data in the cluster can be determined as the cluster center center of the cluster to represent the characteristics of the cluster. For example, one of the text data text1 in cluster c1 can be selected as the cluster center center1 of the cluster, and center1 is used to represent the cluster c1. Therefore, center1 can be used to represent cluster c1 containing the following text data [text1, text5, text10, ...]. Similarly, the cluster centers center2, center3, etc. of other clusters can be determined. Since center represents a cluster and is a text data from the cluster to represent the characteristics of this cluster, the first similarity can be calculated by calculating the text similarity between the center of each cluster and the event theme title. In actual scenarios, the solution for selecting the cluster center center can use algorithms such as DocRank.

[0030] Therefore, when calculating the first similarity between each cluster and the event subject, and filtering the clusters according to the first similarity, the first similarity between the cluster center of each cluster and the event subject can be calculated first, and the first similarity can be compared with the first threshold. The first similarity can be the cosine similarity of the text. Taking center1 as an example, center1 and title can be segmented and word embedded respectively to obtain the corresponding feature information f_ center1 and f_title, and then the cosine similarity between the feature information f_ center1 and f_title is calculated as the first similarity cos2 in the present application.

[0031] After obtaining the first similarity, the first similarity can be compared with the first threshold. The specific value of the first threshold can be set according to the needs of the actual scenario, for example, it can be set to 0.4 in this embodiment. If the first similarity is less than the first threshold, the cluster corresponding to the first similarity is deleted. On the contrary, if the first similarity is greater than or equal to the first threshold, the cluster corresponding to the first similarity is retained. Taking the aforementioned cluster c1 as an example, if the first similarity between its corresponding center1 and title is less than 0.4, it means that the text data in the cluster is not similar enough to the event theme in terms of text content, and cannot be used to generate the event context of the event theme. Therefore, the cluster can be deleted, thereby filtering out the cluster c1. If the first similarity between the corresponding center1 and title of cluster c1 is greater than or equal to 0.4, it means that the text data in the cluster has met the requirements of this processing in terms of text content. The similarity between the event theme can be used to generate the event context of the event theme, so the cluster can be retained and continue to be used as sample data for subsequent processing.

[0032] Step S103: extract summary information of each cluster, calculate the second similarity between the summary information of each cluster, and merge the clusters according to the second similarity.

[0033] Similarly, when processing each cluster in this step, the cluster center of each cluster can also be used. Figure 2 The schematic diagram of the process of extracting summary information and merging clusters in this embodiment is shown, which specifically includes the following processing steps: Step S201, first select a cluster center in each cluster. For example, for cluster c1, cluster c2 and cluster c3, select one text as the cluster center of each cluster, denoted as center1, center2 and center3.

[0034] Step S202, when extracting the summary information of each cluster, the cluster center of each cluster can be input into the large model (LLM, Large Language Model) to obtain the summary information abs1, abs2 and abs3 of each cluster. In the actual scenario, the prompt for the summary generation task can be pre-constructed according to the actual needs. For example, the prompt constructed in this embodiment is: "Your task is to summarize the given text in a concise sentence. The content of the summary must meet the following requirements: 1. The summary content is strictly prohibited from containing information that does not appear or mention in the original text; 2. The summary content must be concise, with emphasis and logical coherence, and the sentences should be connected naturally, and no self-expression; 3. The number of words must not exceed 40. The given text is: {center1}", after the prompt is input into the large model, the summary information abs1 about the text center1 can be obtained, and the summary information will meet the above 3 requirements and can represent the main content of cluster c1.

[0035] Step S203, after extracting the summary information corresponding to each cluster, the second similarity between these summary information can be calculated in pairs. For example, if the clusters in this embodiment are cluster c1, cluster c2 and cluster c3, and their corresponding summary information is abs1, abs2 and abs3 respectively, then the second similarity between abs1 and abs2, the second similarity between abs1 and abs3, and the second similarity between abs2 and abs3 can be calculated respectively. Similarly, the first similarity can also be the cosine similarity of the text. Taking the summary information abs1 corresponding to center1 and the summary information abs3 corresponding to center3 as examples, they can be segmented and word embedded respectively to obtain the corresponding feature information f_abs1 and f_abs3, and then the cosine similarity between the feature information f_abs1 and f_abs3 is calculated as the second similarity cos2 in the solution of this application.

[0036] Step S204, after obtaining the second similarity, the second similarity can be compared with the second threshold. The specific value of the second threshold can also be set according to the needs of the actual scenario, for example, it can be set to 0.5 in this embodiment. If the second similarity is less than the second threshold, the two clusters corresponding to the second threshold are retained; if the second similarity is greater than or equal to the second threshold, the two clusters corresponding to the second threshold are merged. Taking the aforementioned clusters c1 and c3 as an example, if the second similarity between the corresponding summary information abs1 and abs3 is less than 0.5, it means that the similarity between the summary information of the sample data in the two clusters does not meet the requirements of merging, and the two clusters may represent different event contents. At this time, the two clusters c1 and c3 are retained. If the second similarity between the summary information abs1 and abs3 corresponding to clusters c1 and c3 is greater than or equal to 0.5, it means that the similarity between the summary information of the sample data in the two clusters meets the requirements of merging, and the two clusters may represent the same event content. At this time, there is no need to retain the two clusters c1 and c3 at the same time, and they can be merged into one cluster. The specific way of merging can be to delete one of them. For example, in this embodiment, for clusters c1 and c3, cluster c1 can be deleted and only cluster c3 can be retained. Since the scheme extracts summary information of samples before merging clusters, the interference of noise data on the merging process can be effectively reduced.

[0037] Step S104, sorting the clusters according to their release time. The release time of a cluster refers to the release time of the text data contained in a certain cluster. In this embodiment, the earliest release time in the cluster can be used as the release time of the cluster. For example, for cluster c1, the release times of the three text data texts it contains are t1, t2 and t3 respectively. If t2 is earlier than t1 and t3, t2 can be used as the release time of cluster c1. After determining the release time of each cluster, they are sorted in the order of release time. The following is the result of sorting several clusters by release time according to the scheme of this embodiment. If the cluster center center of each cluster represents the corresponding cluster, the sorting result can be expressed as [center3, center6, center21, center44], which means that cluster c3 has the earliest release time, and cluster c44 has the latest release time.

[0038] Step S105, after confirming that the cluster meets the preset conditions, an event context about the event theme is generated. The preset conditions include two conditions, namely, the identification of multiple events and the judgment of theme association, wherein the identification of multiple events refers to determining whether the cluster center contains a single event theme, and the judgment of theme association refers to determining whether the cluster center is associated with the event theme being processed this time. Only when the above two conditions are met at the same time, that is, the cluster center of a certain cluster contains only a single event theme and is associated with the event theme being processed this time, is it confirmed that the cluster meets the preset conditions. At this time, the event context of the event theme can be generated based on this cluster.

[0039] Because in actual business scenarios, many news may not only report one event, but may report several events together in one news release. For the solution of this application, this form of news text will affect the sorting of event context. Therefore, when generating event context, you can filter this type of text data containing multiple event topics by multi-event recognition. If the center of a cluster contains only one event topic, then keep the center of the cluster. If the center of a cluster contains multiple event topics, then delete the center. In addition, in actual business scenarios, there are many news contents that may be irrelevant to the event topic of the event context we need to generate. Therefore, before generating the event context, you can further perform a topic association judgment to determine whether the center of the cluster is related to the event topic of the event context generated this time. If it is related, keep the center. If it is not, delete the center.

[0040] When generating an event context, it is possible to first determine whether each cluster meets the preset conditions. If the preset conditions are met, the cluster is retained, and an event context about the event topic is generated based on the retained cluster. If the preset conditions are not met, the cluster is deleted and is not used to generate an event context to avoid adding noise information, resulting in the generated event context lacking logic and accuracy. Among them, the specific form of the event context may include event information sorted by release time. For example, in this embodiment, taking a certain event topic "title1" as an example, its corresponding event context may be specifically as follows: At 8:00 on December 14, AAA fainted and was rushed to the hospital for treatment; At 19:00 on December 14, BBB decided to arrest AAA; At 14:00 on December 15, CCC went to the hospital to visit AAA.

[0041] In actual scenarios, the identification of multiple events and the determination of topic associations can also be achieved using large models. By pre-building the corresponding prompts and inputting them into the corresponding large models, the required results can be obtained. Figure 3 and Figure 4 Schematic diagrams of the multi-event recognition model and the topic association judgment model are shown respectively. Among them, the prompt word prompt1 input to the multi-event recognition model can be "If you are a news analyst, now you need to analyze whether the given text revolves around only one event, and then give the final result: yes or no. If the text only tells about one event, the answer is yes; if the text tells about no less than three events, the answer is no, and the given text is: center3." The prompt word prompt2 input to the topic association judgment model can be "If you are a news analyst, now you are given two texts: [topic, text]. First, you analyze what the event described in the text is, and then determine whether the event described in the text is related to the topic topic, and then give the final result: yes or no. You must summarize the events contained in the text, and then determine whether the event is related to the topic. If so, return yes, otherwise return no; the event in the text can be a specific explanation of the topic or an extension of the topic. If the two events are consistent, return yes, otherwise return no; it is necessary to analyze whether the text event is related to the topic before giving an answer, and the answer must be yes or no." The above two models can realize the recognition of multiple events and the judgment of topic association in this embodiment.

[0042] Based on another aspect of the present application, an embodiment of the present application also provides an event context generation device, which includes a memory for storing computer program instructions and a processor for executing computer program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the aforementioned event context generation method.

[0043] In particular, the methods and / or embodiments in the embodiments of the present application may be implemented as computer software programs. For example, the embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program code for executing the method shown in the flowchart. When the computer program is executed by the processing unit, the above functions defined in the method of the present application are executed.

[0044] It should be noted that the computer-readable medium described in this application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, 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 above. In this application, a computer-readable medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device.

[0045] In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0046] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0047] The flow chart or block diagram in the accompanying drawings shows the possible architecture, function and operation of the equipment, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated system for hardware that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0048] As another aspect, the present application further provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more computer program instructions, which may be executed by a processor to implement the methods and / or technical solutions of the above multiple embodiments of the present application.

[0049] It should be noted that the present application can be implemented in software and / or a combination of software and hardware, for example, it can be implemented using an application-specific integrated circuit (ASIC), a general-purpose computer or any other similar hardware device. In some embodiments, the software program of the present application can be executed by a processor to implement the above steps or functions. Similarly, the software program of the present application (including related data structures) can be stored in a computer-readable recording medium, for example, a RAM memory, a magnetic or optical drive or a floppy disk and the like. In addition, some steps or functions of the present application can be implemented in hardware, for example, as a circuit that cooperates with a processor to perform various steps or functions.

[0050] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or basic features of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present application is limited by the attached claims rather than the above description, so it is intended to include all changes that fall within the meaning and scope of the equivalent elements of the claims in the present application. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the device claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

Claims

1. A method for generating an event context, characterized in that: The method comprises: Performing text clustering on text data related to the event topic to obtain multiple clusters, each cluster including at least one piece of text data; Calculating a first similarity between each cluster and the event subject, and filtering the clusters according to the first similarity; Extracting summary information of each cluster, calculating a second similarity between the summary information of each cluster, and merging the clusters according to the second similarity; sorting the clusters according to their release time; After confirming that the cluster meets the preset conditions, an event context about the event topic is generated, wherein the preset conditions include that the cluster center of the cluster contains a single event topic and that the cluster center of the cluster is associated with the event topic, and the event context includes event information sorted by release time.

2. The method according to claim 1, characterized in that Before obtaining multiple clusters, text clustering of media data related to event topics also includes: Perform data cleaning on media data related to event topics.

3. The method according to claim 1, characterized in that: Perform text clustering on media data related to the event theme to obtain multiple clusters, including: The SinglePass text clustering algorithm is used to perform text clustering on media data related to event topics and obtain multiple clusters.

4. The method according to claim 1, characterized in that: Calculating a first similarity between each cluster and the event subject, and filtering the clusters according to the first similarity, including: Calculating a first similarity between a cluster center of each cluster and an event topic, and comparing the first similarity with a first threshold; If the first similarity is less than a first threshold, deleting the cluster corresponding to the first similarity; If the first similarity is greater than or equal to a first threshold, the cluster corresponding to the first similarity is retained.

5. The method according to claim 1, characterized in that Extract summary information for each cluster, including: The cluster center of each cluster is input into the large model to obtain summary information of each cluster.

6. The method according to claim 1, characterized in that Calculating a second similarity between summary information of each cluster, and merging the clusters according to the second similarity; Calculating a second similarity between summary information of cluster centers of each cluster, and comparing the second similarity with a second threshold; If the second similarity is less than a second threshold, retaining two clusters corresponding to the second threshold; If the second similarity is greater than or equal to a second threshold, two clusters corresponding to the second threshold are merged.

7. The method according to claim 1, characterized in that After confirming that the cluster meets the preset conditions, an event context about the event topic is generated, including: Determine whether each cluster meets the preset conditions; If the preset conditions are met, the cluster is retained, and an event context about the event topic is generated according to the retained cluster.

8. The method according to claim 7, characterized in that The method further comprises: If the preset condition is not met, the cluster is deleted.

9. An event context generating device, wherein: The device comprises a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the method according to any one of claims 1 to 8.

10. A computer readable medium having computer program instructions stored thereon, wherein the computer program instructions can be executed by a processor to implement the method according to any one of claims 1 to 8.

Images