Extended event display method, device and equipment and computer readable storage medium

Abstract

The application provides an extended event display method and device, equipment and a computer readable storage medium, which are applied to various scenes such as cloud technology, artificial intelligence, intelligent transportation and vehicle-mounted devices. The method comprises the following steps: obtaining a to-be-processed event by extracting event information of a hot event, and determining a core event from the to-be-processed event; performing keyword extraction on the core event to obtain keywords corresponding to the core event; searching and matching based on the to-be-processed event by using the keywords to obtain a candidate event; performing filtering on the candidate event in at least one dimension to obtain a context event; and performing extended event display according to the context event and the core event. Through the application, the efficiency and accuracy of the extended event display can be improved.

Description

Technical Field

[0001] This application relates to terminal technology, and in particular to an extended event display method, apparatus, device, and computer-readable storage medium. Background Technology

[0002] Currently, for some news events that have a long duration, machine learning can be used to extract important stages of the event's development as contextual information. This contextual information can then be used to expand the event's presentation or recommendations, allowing users to intuitively understand the event's development process through the displayed expanded events.

[0003] The common method for generating event context information using related technologies is to first generate events from articles, and then use similarity-based models or large-scale clustering methods to link similar events into a context. This large-scale clustering method is inefficient and can only process a small number of events within a certain timeframe. When the number of events exceeds tens of thousands, the accuracy of related technologies drops significantly, failing to meet the needs of routinely generating long-term contexts on a large scale, thus reducing the accuracy of extended event display. Summary of the Invention

[0004] This application provides an extended event display method, apparatus, device, and computer-readable storage medium, which can improve the efficiency and accuracy of extended event display.

[0005] The technical solution of this application embodiment is implemented as follows:

[0006] This application provides an extended event display method, including:

[0007] By extracting event information from hot events, events to be processed are obtained, and core events are identified from these events.

[0008] Keyword extraction is performed on the core events to obtain the keywords corresponding to the core events;

[0009] Using the keywords, search and match based on the events to be processed to obtain candidate events;

[0010] The candidate events are filtered in at least one dimension to obtain contextual events;

[0011] Extended events are displayed based on the contextual events and the core events.

[0012] This application provides an extended event display device, including:

[0013] The extraction module is used to extract event information from hot events to obtain events to be processed, and to identify core events from the events to be processed; and to extract keywords from the core events to obtain the keywords corresponding to the core events.

[0014] The matching module is used to search and match based on the keywords and the event to be processed to obtain candidate events;

[0015] A filtering module is used to filter the candidate events in at least one dimension to obtain contextual events;

[0016] The display module is used to display extended events based on the context events and the core events.

[0017] In the above-mentioned device, the extraction module is further configured to perform named entity recognition on the core event title corresponding to the core event to obtain at least one named entity; among the at least one named entity, a named entity of a preset noun type is used as the keyword.

[0018] In the above-mentioned device, the extended event display device further includes a word weight analysis module; the word weight analysis module is used to perform word weight analysis on the core event title when the number of keywords is less than a preset number threshold, to obtain at least one weighted word; among the at least one weighted word, at least one verb-type weighted word is determined as the keyword.

[0019] In the above-mentioned device, the at least one dimension includes at least one of the following: semantic dimension, keyword dimension, event popularity dimension, and event occurrence time dimension; the filtering module is further configured to score the candidate event in at least one dimension to obtain a score for the candidate event in at least one dimension; obtain a comprehensive score based on the at least one dimension score; and filter the candidate events based on the comprehensive score to obtain contextual events.

[0020] In the above-described device, the semantic dimension includes a semantic similarity dimension; the at least one dimension score includes a semantic similarity score; the filtering module is further configured to use a preset semantic analysis model to perform semantic analysis on the core event title of the core event and the candidate event title of the candidate event to obtain the core semantic features corresponding to the core event and the candidate semantic features corresponding to the candidate event; obtain the semantic similarity between the core event and the candidate event by calculating the feature vector distance between the core semantic features and the candidate semantic features; and score the candidate event based on the semantic similarity to obtain the semantic similarity score.

[0021] In the above device, the semantic dimension includes a semantic clustering dimension; the at least one dimension score includes a semantic clustering score; the filtering module is further configured to perform density clustering on the core event and the candidate event based on the semantic similarity between the core event and the candidate event to obtain at least one event cluster; and to score the candidate event according to the event cluster in which the core event belongs to obtain the semantic clustering score of the candidate event.

[0022] In the above-described device, the at least one dimension score includes: keyword score; the filtering module is further configured to calculate the word frequency of the keyword using a preset corpus, and calculate the inverse document frequency corresponding to the keyword based on the word frequency; take the keywords contained in the candidate event title of the candidate event as internal keywords, and score the candidate event based on the inverse document frequency of the internal keywords to obtain the keyword score; the keyword score is inversely proportional to the inverse document frequency.

[0023] In the above-mentioned device, the filtering module is further used to adjust the weight of the inverse document frequency of the internal keywords by a preset coefficient to obtain the weighted inverse document frequency; and to use a preset frequency adjustment factor to sum and average the minimum values ​​of the weighted inverse document frequencies to obtain the keyword score.

[0024] In the above device, the event information includes: event popularity; the filtering module is further configured to determine the corresponding target popularity interval in at least one preset popularity interval based on the event popularity of the candidate event; and to obtain the popularity score corresponding to the candidate event based on the preset popularity coefficient corresponding to the target popularity interval.

[0025] In the above device, the event information includes: the event occurrence time; the at least one dimension score includes: the time dimension score; the filtering module is further used to calculate the time difference between the event occurrence time of the candidate event and the context event; based on the time difference, determine the corresponding target time difference interval in at least one preset time difference interval, and use the preset score corresponding to the target time difference interval as the time score corresponding to the candidate event.

[0026] In the above-mentioned device, the extraction module is further configured to obtain at least one search event with an event popularity higher than a preset popularity threshold from the background search data corresponding to the preset application as a hot event; extract the event title of each search event in the at least one search event, as well as at least one of the event popularity and the event occurrence time of each search event, and use the event popularity and the event occurrence time as the event information.

[0027] In the aforementioned device, the display module is further configured to sort the events in time, generate an event timeline based on the sorting results and the core events, and display extended events through the event timeline.

[0028] This application provides an electronic device, including:

[0029] Memory, used to store executable instructions;

[0030] The processor, when executing executable instructions stored in the memory, implements the extended event display method provided in the embodiments of this application.

[0031] This application provides a computer-readable storage medium storing executable instructions for implementing the extended event display method provided in this application when executed by a processor.

[0032] This application provides a computer program product, including a computer program or instructions, characterized in that, when the computer program or instructions are executed by a processor, they implement the extended event display method provided in this application.

[0033] The embodiments of this application have the following beneficial effects:

[0034] This application embodiment recalls candidate events related to the core event from the events to be processed based on the keywords of the core event, and then filters the candidate events in at least one dimension to obtain contextual events. This achieves noise reduction processing of the candidate events to obtain contextual events, and combines the core event with the contextual events to display extended events. This not only reduces the computational workload of directly generating event context through large-scale clustering and improves the efficiency of filtering contextual events, thereby improving the efficiency of extended event display; but also significantly improves the accuracy of the filtered contextual events through keyword matching and at least one dimension of filtering and noise reduction, achieving accurate event context construction under a large number of events, thereby improving the accuracy of extended event display. Attached Figure Description

[0035] Figure 1 This is an optional structural diagram of the extended event display system architecture provided in the embodiments of this application;

[0036] Figure 2 This is an optional structural schematic diagram of the extended event display device provided in the embodiments of this application;

[0037] Figure 3 This is an optional flowchart illustrating the extended event display method provided in the embodiments of this application;

[0038] Figure 4This is an optional effect diagram of an event timeline provided in an embodiment of this application;

[0039] Figure 5 This is an optional flowchart illustrating the extended event display method provided in the embodiments of this application;

[0040] Figure 6 This is an optional flowchart illustrating a preset semantic analysis model training method provided in an embodiment of this application;

[0041] Figure 7 This is an optional flowchart illustrating the extended event display method provided in the embodiments of this application;

[0042] Figure 8 This is an optional flowchart illustrating the extended event display method provided in the embodiments of this application;

[0043] Figure 9 This is an optional flowchart illustrating the extended event display method provided in the embodiments of this application;

[0044] Figure 10 This is an optional flowchart illustrating the extended event display method provided in the embodiments of this application;

[0045] Figure 11 This is an optional flowchart illustrating the contextual event filtering method provided in the embodiments of this application;

[0046] Figure 12 This is an optional schematic diagram of the DBSCAN clustering process provided in the embodiments of this application. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0048] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0049] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0051] In the implementation of this application, the collection and processing of relevant data should strictly comply with the requirements of relevant laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0052] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0053] 1) Artificial Intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. In other words, AI is a comprehensive technology within computer science used to capture the essence of intelligence and produce new intelligent machines that can react in a way similar to human intelligence. Furthermore, AI is used to study the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities. Moreover, AI technology is a comprehensive discipline involving a wide range of fields, encompassing both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operating / interactive systems, and mechatronics. AI software technologies mainly include computer vision, speech processing, natural language processing, and machine learning (ML) / deep learning.

[0054] 2) Natural Language Processing (NLP) is an important field within computer science and artificial intelligence. It refers to the study of theories and methods that enable effective communication between humans and computers using natural language. Therefore, NLP is a science integrating linguistics, computer science, and mathematics; consequently, research in NLP involves natural language, that is, the language people use in daily life, thus NLP is closely related to linguistics. NLP technologies typically include Machine Reading Comprehension (MRC), text processing, semantic understanding, machine translation, question answering, and knowledge graphs.

[0055] 3) Machine learning is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory, among others. It studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence. Its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0056] 4) Named Entity Recognition (NER), also known as entity recognition, entity segmentation, and entity extraction, is used to locate and classify named entities in text into predefined categories, such as people, organizations, locations, time expressions, quantities, currency values, percentages, etc. Typically, the task of NER is to identify three main categories (entity, time, and number) and seven subcategories (person names, organization names, place names, time, date, currency, and percentage) of named entities in the text to be processed. In this embodiment, named entity recognition is used to obtain entities of preset entity types, such as person names and place names.

[0057] 5) Density-Based Spatial Clustering of Applications with Noise (DBSCAN): A representative density-based clustering algorithm.

[0058] 6) Inverse Document Frequency (IDF): The higher the inverse document frequency, the less information gain the corresponding words have, and the smaller their contribution to text classification.

[0059] 7) Bidirectional Encoder Representations from Transformers (BERT): A pre-training technique for Natural Language Processing (NLP).

[0060] Currently, related technologies for generating event timelines typically involve aggregating articles within time windows, denoising and clustering them to create article clusters, and finally selecting representative articles to generate events. Furthermore, when generating timelines from events, a similarity-based model is generally used to determine whether two events belong to the same timeline; alternatively, large-scale clustering is used to directly group all events into multiple clusters, each cluster representing a timeline. It's clear that these methods require generating events from articles and then connecting them into timelines, both steps involving significant challenges and low efficiency. Moreover, directly generating timelines through similarity metrics or clustering can only process a small number of events within a given timeframe. When the event database exceeds tens of thousands, the accuracy of these technologies drops significantly, failing to meet the demands of routinely generating large-scale timelines spanning long periods. Therefore, current technologies for obtaining timeline events are both inefficient and inaccurate, resulting in low efficiency and accuracy in expanding event display based on timeline events.

[0061] This application provides an extended event display method, apparatus, device, and computer-readable storage medium, which can improve the efficiency and accuracy of extended event display. The following describes exemplary applications of the electronic devices provided in this application. These electronic devices can be implemented as various types of user terminals such as smartphones, smartwatches, laptops, tablets, desktop computers, set-top boxes, mobile devices (e.g., mobile phones, portable music players, personal digital assistants, dedicated messaging devices, portable gaming devices), smart voice interaction devices, smart home appliances, and in-vehicle terminals, or as servers. The following describes exemplary applications when the electronic device is implemented as a server.

[0062] See Figure 1 , Figure 1 This is an optional architecture diagram of the event network system 100 provided in the embodiments of this application. The terminal 400 is connected to the server 200 through the network 300, which can be a wide area network or a local area network, or a combination of the two.

[0063] Server 200 is used to extract event information from hot events to obtain events to be processed, and to identify core events from these events; to extract keywords from the core events to obtain corresponding keywords; to match recall events from the events to be processed based on the keywords; to filter the recall events in at least one dimension to obtain candidate events; and to display extended events based on the candidate events and the core events. For example, an event timeline is generated and displayed based on the candidate events and the core events, or related terms are recommended and displayed at preset page positions corresponding to the core events. Here, server 200 can push the event timeline or related terms as extended events to terminal 400, or it can save the extended events in the corresponding storage space of server 200 for terminal 400 to access.

[0064] Terminal 400 is used to access server 200 and obtain extended events through a web client or application; or, to receive extended events pushed by server 200 and then display the extended events on the graphical interface 410 of the web client or application.

[0065] In some embodiments, server 200 may be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Terminal 400 may be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, smart voice interaction device, smart home appliance, or in-vehicle terminal, but is not limited to these. Terminals and servers can be directly or indirectly connected via wired or wireless communication, which is not limited in this embodiment.

[0066] See Figure 2 , Figure 2 This is a schematic diagram of the structure of the server 200 provided in the embodiments of this application. Figure 2 The server 200 shown includes at least one processor 210, memory 250, at least one network interface 220, and a user interface 230. The various components in server 200 are coupled together via a bus system 240. It is understood that the bus system 240 is used to implement communication between these components. In addition to a data bus, the bus system 240 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 2 The general labeled all buses as Bus System 240.

[0067] Processor 210 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., where the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0068] User interface 230 includes one or more output devices 231 that enable the presentation of media content, including one or more speakers and / or one or more visual displays. User interface 230 also includes one or more input devices 232, including user interface components that facilitate user input, such as a keyboard, mouse, microphone, touch screen display, camera, other input buttons and controls.

[0069] The memory 250 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 250 may optionally include one or more storage devices physically located away from the processor 210.

[0070] The memory 250 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 250 described in this application embodiment is intended to include any suitable type of memory.

[0071] In some embodiments, memory 250 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0072] Operating system 251 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0073] The network communication module 252 is used to reach other computing devices via one or more (wired or wireless) network interfaces 220, exemplary network interfaces 420 including: Bluetooth, WiFi, and Universal Serial Bus (USB), etc.

[0074] Presentation module 253 is configured to enable the presentation of information (e.g., a user interface for operating peripheral devices and displaying content and information) via one or more output devices 231 associated with user interface 230 (e.g., a display screen, a speaker, etc.).

[0075] The input processing module 254 is used to detect and translate one or more user inputs or interactions from one or more input devices 232.

[0076] In some embodiments, the apparatus provided in this application can be implemented in software. Figure 2 An extended event display device 255 stored in memory 250 is shown. It can be software in the form of programs and plug-ins, including the following software modules: extraction module 2551, matching module 2552, filtering module 2553 and display module 2554. These modules are logical and can therefore be arbitrarily combined or further split according to the functions they implement.

[0077] The functions of each module will be explained below.

[0078] In other embodiments, the apparatus provided in this application can be implemented in hardware. As an example, the apparatus provided in this application can be a processor in the form of a hardware decoding processor, which is programmed to execute the extended event display method provided in this application. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0079] In some embodiments, the terminal or server can implement the extended event display method provided in this application by running a computer program. For example, the computer program can be a native program or software module in an operating system; it can be a native application (APP), that is, a program that needs to be installed in the operating system to run, such as a social application APP or a messaging APP; it can also be a mini-program, that is, a program that only needs to be downloaded to a browser environment to run; or it can be a mini-program or web client program that can be embedded in any APP. In short, the above-mentioned computer program can be any form of application, module or plugin.

[0080] The extended event display method provided in this application will be described in conjunction with exemplary applications and implementations of the server provided in the embodiments of this application. Furthermore, the event integration method provided in the embodiments of this application can be applied to various scenarios such as cloud technology, artificial intelligence, smart transportation, and in-vehicle systems.

[0081] See Figure 3 , Figure 3 This is an optional flowchart illustrating an extended event display method provided in an embodiment of this application, which will be combined with... Figure 3 The steps shown are explained.

[0082] S101. By extracting event information from hot events, events to be processed are obtained, and core events are identified from the events to be processed.

[0083] In this embodiment, the electronic device can obtain hot events from the backend server of a preset application. Hot events may include at least one event whose popularity exceeds a preset popularity threshold. The electronic device extracts event information from each hot event and generates a corresponding event to be processed based on the event information.

[0084] In some embodiments, the electronic device can select at least one search event with a popularity exceeding a preset popularity threshold from the background search data corresponding to a preset application as a hot event. For example, the electronic device can select the top M topics with high search rankings and / or high levels of interaction and discussion from the background search data of news applications or instant messaging social applications, such as ranking information for topic events. Here, M is a positive integer greater than or equal to 1. The electronic device extracts information needed to generate an event timeline from at least one search event as event information.

[0085] Here, the popularity of an event can be determined based on at least one of the data that characterizes the attention given to an event, such as search volume, click volume, readership, mention volume, and interaction volume (e.g., comment volume, repost volume, like volume).

[0086] In some embodiments, since hot events usually have a time continuity and the same hot event may correspond to different event titles at different time stages, electronic devices can not only obtain hot events from current background search data, such as the latest hot search list data of the day or one hour, but also obtain hot events from preset time ranges, such as hot search list data within the last six months, so as to ensure that the obtained hot events can contain event information within different time spans.

[0087] In some embodiments, the electronic device may use at least one of the event popularity and event occurrence time, the event title and the content identifier of each search event in at least one search event as event information.

[0088] In some embodiments, event popularity may include the highest event popularity, such as the highest historical search volume and readership of the event; content identifiers may include the article ID corresponding to the event, such as the article ID of the content document corresponding to the event in the background data. Electronic devices can construct a database of events to be processed based on the event title and content identifier extracted from each event in the hot events, as well as the event popularity and event occurrence time, as shown in Table 1 below:

[0089]

[0090] Table 1

[0091] It should be noted that the article ID corresponding to each event title in Table 1 can be one or more, and this application embodiment does not limit it.

[0092] It can be seen that by directly using the background search data of the preset application as the source of hot events, events to be processed can be generated directly based on existing ranking data. Compared with the current process of generating events through articles in related technologies, this saves the event processing process and improves the efficiency of generating contextual events.

[0093] In some embodiments, the electronic device can select a target event as the core event by receiving a selection instruction for a target event among the events to be processed. The electronic device can also determine the core event from the events to be processed based on event information, such as the popularity information of each event. The specific selection depends on the actual situation, and this application embodiment does not limit the choice.

[0094] S102. Extract keywords from the core events to obtain the keywords corresponding to the core events.

[0095] In this embodiment, the electronic device can use a core word algorithm to extract keywords from the event information of core events, obtaining at least one word as the keyword corresponding to the core event. The electronic device can identify entities or words with high weight as keywords from the event information of core events, and use these keywords to recall candidate events with high relevance to the core events from the events to be processed.

[0096] In some embodiments, the event information includes an event title; the process of the electronic device extracting keywords from the core event to obtain the keywords corresponding to the core event can be implemented through S1021-S1022, which will be explained in conjunction with each step.

[0097] S1021. Perform named entity recognition on the core event title of the core event to obtain at least one named entity.

[0098] In this embodiment of the application, the electronic device obtains the core event title from the event information of the core event, and performs named entity recognition on the core event title to obtain at least one named entity.

[0099] In some embodiments, electronic devices can use a character-word convolutional neural network (CWCNN) to perform named entity recognition on core event titles to obtain named entities; alternatively, they can use a long short-term memory network (LSTM) combined with a conditional random field (CRF) model, or other natural language processing models to perform named entity recognition to obtain named entities. The specific choice depends on the actual situation, and this application embodiment does not limit the choice.

[0100] S1022. In at least one named entity, a named entity of the preset noun type is used as a keyword.

[0101] In this embodiment of the application, the electronic device can use at least one named entity of a preset noun type as a keyword; for example, at least one named entity can use a person's name and a place name as keywords.

[0102] In some embodiments, the electronic device may also use other pre-defined noun types of named entities in at least one named entity as keywords, such as brand names, major event names, etc., as keywords, according to the actual needs of the project. The specific selection is based on the actual situation and is not limited in this application embodiment.

[0103] In some embodiments, if the number of keywords obtained by the electronic device through the named entity extraction process of S1021-S1022 is less than a preset threshold, the electronic device can further extract keywords from the core event title through word weight analysis, which can be implemented through S1023-S1024, as follows.

[0104] S1023. Perform word weight analysis on the core event titles to obtain at least one weighted word.

[0105] In this embodiment of the application, the electronic device can perform word weight analysis on the words contained in the core event title to obtain at least one weighted word.

[0106] In some embodiments, electronic devices can perform word weight analysis on core event titles using an eXtreme Gradient Boosting (xgboost) classification model, obtaining words with weights higher than a preset weight threshold as at least one weighted word. Other ensemble learning models can also be selected for word weight analysis based on actual circumstances; the specific selection depends on the actual situation and is not limited in this application.

[0107] S1024. Among at least one weighted vocabulary, identify at least one weighted vocabulary of the verb type as a keyword.

[0108] In this embodiment of the application, the electronic device adds verb-type words from at least one weighted vocabulary to the keywords.

[0109] In some embodiments, the preset quantity threshold can be three. If the number of keywords obtained through named entity recognition is greater than or equal to three, the electronic device can add keywords without performing the word weight analysis process in S1023-S1024. If the preset quantity threshold is less than three, the electronic device can add all weighted words of at least one verb type obtained through word weight analysis to the keywords. In some embodiments, the electronic device can also disregard the number of keywords and directly obtain keywords through S1021-S1024 by combining named entity extraction with word weight analysis. The specific selection depends on the actual situation, and this application embodiment does not limit it.

[0110] S103. Using keywords, search and match based on the events to be processed to obtain candidate events.

[0111] In this embodiment of the application, the electronic device can use keywords to match events to be processed, and select at least one event containing the keyword as a candidate event.

[0112] In some embodiments, electronic devices can also use keywords to perform web searches, extract event information from articles or web content that match the keywords, and generate candidate events based on the extracted event information. For example, from the searched articles or web content, the title of the article or web content is extracted as the event title; the link of the article or web content is extracted as the content identifier; the number of views and / or interactions of the article or web content is extracted as the event popularity; the publication time of the article or web content is extracted as the event occurrence time; and then candidate events are generated based on the event title, content identifier, event popularity, and event occurrence time. This allows for the retrieval and generation of candidate events across a wider network using keywords.

[0113] In some embodiments, electronic devices may also perform semantic expansion based on keywords to obtain similar words or synonyms corresponding to the keywords, and further expand the keywords based on similar words or synonyms to more accurately match a larger number of candidate events.

[0114] S104. Filter the candidate events in at least one dimension to obtain the context events.

[0115] In this embodiment of the application, the electronic device can perform at least one dimension of screening on the candidate events recalled by keyword matching of the core event, perform noise reduction processing on the candidate events, and remove candidate events with low relevance to the core event through evaluation of at least one dimension to obtain the context event.

[0116] In some embodiments, at least one dimension may include at least one of a semantic dimension, a keyword dimension, and an event structure information dimension. The semantic dimension represents filtering candidate events based on semantic similarity; the keyword dimension represents filtering based on the matching keywords contained in the candidate events; and the event structure information dimension represents filtering based on structured information in the event information, such as event occurrence time and event popularity. The electronic device can score candidate events based on at least one of these dimensions to obtain a score for each candidate event in at least one dimension; obtain a comprehensive score based on the scores for each dimension; and filter candidate events based on the comprehensive score to obtain contextual events.

[0117] It should be noted that, in the embodiments of this application, the process of the electronic device scoring candidate events based on at least one dimension does not have a specific order. It can be executed in any order, or the scoring of at least one dimension can be performed in parallel, depending on the actual engineering needs. The specific choice depends on the actual situation, and this embodiment of the application does not impose any limitations.

[0118] In some embodiments, the electronic device may sum the scores of at least one dimension to obtain a comprehensive score; or it may perform weight adjustment or other forms of algorithm processing on the scores of at least one dimension to obtain a comprehensive score. The specific choice is made according to the actual situation, and the embodiments of this application do not limit it.

[0119] In this embodiment, the electronic device can sort candidate events based on a comprehensive score corresponding to at least one dimension, and determine the top N events as context events according to the needs of the actual application. Here, N is a positive integer greater than or equal to 1.

[0120] S105. Expand the event display based on candidate events and core events.

[0121] In this embodiment of the application, the electronic device can use candidate events as extended events corresponding to the core event, and display the extended events based on the candidate events and the core event.

[0122] In some embodiments, electronic devices can sort event timelines, generate event timelines based on the sorting results and core events, and then display extended events through these event timelines.

[0123] For example, the extended event display generation method of this application embodiment can be used in search application scenarios. When a user uses a terminal to search for events with continuous information, the method in this application embodiment can be used to generate and display the event context corresponding to the event searched by the user, such as... Figure 4 As shown. Figure 4 A diagram illustrating an event timeline is shown. Figure 4 On the page, the event timeline includes a core event 40 and several related events 41 displayed in a list below the core event 40. Each related event is ordered according to its occurrence time, allowing users to understand the sequential development of events. Furthermore, within one of the related events, such as... Figure 4 The context event 42 shown in the figure receives a user's viewing operation, such as a click operation. In the case of this operation, the electronic device can enter the article page corresponding to the context event 42 based on the content identifier, such as the article ID. Figure 4 Control 43, namely the "Expand More" button, can expand all network events in the current network event list when a click operation is received, thereby providing the user with complete event network information and realizing expanded event display.

[0124] In some embodiments, the extended event display method in this application can also be used in information flow scenarios for search recommendations; for example, when a user completes reading an article, search terms related to the article's context can be recommended at the bottom of the article to achieve extended event display, which satisfies the user's need to understand related events and increases the frequency of user searches.

[0125] In some embodiments, the extended event display method in this application can also be used in the scenario of related search. When a user searches for all related events, the relevant event entries can be directly displayed in the search results to realize extended event display, guide users to pay attention to related events, and increase the frequency of user searches.

[0126] In some embodiments, the extended event display in this application can also be used in information flow distribution scenarios. By directly displaying key event information and corresponding articles, the extended event display can meet users' needs to understand the cause and effect of events, increase users' interest in clicking on articles, and facilitate article distribution.

[0127] It is understood that, according to the keywords of the core event, the embodiments of this application recall candidate events related to the core event from the events to be processed, and then filter the candidate events in at least one dimension to obtain contextual events. This achieves noise reduction processing of the candidate events to obtain contextual events, and combines the core event with the contextual events to display extended events. This not only reduces the computational amount of directly generating event context through large-scale clustering and improves the efficiency of filtering contextual events, thereby improving the efficiency of displaying extended events, but also significantly improves the accuracy of the filtered contextual events through keyword matching and at least one dimension of filtering and noise reduction. This achieves accurate event context construction under a large number of events, thereby improving the accuracy of displaying extended events.

[0128] In some embodiments, the semantic dimension in at least one dimension includes a semantic similarity dimension; the score of at least one dimension includes a semantic similarity score. The process by which an electronic device scores candidate events in at least one dimension to obtain the score of at least one dimension corresponding to the candidate events can be as follows: Figure 5 As shown, this is achieved through steps S201-S203, which will be explained in conjunction with each step.

[0129] S201. Using a pre-defined semantic analysis model, perform semantic analysis on the core event title of the core event and the candidate event title of the candidate event to obtain the core semantic features corresponding to the core event and the candidate semantic features corresponding to the candidate event.

[0130] In this embodiment of the application, the electronic device can obtain the core event title of the core event and the candidate event title of the candidate event, and use a preset semantic analysis model to perform semantic analysis on the core event title and the candidate event title to obtain the core semantic features corresponding to the core event and the candidate semantic features corresponding to the candidate event.

[0131] In some embodiments, the preset semantic analysis model can be the BERT model. Electronic devices can use the BERT model to perform semantic analysis on the core event title and the candidate event title respectively, and obtain the first dimension cls vector output by the BERT model for the core event title and the candidate event title respectively, such as a 768-dimensional feature vector, as the core semantic feature and the candidate semantic feature.

[0132] In some embodiments, before executing S201, the electronic device may acquire an initial semantic analysis model and training samples; for example, it may acquire 10,000 manually annotated training samples of related events; wherein the training samples include a first event title and a second event title, the first event title and the second event title being the event titles of related events. The electronic device constructs a training model with a dual-tower structure containing the first initial semantic analysis model and the second initial semantic analysis model based on the initial semantic analysis model, such as... Figure 6 As shown. For example, the first initial semantic analysis model and the second initial semantic analysis model can be BERT models.

[0133] In some embodiments, such as Figure 6 As shown, in each training round, the electronic device can perform semantic analysis on the first event title in the training samples using a first initial semantic analysis model to obtain the corresponding cls vector, which serves as the first semantic feature. Similarly, it can perform semantic analysis on the second event title in the training samples using a second initial semantic analysis model to obtain the corresponding cls vector, which serves as the second semantic feature. The electronic device calculates the cosine similarity between the first and second semantic features and performs cross-entropy loss calculation based on the cosine similarity, which serves as the training loss for each round. Based on the training loss, the electronic device adjusts the training model, which includes the first and second initial semantic analysis models, until the preset training conditions are met, resulting in a pre-trained preset semantic analysis model.

[0134] S202. By calculating the feature vector distance between the core semantic features and the candidate semantic features, the semantic similarity between the core event and the candidate event is obtained.

[0135] In this embodiment of the application, the electronic device can calculate the feature vector distance between the core semantic features and the candidate semantic features. For example, the cosine distance is calculated as the semantic similarity between the core event and the candidate event.

[0136] In some embodiments, electronic devices may also calculate the semantic similarity between core semantic features and candidate semantic features using other methods. The specific method to be selected depends on the actual situation and is not limited in this application embodiment.

[0137] S203. Score the candidate events based on semantic similarity to obtain a semantic similarity score.

[0138] In this embodiment of the application, the electronic device can score candidate events based on semantic similarity to obtain a semantic similarity score.

[0139] In some embodiments, the electronic device can normalize the semantic similarity between the calculated core semantic features and candidate semantic features to a numerical range of [0,1] to obtain a semantic similarity score.

[0140] In some embodiments, the semantic similarity score is at most 1 point, but it can also be set according to the needs of the actual project. The specific choice is made according to the actual situation, and this application embodiment does not limit it.

[0141] Understandably, combining semantic similarity to filter candidate events can improve the accuracy of identifying contextual events from the candidate events, thereby improving the accuracy of expanding event display based on candidate events and core events.

[0142] In some embodiments, the semantic dimension includes: a semantic clustering dimension; at least one dimension score includes: a semantic clustering score; the process by which an electronic device scores candidate events in at least one dimension to obtain at least one dimension score corresponding to the candidate events can be as follows: Figure 7 As shown, steps S301-S302 will be explained in conjunction with each step.

[0143] S301. Based on the semantic similarity between core events and candidate events, perform density clustering on core events and candidate events to obtain at least one event cluster.

[0144] In some embodiments, the electronic device may perform DBSCAN density clustering on the candidate events based on the semantic similarity between the core event and the candidate events to obtain at least one event cluster.

[0145] In some embodiments, the electronic device can calculate the semantic similarity between the core event and the candidate event based on the method in S201-S202, and perform neighborhood size measurement in DBSCAN clustering based on the semantic similarity, thereby realizing DBSCAN clustering based on semantic similarity.

[0146] In some embodiments, the preset minimum density sample number for DBSCAN clustering can be 1; the preset density threshold can be 0.2, and the specific selection can be made according to the actual situation. This application embodiment does not limit the specific selection.

[0147] S302. Based on the event cluster to which the core event belongs, score the candidate events to obtain the semantic clustering score of the candidate events.

[0148] In this embodiment of the application, candidate events belonging to the same event cluster as the core event are more relevant to the core event. Electronic devices can score candidate events by increasing the scores of candidate events belonging to the same event cluster as the core event, thereby obtaining the semantic clustering score of the candidate events.

[0149] In some embodiments, electronic devices can use DBSCAN density clustering to assign a semantic clustering score of 1 to candidate events that belong to the same event cluster as the core event, and assign no score to candidate events that do not belong to the same event cluster as the core event, thereby obtaining the semantic clustering score of the candidate events.

[0150] Understandably, scoring and filtering candidate events through semantic density clustering can significantly improve the scores of candidate events belonging to the same cluster as the core event, thereby increasing the accuracy of identifying contextual events from the candidate events and ultimately improving the accuracy of displaying extended events based on candidate events and the core event. Furthermore, the difference between this application's embodiment and current related technologies that use clustering methods to obtain contextual events lies in the fact that related technologies obtain multiple clusters through semantic clustering, with events within each cluster representing contextual events in the same context. This direct clustering results in lower accuracy for contextual events. In contrast, the semantic density clustering in this application serves a noise reduction function. From candidate events obtained based on keyword matching, candidate events not belonging to the same event cluster as the core event are removed by lowering their scores, thus significantly improving the accuracy of identifying contextual events from the candidate events and consequently improving the accuracy of displaying extended events based on candidate events and the core event.

[0151] In some embodiments, where the semantic dimension includes a keyword dimension; at least one dimension score may include: keyword scoring; the process by which an electronic device scores candidate events in at least one dimension to obtain at least one dimension score corresponding to the candidate events can be as follows: Figure 8 As shown, this is achieved through steps S401-S402, which will be explained in conjunction with each step.

[0152] S401. Calculate the word frequency of keywords using a pre-set corpus, and calculate the inverse document frequency of the keywords based on the word frequency.

[0153] In this embodiment of the application, the electronic device can obtain the word frequency of the keyword by segmenting the event title in the preset corpus, and calculate the inverse document frequency corresponding to the keyword based on the word frequency.

[0154] In some embodiments, the preset corpus may contain 1 million historical event titles, and the electronic device can calculate the inverse document frequency according to formula (1), as follows:

[0155] (1)

[0156] In formula (1), The event titles are preset in the corpus, such as 1 million; The number of candidate event titles containing this word. The number of keywords, Characterization The first keyword indivual. For the first The inverse document frequency of each keyword. It can be seen that the more common the keyword is in the candidate event title, the larger the denominator of formula (1) is, the smaller the inverse document frequency is, and the closer it is to 0.

[0157] S402. Use the keywords contained in the candidate event titles as internal keywords, and score the candidate events based on the inverse document frequency of the internal keywords to obtain the keyword score; the keyword score is inversely proportional to the inverse document frequency.

[0158] In this embodiment of the application, the electronic device can use the keywords contained in the candidate event title as internal keywords, and based on the inverse document frequency of the internal keywords, score the candidate event according to the IDF value of the keywords hit in the candidate event title to obtain the keyword score.

[0159] In some embodiments, the electronic device can adjust the weight of the inverse document frequencies of internal keywords by a preset coefficient to obtain the weighted inverse document frequencies; using a preset frequency adjustment factor, the minimum value of the weighted inverse document frequencies is summed and averaged to obtain the keyword score. For example, the above process can be implemented using formula (2), as follows:

[0160] (2)

[0161] In formula (2), For example, using preset coefficients, It can be 5, which means that the inverse document frequency is weighted by 5 times. For example, to preset the frequency adjustment factor, It can be 1. Keyword scores are calculated. It can be seen that the higher the inverse document frequency, the lower the keyword score; that is, keyword score is inversely proportional to inverse document frequency.

[0162] In some embodiments, the keyword score can be up to 2 points. The preset coefficient and preset frequency adjustment factor can also be adjusted according to the actual needs of the project. The specific selection is based on the actual situation and is not limited in this application embodiment.

[0163] Understandably, filtering candidate events by inverse document frequency of keywords can improve the accuracy of identifying contextual events from the candidate events, thereby improving the accuracy of displaying extended events based on candidate events and core events.

[0164] In some embodiments, event information includes: event popularity; where at least one dimension includes an event popularity dimension, the at least one dimension score includes: popularity dimension score; the process by which an electronic device scores candidate events in at least one dimension to obtain the at least one dimension score corresponding to the candidate events can be as follows: Figure 9 As shown, this is achieved through steps S501-S502, which will be explained in conjunction with each step.

[0165] S501. Based on the event popularity of the candidate event, determine the corresponding target popularity interval in at least one preset popularity interval.

[0166] S502. Based on the preset popularity coefficient corresponding to the target popularity range, obtain the popularity score corresponding to the candidate event.

[0167] In this embodiment of the application, the electronic device can combine at least one preset score value corresponding to at least one preset popularity interval, and score the candidate event based on the popularity of the event in the event information, such as the historical highest popularity.

[0168] In this embodiment of the application, the electronic device determines the corresponding target popularity interval within at least one preset popularity interval based on the popularity of the candidate event, and obtains the popularity score corresponding to the candidate event based on the preset popularity coefficient corresponding to the target popularity interval. For example, as shown in formula (3), it is as follows:

[0169] (3)

[0170] In formula (3), The event popularity is defined as 200 and 50, which are the preset popularity range thresholds after numerical unit normalization. According to formula (3), electronic devices can assign 1 point to candidate events with a popularity higher than 2 million; and assign 1 point to events with a popularity between 50 and 2 million. The coefficient is used to calculate the score, and those with less than 500,000 popularity points receive 0 points, thus obtaining the popularity score. .

[0171] Understandably, filtering candidate events based on their popularity can improve the accuracy of identifying relevant events from among the candidate events, thereby improving the accuracy of expanding the event display based on the candidate events and the core events.

[0172] In some embodiments, event information includes: event occurrence time; where at least one dimension includes the event occurrence time dimension, the at least one dimension score includes: time dimension score; the process by which an electronic device scores candidate events in at least one dimension to obtain the at least one dimension score corresponding to the candidate events can be as follows: Figure 10 As shown, this is achieved through steps S601-S603, which will be explained in conjunction with each step.

[0173] S601. Calculate the time difference between the occurrence times of the candidate event and the core event.

[0174] S602. Based on the time difference, determine the corresponding target time difference interval in at least one preset time difference interval, and use the preset score corresponding to the target time difference interval as the time score corresponding to the candidate event.

[0175] In this embodiment, the electronic device can score candidate events based on their occurrence time to obtain a time score for the candidate event. The time difference between the occurrence time of the candidate event and the occurrence time of the core event is inversely proportional to the time score. For example, as shown in formula (4), the time score is as follows:

[0176] (4)

[0177] In formula (4), Given the time difference between the occurrence times of candidate events and contextual events, electronic devices can assign a score of 1 for events with a time difference within 24 hours, 0.5 for events with a time difference within 72 hours, and 0 for events with a time difference of 0, based on formula (4), thus obtaining the time score for the candidate events. .

[0178] It is understandable that filtering candidate events by using the time difference between the occurrence times of candidate events and core events can improve the accuracy of identifying contextual events from the candidate events, thereby improving the accuracy of displaying extended events based on candidate events and core events.

[0179] The following will describe an exemplary application of the embodiments of this application in a real-world application scenario.

[0180] The extended event display method in this embodiment, when applied to displaying event context information in a search scenario, can generate an event context based on contextual events and core events, and then display extended events through the event context. This provides users with additional information beyond the search terms, proactively uncovering users' related reading needs while satisfying their search requirements, thereby enhancing the user value of the search results page. Figure 11 As shown, in the above search scenario, the event context can be generated and the expanded event display can be achieved through the processes S701-S707, as follows:

[0181] S701, Identify core events.

[0182] In S701, upon receiving an instruction to generate an event timeline, the electronic device retrieves background search ranking data, such as search leaderboards, from the backend server of the instant messaging social network platform. The electronic device extracts trending events from the search leaderboards and event information from these trending events, constructing a timeline based on the event information. Figure 11 The event library is shown below. The events in the event library are those awaiting processing.

[0183] In S701, electronic devices can select the event to be processed corresponding to the core event from the event library as the core event according to the core event selection instruction.

[0184] S702, Get the core event title.

[0185] In S702, the electronic device obtains the core event title from the event information of the core event.

[0186] S703, Keyword Extraction.

[0187] In S703, the electronic device identifies core events from the event library and extracts keywords from the core events.

[0188] Here, the process in S703 is consistent with the process description in S102 above, and will not be repeated here.

[0189] S704, Keyword Recall.

[0190] In S704, the electronic device recalls candidate events containing keywords from the event database based on the keywords of the core events.

[0191] The process in S704 is the same as the process described in S103 above, and will not be repeated here.

[0192] S705, at least one dimension of scoring.

[0193] In S705, electronic devices can use event similarity, keyword IDF, DBSCAN clustering, historical highest popularity, and event occurrence time as scoring factors for at least one dimension. Based on the scoring factors for at least one dimension, the recalled candidate events are scored, and the comprehensive score corresponding to the candidate events is obtained based on the scores corresponding to at least one dimension.

[0194] It should be noted that the scoring process for electronic devices based on event similarity, keyword IDF, DBSCAN clustering, historical highest popularity, and event occurrence time does not distinguish the execution order.

[0195] S706, rating filter.

[0196] In S706, the electronic device filters candidate events based on their overall scores according to a preset scoring filtering strategy to obtain contextual events. For example, the electronic device can sort the candidate events and select the highest-ranked candidate event as the contextual event.

[0197] Here, event similarity is equivalent to the semantic similarity dimension mentioned above, keyword IDF is equivalent to the keyword dimension mentioned above, DBSCAN clustering is equivalent to the semantic clustering dimension mentioned above, and historical highest popularity is equivalent to the event popularity dimension mentioned above. The process of S704-S705 is consistent with the process of S104 mentioned above, and will not be repeated here.

[0198] S707, Show the sequence of events.

[0199] In S706, electronic devices can generate and display event contexts based on contextual events obtained from scoring filtering.

[0200] In some embodiments, Figure 11 The DBSCAN clustering process corresponding to S705 in the diagram can be illustrated as follows: Figure 12 As shown. Figure 12 Different points represent candidate events and core events, respectively, and the distance between points represents the semantic similarity between events. Region 120 represents the range of the DBSCAN clustering threshold. During the DBSCAN clustering process, electronic devices can iteratively expand the clustering range from high-density areas where events are clustered until no new candidate events are added within the region. Figure 12Region 121, centered on the core event, represents the range of contextual events determined based on the distance between candidate events and the core event, along with at least one scoring factor. Candidate events within region 122 are the contextual events selected through screening. In practical applications, the semantic difference between the core event and candidate events can be significant. This embodiment uses DBSCAN clustering to conditionally shift the semantics of the core event, while limiting the degree of shift using other scoring factors, thereby improving the accuracy of selecting contextual events from candidate events.

[0201] In some embodiments, the applicant conducted a comparative experiment on a purely manual selection scheme, a scheme that only performs DBSCAN semantic clustering screening, and a complete scheme based on five dimensions: semantic similarity, semantic clustering, keywords, event popularity, and event occurrence time. Using 84 event threads and a total of 1000 screening events, the accuracy, recall, and average manual operation time per thread were compared. The results are shown in Table 2 below:

[0202]

[0203] Table 2

[0204] As can be seen, the solution in this application embodiment can significantly reduce the time required for manual configuration and improve operational efficiency. Simultaneously, using keyword recall and scoring methods effectively improves the recall rate compared to manually searching the event database, thereby reducing the operational burden and allowing for greater focus on the screening process after candidate event recall. Furthermore, using features such as keyword IDF scores and event similarity scores can also significantly improve the accuracy of contextual events, reducing the operational review burden and increasing efficiency by 50%.

[0205] The following description continues to illustrate the exemplary structure of the extended event display device 255 provided in the embodiments of this application as a software module. In some embodiments, such as... Figure 2 As shown, the software modules stored in the extended event display device 255 of the memory 250 may include:

[0206] The extraction module 2551 is used to extract event information of hot events to obtain events to be processed, and to identify core events from the events to be processed; and to extract keywords from the core events to obtain the keywords corresponding to the core events.

[0207] Matching module 2552 is used to search and match based on the event to be processed using the keywords to obtain candidate events;

[0208] The filtering module 2553 is used to filter the candidate events in at least one dimension to obtain contextual events;

[0209] The display module 2554 is used to generate an event context based on the context events and the core events.

[0210] In some embodiments, the extraction module 2551 is further configured to perform named entity recognition on the core event title corresponding to the core event to obtain at least one named entity; among the at least one named entity, a named entity of a preset noun type is used as the keyword.

[0211] In some embodiments, the extended event display device 255 further includes a word weight analysis module; the word weight analysis module is used to perform word weight analysis on the core event title when the number of keywords is less than a preset number threshold, to obtain at least one weighted word; and to determine at least one verb-type weighted word as the keyword among the at least one weighted word.

[0212] In some embodiments, the at least one dimension includes at least one of a semantic dimension, a keyword dimension, an event popularity dimension, and an event occurrence time dimension; the filtering module 2553 is further configured to score the candidate event in at least one dimension to obtain a score for the candidate event in at least one dimension; obtain a comprehensive score based on the at least one dimension score; and filter the candidate events based on the comprehensive score to obtain contextual events.

[0213] In some embodiments, the semantic dimension includes a semantic similarity dimension; the at least one dimension score includes a semantic similarity score; the filtering module 2553 is further configured to use a preset semantic analysis model to perform semantic analysis on the core event title of the core event and the candidate event title of the candidate event to obtain the core semantic features corresponding to the core event and the candidate semantic features corresponding to the candidate event; obtain the semantic similarity between the core event and the candidate event by calculating the feature vector distance between the core semantic features and the candidate semantic features; and score the candidate event based on the semantic similarity to obtain the semantic similarity score.

[0214] In some embodiments, the semantic dimension includes a semantic clustering dimension; the at least one dimension score includes a semantic clustering score; the filtering module 2553 is further configured to perform density clustering on the core event and the candidate event based on the semantic similarity between the core event and the candidate event to obtain at least one event cluster; and to score the candidate event according to the event cluster in which the core event belongs to obtain the semantic clustering score of the candidate event.

[0215] In some embodiments, the at least one dimension score includes: keyword score; the filtering module 2553 is further configured to calculate the word frequency of the keyword through a preset corpus, and calculate the inverse document frequency corresponding to the keyword based on the word frequency; take the keywords contained in the candidate event title of the candidate event as internal keywords, and score the candidate event based on the inverse document frequency of the internal keywords to obtain the keyword score; the keyword score is inversely proportional to the inverse document frequency.

[0216] In some embodiments, the filtering module 2553 is further configured to adjust the weight of the inverse document frequency of the internal keywords by a preset coefficient to obtain the weighted inverse document frequency; and to use a preset frequency adjustment factor to sum and average the minimum values ​​of the weighted inverse document frequencies to obtain the keyword score.

[0217] In some embodiments, the event information includes: event popularity; the filtering module 2553 is further configured to determine the corresponding target popularity interval in at least one preset popularity interval based on the event popularity of the candidate event; and to obtain the popularity score corresponding to the candidate event based on the preset popularity coefficient corresponding to the target popularity interval.

[0218] In some embodiments, the event information includes: the event occurrence time; the at least one dimension score includes: the time dimension score; the filtering module 2553 is further configured to calculate the time difference between the event occurrence times of the candidate event and the context event; based on the time difference, determine a corresponding target time difference interval in at least one preset time difference interval, and use the preset score corresponding to the target time difference interval as the time score corresponding to the candidate event.

[0219] In some embodiments, the extraction module 2551 is further configured to obtain at least one search event with an event popularity higher than a preset popularity threshold from the background search data corresponding to the preset application as a hot event; extract the event title of each search event in the at least one search event, as well as at least one of the event popularity and the event occurrence time of each search event, and use the event popularity and the event occurrence time as the event information.

[0220] In some embodiments, the display module 2554 is further configured to sort the network events by time, generate an event network based on the sorting result and the core events, and display extended events through the event network.

[0221] It should be noted that the description of the above device embodiments is similar to the description of the above method embodiments, and has similar beneficial effects. For technical details not disclosed in the device embodiments of this application, please refer to the description of the method embodiments of this application for understanding.

[0222] This application provides a computer-readable storage medium storing executable instructions. When these executable instructions are executed by a processor, they cause the processor to perform the method provided in this application, for example... Figure 3 , 5 The methods shown in 7, 8, 9, and 10.

[0223] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0224] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0225] As an example, executable instructions may, but do not necessarily, correspond to files in the file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborative files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0226] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0227] In summary, this application's embodiments recall candidate events related to the core event from the events to be processed based on the keywords of the core event, and then filter the candidate events according to at least one dimension to obtain contextual events. This achieves noise reduction processing of candidate events to obtain contextual events, and combines core events and contextual events for extended event display. This not only reduces the computational load of directly generating event context through large-scale clustering and improves the efficiency of filtering contextual events, thereby improving the efficiency of extended event display; but also significantly improves the accuracy of the filtered contextual events through keyword matching and at least one dimension of filtering and noise reduction, achieving accurate event context construction under a large number of events, thereby improving the accuracy of extended event display.

[0228] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. An extended event presentation method, characterized by, include: By extracting event information from hot events, events to be processed are obtained, and core events are identified from these events. Keyword extraction is performed on the core events to obtain the keywords corresponding to the core events; Using the keywords, search and match based on the events to be processed to obtain candidate events; Based on the semantic similarity between the core events and the candidate events, density clustering is performed on the core events and the candidate events, iteratively expanding the clustering range starting from the high-density regions where the candidate events and core events cluster, to obtain at least one event cluster, until no new candidate events are added iteratively within the region. During the density clustering process, the semantics of the core events are conditionally shifted, and the degree of shift is limited by scoring factors in dimensions other than the semantic clustering dimension. Based on the event cluster to which the core event belongs, the candidate events are scored to obtain the semantic clustering score of the candidate events; the candidate events are scored according to the other dimensions to obtain at least one other dimension score corresponding to the candidate events. A comprehensive score is obtained based on the semantic clustering score and the score of at least one other dimension; Based on the comprehensive score, the candidate events are filtered to obtain contextual events; Extended events are displayed based on the contextual events and the core events.

2. The method of claim 1, wherein, The step of extracting keywords from the core event to obtain the keywords corresponding to the core event includes: Named entity recognition is performed on the core event titles corresponding to the core events to obtain at least one named entity; In the at least one named entity, a named entity of a preset noun type is used as the keyword.

3. The method of claim 2, wherein, The method further includes: If the number of keywords is less than a preset threshold, perform word weight analysis on the core event title to obtain at least one weighted word. Among the at least one weighted vocabulary, at least one weighted vocabulary of the verb type is determined as the keyword.

4. The method according to any one of claims 1 to 3, characterized in that, The at least one dimension includes at least one of the following: semantic dimension, keyword dimension, event popularity dimension, and event occurrence time dimension.

5. The method of claim 4, wherein, The semantic dimension includes: a semantic similarity dimension; the at least one dimension score includes: a semantic similarity score; The step of scoring the candidate event according to the other dimensions to obtain at least one other dimension score corresponding to the candidate event includes: Using a preset semantic analysis model, semantic analysis is performed on the core event title of the core event and the candidate event title of the candidate event to obtain the core semantic features corresponding to the core event and the candidate semantic features corresponding to the candidate event. The semantic similarity between the core event and the candidate event is obtained by calculating the feature vector distance between the core semantic features and the candidate semantic features. The candidate events are scored based on the semantic similarity to obtain the semantic similarity score.

6. The method according to claim 4, characterized in that, The semantic dimension includes a semantic clustering dimension; the at least one dimension score includes a semantic clustering score.

7. The method of claim 4, wherein, The at least one dimension score includes: keyword score; The step of scoring the candidate event according to the other dimensions to obtain at least one other dimension score corresponding to the candidate event includes: The word frequency of the keyword is calculated using a pre-set corpus, and the inverse document frequency corresponding to the keyword is calculated based on the word frequency. The keywords contained in the candidate event titles of the candidate events are used as internal keywords. The candidate events are scored based on the inverse document frequency of the internal keywords to obtain the keyword score. The keyword score is inversely proportional to the inverse document frequency. The step of scoring the candidate events based on the inverse document frequency to obtain the keyword score includes: The inverse document frequencies of the internal keywords are weighted by a preset coefficient to obtain the weighted inverse document frequencies. Using a preset frequency adjustment factor, the minimum value of the adjusted inverse document frequency is summed and averaged to obtain the keyword score.

8. The method of claim 4, wherein, The event information includes: event popularity; the at least one dimension score includes: popularity dimension score; The step of scoring the candidate event according to the other dimensions to obtain at least one other dimension score corresponding to the candidate event includes: Based on the event popularity of the candidate events, the corresponding target popularity interval is determined in at least one preset popularity interval; The popularity score corresponding to the candidate event is obtained based on the preset popularity coefficient corresponding to the target popularity range.

9. The method of claim 4, wherein, The event information includes: the time of the event; the at least one dimension score includes: the time dimension score; The step of scoring the candidate event according to the other dimensions to obtain at least one other dimension score corresponding to the candidate event includes: Calculate the time difference between the occurrence times of the candidate event and the core event; Based on the time difference, a corresponding target time difference interval is determined in at least one preset time difference interval, and the preset score corresponding to the target time difference interval is used as the time score corresponding to the candidate event.

10. The method of claim 1, wherein, The process of extracting event information from hot events to obtain events to be processed includes: From the background search data corresponding to the preset application, at least one search event with an event popularity higher than the preset popularity threshold is selected as a hot event; Extract the event title and content identifier of each search event from the at least one search event, as well as at least one of the event popularity and event occurrence time of each search event, and use the event popularity and at least one of the event occurrence time, the event title and the content identifier as the event information.

11. The method according to any one of claims 5-10, characterized in that, The extended event display based on the contextual events and the core events includes: The events in the timeline are sorted, and an event timeline is generated based on the sorting results and the core events. Extended events are then displayed through the event timeline.

12. An extended event presentation device, characterized by include: The extraction module is used to extract event information from hot events, obtain events to be processed, and identify core events from the events to be processed. The core events are then analyzed to extract keywords, which are then used to obtain the keywords corresponding to the core events. The matching module is used to search and match based on the keywords and the event to be processed to obtain candidate events; The filtering module is used to perform density clustering on the core events and candidate events based on their semantic similarity, iteratively expanding the clustering range from high-density regions where the candidate events and core events cluster together, to obtain at least one event cluster, until no new candidate events are added iteratively within the region. During the density clustering process, the semantics of the core events are conditionally shifted, and the degree of shift is limited by scoring factors in dimensions other than the semantic clustering dimension. Based on the event cluster to which the core event belongs, the candidate events are scored to obtain the semantic clustering score of the candidate events; the candidate events are scored according to the other dimensions to obtain at least one other dimension score corresponding to the candidate events. A comprehensive score is obtained based on the semantic clustering score and the score of at least one other dimension; Based on the comprehensive score, the candidate events are filtered to obtain contextual events; The display module is used to display extended events based on the context events and the core events.

13. The apparatus according to claim 12, characterized in that, The extraction module is also used to perform named entity recognition on the core event title corresponding to the core event to obtain at least one named entity; In the at least one named entity, a named entity of a preset noun type is used as the keyword.

14. The apparatus of claim 13, wherein, The device further includes: The word weight analysis module is used to perform word weight analysis on the core event title when the number of keywords is less than a preset threshold, and obtain at least one weighted word. Among the at least one weighted vocabulary, at least one weighted vocabulary of the verb type is determined as the keyword.

15. An electronic device, comprising: include: Memory, used to store executable instructions; A processor, when executing executable instructions stored in the memory, implements the method according to any one of claims 1 to 11.

16. A computer-readable storage medium, characterized in that, It stores executable instructions for implementing the method according to any one of claims 1 to 11 when executed by a processor.

17. A computer program product comprising computer programs or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the method described in any one of claims 1 to 11.

Images