Method and apparatus for pushing events, and computing device and medium

Abstract

The present disclosure relates to and provides a method and apparatus for pushing events, and a computing device, a computer-readable storage medium and a computer program product. The method comprises: on the basis of a target model, determining a category of relationship between a first event and each event in a set of candidate events, wherein the category of relationship at least comprises one of identical, similar and unrelated. The method further comprises: on the basis of the determined category of relationship, using the first event to update the set of candidate events. The method further comprises: on the basis of the updated set of candidate events, pushing a plurality of recommended events.

Description

Methods, apparatus, computing devices, and media for pushing events Technical Field

[0001] This disclosure relates to the field of computer technology, and more specifically, to a method, apparatus, computing device, computer-readable storage medium, and computer program product for pushing events. Background Technology

[0002] With the development of information technology, news push applications have become widespread, primarily targeting a large user base with a strong interest in news in specific fields. In the daily news push service process, several recent and highly influential news events are typically selected from a specific field, then integrated into a recommendation list and pushed to users to meet their news needs. However, selecting appropriate news events to push to users and improving the user experience of these applications presents a challenge for relevant technical personnel. Summary of the Invention

[0003] This disclosure provides a method, apparatus, computing device, computer-readable storage medium, and computer program product for pushing events. The technical solution of this disclosure is based on a target model to determine whether there are identical, similar, or unrelated relationships between events, providing a basis for event merging and deduplication, avoiding the pushing of identical or similar events to users, thereby improving user experience.

[0004] According to a first aspect of this disclosure, a method for pushing events is provided. The method includes determining a relationship category between a first event and each event in a candidate event set based on a target model, wherein the relationship category includes at least one of identical, similar, and unrelated. The method further includes updating the candidate event set with the first event based on the determined relationship category. The method also includes pushing multiple recommended events based on the updated candidate event set.

[0005] According to a second aspect of this disclosure, an apparatus for pushing events is provided. The apparatus includes a relationship classification unit configured to determine, based on a target model, a relationship category between a first event and each event in a candidate event set, wherein the relationship category includes at least one of identical, similar, and unrelated. The apparatus further includes an update unit configured to update the candidate event set with the first event based on the determined relationship category. The apparatus further includes a push unit configured to push multiple recommended events based on the updated candidate event set.

[0006] According to a third aspect of this disclosure, a computing device is provided, comprising: at least one processing unit; at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, causing the computing device to perform the method as described in the first aspect of this disclosure.

[0007] According to a fourth aspect of this disclosure, a non-transient computer storage medium is provided, including machine-executable instructions that, when executed by a device, cause the device to perform the method as described in the first aspect of this disclosure.

[0008] According to a fifth aspect of this disclosure, a computer program product is provided, including machine-executable instructions that, when executed by a device, cause the device to perform the method as described in the first aspect of this disclosure.

[0009] It should be understood that the summary section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0010] The above and other objects, features, and advantages of embodiments of the present disclosure will become more readily understood from the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the present disclosure will be described by way of example and non-limitation, wherein:

[0011] Figure 1 shows a schematic diagram of an event push system according to an embodiment of the present disclosure;

[0012] Figure 2 shows a schematic flowchart of a method for pushing events according to an embodiment of the present disclosure;

[0013] Figure 3 illustrates a schematic diagram of a process for fine-tuning a language model according to an embodiment of the present disclosure;

[0014] Figure 4 illustrates a schematic flowchart of a process for constructing a training dataset for a language model according to an embodiment of the present disclosure.

[0015] Figure 5 shows a block diagram of an apparatus for pushing events according to an embodiment of the present disclosure; and

[0016] Figure 6 shows a block diagram of an electronic device according to an embodiment of the present disclosure.

[0017] In all the accompanying figures, the same or similar reference numerals denote the same or similar elements. Detailed Implementation

[0018] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0019] In the description of embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects unless explicitly stated. Other explicit and implicit definitions may also be included below.

[0020] As used in this paper, the term "model" refers to a model that learns the relationship between inputs and outputs from training data, enabling it to generate corresponding outputs for a given input after training. Model generation can be based on machine learning techniques. Deep learning is a machine learning algorithm that processes inputs and provides corresponding outputs using multiple layers of processing units. A neural network model is an example of a deep learning-based model. In this paper, "model" may also be referred to as a "machine learning model," "learning model," "machine learning network," or "learning network," and these terms are used interchangeably.

[0021] A neural network is a machine learning network based on deep learning. A neural network processes input and provides a corresponding output, typically consisting of an input layer, an output layer, and one or more hidden layers between the input and output layers. Neural networks used in deep learning applications often include many hidden layers, thus increasing the network's depth. The layers of a neural network are connected sequentially, so that the output of the previous layer is provided as the input to the next layer. The input layer receives the input to the neural network, while the output layer's output serves as the final output. Each layer of a neural network includes one or more nodes (also called processing nodes or neurons), each node processing the input from the layer above.

[0022] Machine learning typically comprises three phases: training, testing, and usage (also known as inference). In the training phase, a given model is trained using a large amount of training data, iteratively updating parameter values ​​until the model can consistently generate inferences that meet the expected goals from the training data. Through training, the model can be considered to have learned the relationship between inputs and outputs (also known as the input-output mapping) from the training data. The parameter values ​​of the trained model are determined. In the testing phase, test inputs are applied to the trained model to test whether it can provide the correct output, thus determining the model's performance. In the usage phase, the model can be used to process actual inputs based on the trained parameter values ​​to determine the corresponding output.

[0023] Most existing news push applications use news reports as a medium, pushing news events based on the popularity of the reports and topics of interest to users. For example, some news push applications select multiple latest and highly influential news events in the user's area of ​​interest every day and push them to the user in the form of a trending list or recommended news cards. This push method is prone to the problem that multiple recommended events actually tell the same or repetitive stories. From the server's perspective, storing multiple news reports of the same event wastes storage resources; from the user experience perspective, pushing multiple identical news events to the user at once, or repeatedly pushing the same or similar news events in a short period of time, reduces the efficiency of users obtaining information, and also makes users feel annoyed, affecting the user experience and reducing user stickiness.

[0024] In view of this, embodiments of this disclosure provide a technical solution that uses a model to analyze the relationships between events and pushes events based on these relationships. Overall, the technical solution of this disclosure abstracts original content into events, uses a model to determine whether two events have the same, similar, or unrelated relationship, providing a basis for event merging, deduplication, etc. For example, it avoids pushing content of the same event in a single push, and / or avoids pushing content of similar events within a short period. Based on this approach, the storage requirements on the server are reduced, and users can be provided with content that does not feel repetitive, improving the efficiency of information retrieval and the user experience.

[0025] The exemplary embodiments of this disclosure will be described in detail below with reference to Figures 1 to 6. It should be noted that although the embodiments of this disclosure will be mainly described in conjunction with news event push applications, those skilled in the art should understand that the embodiments of this disclosure are also applicable to scenarios that push any other type of event or text content, and this disclosure does not impose any limitations.

[0026] Figure 1 illustrates a schematic diagram of an event push system 100 according to an embodiment of the present disclosure. In system 100, an event push service 110 receives a document 101 and pushes events to a terminal 120 based on the received document 101. The document 101 may be a set of news reports, which may include text, images, audio, or other audio formats. The events pushed to the terminal 120 may be the corresponding original document 101 or a modified document. The event push service 110 may periodically push events to the terminal 120, and each push may include multiple events, such as current trending news events.

[0027] As shown in the figure, the event push service 110 includes an event extraction module 102 for extracting events from document 101. The news reports represented by document 101 often vary in length and style, and may include background information, commentary, and other elements in addition to details of the news events. Directly storing the original document may not accurately identify the news events and could be influenced by the narrative style, leading to misunderstandings. The event extraction module 102 can convert the various styles of news reports represented by document 101 into a unified format, improving the accuracy of the model's understanding of news events. In some implementations, a pre-trained Large Language Model (LLM) can be used to extract the original news reports, abstracting them into three parts: [title], [intro], and [fact] by writing prompts. In other words, an event can be represented as or include a title, introduction, and facts. Events can be stored in a candidate event set 104. The candidate event set 104 includes events that have undergone deduplication.

[0028] The event push service 110 also includes an event relationship classification module 106 for identifying relationships between events. For example, the event relationship classification module 106 can identify relationships between events output by the event extraction module 102 and the candidate event set 104, and update the candidate event set 104 based on the identified relationships. In some embodiments, the transaction relationship classification model 106 can use a target model 108 to identify relationships between events. For example, a language model can be implemented as a classifier that receives text pairs representing two events as input and outputs a classification result for the text pairs. Relationships between events can be stored in association with the candidate event set 104.

[0029] The event push service 110 also includes an event selection and push module 105, which selects events from the candidate event set 104 and pushes them to the terminal 120. The terminal 120 can subscribe to one or more topics or categories of interest to the event push service 110. The event selection and push module 105 can select events from the corresponding topics or categories based on the subscription information, sort them according to popularity, and then push these top-ranked events to the terminal 120. The event selection and push module 105 can also select events based on the relationships between events in the candidate event set 104 to avoid a single push including the same event or similar events being pushed frequently.

[0030] The embodiments of this disclosure abstract the relationships between events into "same", "similar", and "unrelated". The definitions and uses of these relationships are as follows.

[0031] Same: If Event 1 and Event 2 describe the same main event, but differ slightly in their textual expression, then Event 1 and Event 2 can be considered to be "same". If the relationship between two events is identified as "same", then the two events can be merged, or either one can be deleted, to reduce storage space.

[0032] Similarity: Also known as "repetition," if Event 1 and Event 2 describe different but similar main events, then Event 1 and Event 2 are considered to have a "similar" relationship. Users may experience a sense of repetition when reading related news events. Two events with a "similar" relationship should not be pushed out in the same time or on the same day to reduce the user's repetitive experience.

[0033] Irrelevant: If the main events described by Event 1 and Event 2 are completely different (e.g., the people, places, times, and facts are obviously different), then Event 1 and Event 2 are considered to be "irrelevant". Two events with an "irrelevant" relationship can be pushed out on the same day, increasing the diversity of event push notifications.

[0034] Figure 2 shows a schematic flowchart of a method 200 for pushing events according to an embodiment of the present disclosure. In some embodiments, method 200 may be implemented by, for example, the event push service 110 shown in Figure 1. It should be understood that method 200 may also include additional actions not shown and / or the actions shown may be omitted, and the scope of the present disclosure is not limited in this respect.

[0035] Method 200 includes: in block 210, determining a relationship category between a first event and each event in a candidate event set, based on a target model, wherein the relationship category includes at least one of identical, similar, and unrelated. The first event may be an event output by event extraction module 102 that has not yet been stored in candidate event set 104. The target model may be a trained or fine-tuned model adapted to identify the relationship category between events. The target model may be a language model or a model of other modalities. For ease of understanding, exemplary embodiments of this disclosure are described below using a language model as an example.

[0036] The candidate event set may include events extracted from a collected set of documents. In some embodiments, method 200 may include obtaining a set of documents, such as collecting recent trending news reports, then generating an event corresponding to each document in the document set, the event including a title, summary, and facts, and constructing a candidate event set based on the generated events. The events in the candidate event set do not include the same events.

[0037] In some embodiments, the Large Language Model (LLM) can be fine-tuned to enable it to identify categories of relationships between events. Fine-tuning involves adjusting the model based on a pre-trained LLM using a task-specific dataset to adapt it for that task.

[0038] Figure 3 illustrates a schematic diagram of the process for fine-tuning a language model according to an embodiment of the present disclosure. In Figure 3, an LLM 302 is pre-trained in an unsupervised manner using a large corpus 301, where the base model of LLM 302 can be a transformer-based model. This results in a pre-trained LLM 304. The pre-trained LLM 304 has learned rich linguistic knowledge and semantic understanding capabilities, such as vocabulary, grammar, and semantic relationships, on the large corpus 301. Through fine-tuning, the model can adjust its parameters based on supervised training data 303 for a specific task, thereby learning specific patterns and knowledge related to that task, resulting in a final model 305. The supervised training data 303 includes events and annotation information between events, where the annotation indicates whether the relationship between two events is the same, similar, or unrelated. In step 210 shown in Figure 2, the final model 305 obtained in this way can be used to predict the relationship category between a first event and other events in the candidate event set.

[0039] Given the vast number of diverse news events generated daily in the real world, for a given news event, there are far more news events "unrelated" to it than those "same" or "similar" to it. Therefore, constructing a balanced training dataset by extracting representative news events from this large pool and collecting the relationships between them is a worthwhile research topic.

[0040] Figure 4 shows a schematic flowchart of a process 400 for constructing a training dataset for a language model according to an embodiment of the present disclosure. The training dataset obtained through process 400 is used as supervised training data 303 to fine-tune the pre-trained LLM 304.

[0041] In box 401, generate multiple seed events. You can prepare a batch of news data and extract it into events. For each topic (such as "technology", "entertainment", etc.), select several (e.g., 5) events as seed events to ensure the diversity and representativeness of the seed events.

[0042] In box 402, for a seed event, retrieve identical and similar events from the event set database. In some embodiments, identical and similar events can be retrieved based on keywords or semantic features of the seed event. For example, for each seed event, recall the top 50 events from the event set using embeddings and / or keywords, and further identify identical and similar events among them, such that each seed event has 10 identical events and several similar events.

[0043] In box 403, determine if there are a sufficient number of identical events. If not, identical events of the seed event can be generated by rewriting the seed event. In box 404, events can be rewritten based on the LLM to generate more identical events. If there are already a sufficient number of identical events, the process proceeds to box 405.

[0044] In box 405, retrieve irrelevant events for the seed event. In some embodiments, events with dates different from the seed event can be retrieved from the event set as irrelevant events for the seed event. Alternatively, all seed events and identical events can be placed in a separate set, and events that do not overlap with the seed event in time can be randomly selected as irrelevant events for the seed event.

[0045] In box 406, events are paired to generate training data. The training dataset is generated by pairing seed events with identical events, similar events, and unrelated events.

[0046] In box 407, it is determined whether all seed events have been traversed. If not, boxes 402 to 406 above are repeated to generate training data based on the next seed event. If all events have been traversed, the training dataset is output in box 408. When fine-tuning the language model, the event relationship judgment is transformed into a text pair classification task. In some embodiments, the model can be fine-tuned using prompt words to enable the model to learn text classification knowledge. Exemplary prompt words are as follows. The language model obtained by fine-tuning in the above manner can be used to classify the relationship between two events.

[0047] Returning to Figure 2, in box 220, the candidate event set is updated with the first event based on the determined relationship category. As mentioned above, the candidate event set does not include identical events, so if the relationship between the first event and an event in the candidate event set is determined to be the same, the first event cannot be directly added to the candidate event set. In some embodiments, in response to determining that the relationship type between the first event and a second event in the candidate event set is the same, the candidate event set is updated by merging the first event and the second event. Merging can be any of the following operations: retaining the second event or discarding the first event, replacing the second event with the first event, or generating a new event based on the first and second events to replace the second event in the candidate event set.

[0048] In some embodiments, in response to determining that the relationship category between the first event and each event in the candidate event set is similar or unrelated, the first event is added to the candidate event set, and the relationship category between the first event and other events is recorded. In this way, the candidate event set does not include identical events, and the relationship information between events is also stored for subsequent push notifications.

[0049] In box 230, multiple recommended events are pushed based on the updated candidate event set. In some embodiments, to avoid repetition, recommended events can be selected from the updated candidate event set so that two or more events with similar relationship categories are not included in a single push or a push within a predetermined time period. Furthermore, since the candidate event set does not include identical events, the push does not include duplicate events. This approach improves the diversity of news pushes and enhances the user experience.

[0050] Exemplary embodiments of this disclosure have been described above with reference to Figures 1 to 4. Compared to existing event push methods, the technical solution of this disclosure uses a model to determine whether two events have the same, similar, or unrelated relationship, providing a basis for event merging, deduplication, etc. For example, a single push may not push content of the same event, and / or content of similar events may not be pushed within a short period of time. Based on this approach, the storage requirements of the server are reduced, and users can be provided with content that does not feel repetitive, improving the efficiency of users obtaining information and the user experience.

[0051] Figure 5 shows a schematic block diagram of an apparatus 500 for event push according to an embodiment of the present disclosure. As shown in Figure 5, the apparatus 500 includes a relationship classification unit 502, an update unit 504, and a push unit 506. The relationship classification unit 502 is configured to determine a relationship category between a first event and each event in a candidate event set based on a target model, wherein the relationship category includes at least one of being the same, similar, or unrelated. The update unit 504 is configured to update the candidate event set with the first event based on the determined relationship category. The push unit 506 is configured to push multiple recommended events based on the updated candidate event set.

[0052] In some embodiments, the update unit 504 may also be configured to update the candidate event set by merging the first event and the second event in the candidate event set in response to determining that the relationship type between the first event and the second event in the candidate event set is the same.

[0053] In some embodiments, the update unit 504 may also be configured to: in response to determining that the relationship category between the first event and each event in the candidate event set is similar or unrelated, add the first event to the candidate event set and record the relationship category between the first event and other events.

[0054] In some embodiments, the update unit 504 may also be configured to: select recommended events from the updated candidate event set, such that two or more events with the same or similar relationship categories are not included in a single push or a push within a predetermined time period.

[0055] In some embodiments, the target model is fine-tuned using a training dataset, and the apparatus further includes a training data acquisition unit configured to: acquire a training dataset; generate a plurality of seed events; for each seed event, retrieve identical and similar events of the seed event in an event set; retrieve events with dates different from those of the seed events in the event set as irrelevant events of the seed events; and generate a training dataset by pairing the seed events with identical, similar, and irrelevant events.

[0056] In some embodiments, the training data acquisition unit may also be configured to generate the same event as the seed event by rewriting the seed event.

[0057] In some embodiments, the training data acquisition unit may also be configured to retrieve identical and similar events of the seed event based on the keywords or semantic features of the seed event.

[0058] In some embodiments, the apparatus 500 may further include an event extraction unit configured to: acquire a document set; generate an event corresponding to each document in the document set, the event including a title, a summary, and facts; and construct a candidate event set based on the generated events.

[0059] It should be noted that further actions or steps shown in Figures 1 to 4 can be implemented using the device 500 shown in Figure 5. For example, device 500 may include more modules or units to implement the actions or steps described above, or some of the units or modules shown in Figure 5 may be further configured to implement the actions or steps described above. This will not be repeated here.

[0060] Figure 6 shows a schematic block diagram of an example device 600 that can be used to implement embodiments of the present disclosure. As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) 602 or loaded from storage unit 606 into random access memory (RAM) 603. Various programs and data required for the operation of device 600 may also be stored in RAM 603. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0061] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0062] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as method 200. For example, in some embodiments, method 200 may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of method 200 described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform method 200 by any other suitable means (e.g., by means of firmware).

[0063] In some embodiments, the methods and processes described above can be implemented as a computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of this disclosure.

[0064] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0065] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper cables, fiber optic cables, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to computer-readable storage media within the respective computing / processing device.

[0066] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​and conventional procedural programming languages. The computer-readable program instructions may execute entirely on a user's computer, partially on a user's computer, as a standalone software package, partially on a user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving 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). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0067] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processing unit of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0068] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0069] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0070] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

[0071] The following are some example implementations of this disclosure.

[0072] Example 1. A method for pushing events, comprising:

[0073] Based on the target model, determine the relationship category between the first event and each event in the candidate event set, wherein the relationship category includes at least one of the following: same, similar, and unrelated.

[0074] Based on the determined relationship category, the candidate event set is updated with the first event; and

[0075] Multiple recommended events are pushed based on the updated set of candidate events.

[0076] Example 2. According to the method of Example 1, wherein updating the candidate event set with the first event based on the determined relationship category includes:

[0077] In response to determining that the relationship type between the first event and the second event in the candidate event set is the same, the candidate event set is updated by merging the first event and the second event.

[0078] Example 3. According to the method of Example 1, wherein updating the candidate event set with the first event based on the determined relationship category includes:

[0079] In response to determining that the relationship category between the first event and each event in the candidate event set is similar or unrelated, the first event is added to the candidate event set, and the relationship category between the first event and the other events is recorded.

[0080] Example 4. The method according to any one of Examples 1 to 3, wherein pushing multiple recommendation events based on the updated candidate event set includes:

[0081] The recommended event is selected from the updated set of candidate events, such that two or more events with the same or similar relationship category are not included in a single push or a push within a predetermined time period.

[0082] Example 5: According to any one of Examples 1 to 4, the target model is fine-tuned using a training dataset, and the method further includes obtaining the training dataset by:

[0083] Generate multiple seed events;

[0084] For each seed event, retrieve the same and similar events of the seed event from the event set;

[0085] Retrieve events from the event set whose dates differ from the seed event date, and treat them as irrelevant events to the seed event; and

[0086] The training dataset is generated by pairing the seed events with the same events, similar events, and unrelated events.

[0087] Example 6. According to the method described in Example 5, the method further includes:

[0088] The same event as the seed event is generated by rewriting the seed event.

[0089] Example 7: According to the method described in Example 5, retrieving identical and similar events of the seed event includes:

[0090] Retrieve identical and similar events based on the keywords or semantic features of the seed event.

[0091] Example 8. The method according to any one of Examples 1 to 7 further includes:

[0092] Get a collection of documents; and

[0093] For each document in the document collection, an event corresponding to that document is generated, the event including a title, summary, and facts; and

[0094] Based on the generated events, the candidate event set is constructed.

[0095] Example 9. An apparatus for pushing events, comprising:

[0096] The relation classification unit is configured to determine, based on the target model, the relation category between the first event and each event in the candidate event set, wherein the relation category includes at least one of being the same, similar, and unrelated.

[0097] An update unit is configured to update the candidate event set with the first event based on the determined relationship category; and

[0098] The push unit is configured to push multiple recommended events based on the updated set of candidate events.

[0099] Example 10. The apparatus according to Example 9, wherein the updating unit comprises:

[0100] The merging unit is configured to update the candidate event set by merging the first event and the second event in the candidate event set in response to determining that the relationship type between the first event and the second event in the candidate event set is the same.

[0101] Example 11. The apparatus according to Example 8, wherein the updating unit comprises:

[0102] In response to determining that the relationship category between the first event and each event in the candidate event set is similar or unrelated, the first event is added to the candidate event set, and the relationship category between the first event and the other events is recorded.

[0103] Example 12, the apparatus according to any one of Examples 9 to 11, wherein the pushing unit comprises:

[0104] The recommended event is selected from the updated set of candidate events, such that two or more events with the same or similar relationship category are not included in a single push or a push within a predetermined time period.

[0105] Example 13. The apparatus according to any one of Examples 9 to 12, wherein the target model is fine-tuned using a training dataset, and the method training data generation unit is configured to:

[0106] Generate multiple seed events;

[0107] For each seed event, retrieve the same and similar events of the seed event from the event set;

[0108] Retrieve events from the event set whose dates differ from the seed event date, and treat them as irrelevant events to the seed event; and

[0109] The training dataset is generated by pairing the seed events with the same events, similar events, and unrelated events.

[0110] Example 14. According to the apparatus described in Example 13, the training data unit includes:

[0111] The rewriting unit is configured to generate the same event as the seed event by rewriting the seed event.

[0112] Example 15, according to the method of Example 13, wherein the training data unit comprises:

[0113] The retrieval unit is configured to retrieve identical and similar events based on the keywords or semantic features of the seed event.

[0114] Example 16. The apparatus according to any one of Examples 9 to 15 further includes an event extraction unit configured to:

[0115] Get a collection of documents; and

[0116] For each document in the document collection, an event corresponding to that document is generated, the event including a title, summary, and facts; and

[0117] Based on the generated events, the candidate event set is constructed.

[0118] Example 17. A computing device, comprising:

[0119] At least one processing unit;

[0120] At least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, causing the computing device to perform an action, the action including:

[0121] Based on the target model, determine the relationship category between the first event and each event in the candidate event set, wherein the relationship category includes at least one of the following: same, similar, and unrelated.

[0122] Based on the determined relationship category, the candidate event set is updated with the first event; and

[0123] Multiple recommended events are pushed based on the updated set of candidate events.

[0124] Example 18. The computing device according to Example 17, wherein updating the candidate event set with the first event based on the determined relationship category includes:

[0125] In response to determining that the relationship type between the first event and the second event in the candidate event set is the same, the candidate event set is updated by merging the first event and the second event.

[0126] Example 19. The computing device according to Example 17, wherein updating the candidate event set with the first event based on the determined relation category includes:

[0127] In response to determining that the relationship category between the first event and each event in the candidate event set is similar or unrelated, the first event is added to the candidate event set, and the relationship category between the first event and the other events is recorded.

[0128] Example 20: A computing device according to any one of Examples 17 to 19, wherein pushing multiple recommendation events based on the updated set of candidate events includes:

[0129] The recommended event is selected from the updated set of candidate events, such that two or more events with the same or similar relationship category are not included in a single push or a push within a predetermined time period.

[0130] Example 21: A computing device according to any one of Examples 17 to 20, wherein the target model is fine-tuned using a training dataset, the action further comprising acquiring the training dataset by:

[0131] Generate multiple seed events;

[0132] For each seed event, retrieve the same and similar events of the seed event from the event set;

[0133] Retrieve events from the event set whose dates differ from the seed event date, and treat them as irrelevant events to the seed event; and

[0134] The training dataset is generated by pairing the seed events with the same events, similar events, and unrelated events.

[0135] Example 22, the computing device according to Example 21, wherein the action further includes:

[0136] The same event as the seed event is generated by rewriting the seed event.

[0137] Example 23, according to the method described in Example 21, wherein retrieving identical and similar events of the seed event includes:

[0138] Retrieve identical and similar events based on the keywords or semantic features of the seed event.

[0139] Example 24. The computing device according to any one of Examples 17 to 23, wherein the action further includes:

[0140] Get a collection of documents; and

[0141] For each document in the document collection, an event corresponding to that document is generated, the event including a title, summary, and facts; and

[0142] Based on the generated events, the candidate event set is constructed.

[0143] Example 25: A computer storage medium including machine-executable instructions that, when executed by a device, cause the device to perform the method described in any one of Examples 1 to 8.

[0144] Example 26: A computer program product including machine-executable instructions that, when executed by a device, cause the device to perform the method described in any one of Examples 1 to 8.

[0145] Although this disclosure has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for pushing events, comprising: Based on the target model, determine the relationship category between the first event and each event in the candidate event set, wherein the relationship category includes at least one of the following: same, similar, and unrelated. Based on the determined relationship category, the candidate event set is updated with the first event; and Multiple recommended events are pushed based on the updated set of candidate events.

2. The method according to claim 1, wherein, Based on the determined relationship category, updating the candidate event set with the first event includes: In response to determining that the relationship type between the first event and the second event in the candidate event set is the same, the candidate event set is updated by merging the first event and the second event.

3. The method according to claim 1, wherein, Based on the determined relationship category, updating the candidate event set with the first event includes: In response to determining that the relationship category between the first event and each event in the candidate event set is similar or unrelated, the first event is added to the candidate event set, and the relationship category between the first event and the other events is recorded.

4. The method according to any one of claims 1 to 3, wherein, Pushing multiple recommendation events based on the updated candidate event set includes: The recommended event is selected from the updated set of candidate events, such that two or more events with the same or similar relationship category are not included in a single push or a push within a predetermined time period.

5. The method according to any one of claims 1 to 4, wherein the target model is fine-tuned using a training dataset, and the method further comprises obtaining the training dataset by: Generate multiple seed events; For each seed event, retrieve the same and similar events of the seed event from the event set; Retrieve events from the event set that have a different date than the seed event, and treat them as irrelevant events to the seed event; as well as The training dataset is generated by pairing the seed events with the same events, similar events, and unrelated events.

6. The method according to claim 5, further comprising: The same event as the seed event is generated by rewriting the seed event.

7. The method according to claim 5, wherein, Retrieving identical and similar events to the seed event includes: Retrieve identical and similar events based on the keywords or semantic features of the seed event.

8. The method according to any one of claims 1 to 7, further comprising: Get the document collection; as well as For each document in the document collection, an event corresponding to that document is generated, the event including a title, a summary, and facts; as well as Based on the generated events, the candidate event set is constructed.

9. An apparatus for pushing events, comprising: The relation classification unit is configured to determine, based on the target model, the relation category between the first event and each event in the candidate event set, wherein the relation category includes at least one of being the same, similar, and unrelated. An update unit is configured to update the candidate event set with the first event based on the determined relationship category; as well as The push unit is configured to push multiple recommended events based on the updated set of candidate events.

10. A computing device, comprising: At least one processing unit; At least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, causing the computing device to perform the method as described in any one of claims 1 to 8.

11. A computer storage medium comprising machine-executable instructions that, when executed by a device, cause the device to perform the method as claimed in any one of claims 1 to 8.

12. A computer program product comprising machine-executable instructions that, when executed by a device, cause the device to perform the method as described in any one of claims 1 to 8.

Images