CN113901226A

CN113901226A - Real-time live broadcast data processing method and computer storage medium

Info

Publication number: CN113901226A
Application number: CN202111488640.8A
Authority: CN
Inventors: 刘士博
Original assignee: Alibaba Damo Institute Hangzhou Technology Co Ltd
Current assignee: Hangzhou Alibaba Cloud Feitian Information Technology Co ltd
Priority date: 2021-12-08
Filing date: 2021-12-08
Publication date: 2022-01-07
Anticipated expiration: 2041-12-08
Also published as: CN113901226B

Abstract

The embodiment of the application provides a real-time live broadcast data processing method and a computer storage medium, wherein the real-time live broadcast data processing method comprises the following steps: obtaining comment messages sent by a user in a live content watching process in real time; semantic analysis is carried out on the comment message within a preset time range, and a user intention corresponding to the comment message is obtained according to the result of the semantic analysis; clustering the comment messages according to the user intention to obtain at least one intention clustering result; and displaying the intention clustering result on a live broadcast information interface of the live broadcast content. Through the embodiment of the application, the requirements of the live users can be timely and effectively met, the experience of the live users is improved, and the stickiness of the live users is increased.

Description

Real-time live broadcast data processing method and computer storage medium

Technical Field

The embodiment of the application relates to the technical field of computers, in particular to a real-time live broadcast data processing method and a computer storage medium.

Background

As the demand for user interaction increases, more and more applications, especially multimedia applications, employ real-time interaction. In this manner, multimedia audience users (e.g., audience users of a live video or audience users of a live audio, etc.) may interact with the multimedia anchor in real-time by commenting on messages.

Taking live video as an example, in a general live video process, the number of online people (possibly thousands of people) in a live broadcast room where many anchor broadcasts are located is very large, and tens of hundreds of or even more comment messages of audience users are generated every second. For the anchor, comment messages are an effective way to understand the viewer's user's thoughts and needs for their play content, play behavior, etc. However, for a large number of comment messages, on one hand, a large amount of labor and time costs are required to find out the core appeal and idea of the audience users; on the other hand, many complaints and ideas are time-efficient, and after a period of time, the complaints of the audience users cannot be timely and effectively satisfied, the experience of the audience users is greatly reduced, and the user stickiness is further reduced.

Disclosure of Invention

In view of the above, embodiments of the present application provide a real-time live data processing scheme to at least partially solve the above problem.

According to a first aspect of an embodiment of the present application, a method for processing live data in real time is provided, including: obtaining comment messages sent by a user in a live content watching process in real time; semantic analysis is carried out on the comment message within a preset time range, and a user intention corresponding to the comment message is obtained according to the result of the semantic analysis; clustering the comment messages according to the user intention to obtain at least one intention clustering result; and displaying the intention clustering result through a live broadcast information interface of the live broadcast content.

According to a second aspect of the embodiments of the present application, there is provided another method for processing live broadcast data, including: displaying live broadcast content and comment information input by a user in the process of watching the live broadcast content on a live broadcast interface; and when the comment intention prompt condition is determined to be met, displaying a user intention prompt message generated according to the comment message on the live broadcast interface.

According to a third aspect of embodiments of the present application, there is provided an electronic apparatus, including: a receiver, a transmitter, and a processor; wherein: the receiver is used for receiving comment messages input by a user in the process of watching the live broadcast content and sending the comment messages to the processor; the processor is used for acquiring the comment message in real time; semantic analysis is carried out on the comment message within a preset time range, and a user intention corresponding to the comment message is obtained according to the result of the semantic analysis; clustering the comment messages according to the user intention to obtain at least one intention clustering result; and the transmitter is used for transmitting the intention clustering result to user equipment so as to display the intention clustering result on a live broadcast interface of live broadcast content through the user equipment.

According to a fourth aspect of embodiments of the present application, there is provided another electronic apparatus, including: an input device, a display and a processor; wherein: the input device is used for receiving comment messages input by a user in the process of watching the live broadcast content and sending the comment messages to the processor; the processor is used for displaying live broadcast content and the comment message on a live broadcast interface; when the comment intention prompting condition is determined to be met, acquiring a user intention prompting message generated according to the comment message; the display is used for displaying the user intention prompt message on the live broadcast interface.

According to a fifth aspect of embodiments of the present application, there is provided a computer storage medium having stored thereon a computer program which, when executed by a processor, implements a live data processing method as described in the first or second aspect.

According to a sixth aspect of embodiments of the present application, there is provided a computer program product, which includes computer instructions for instructing a computing device to execute operations corresponding to the live broadcast data processing method according to the first aspect or the second aspect.

According to the real-time live broadcast data processing scheme provided by the embodiment of the application, comment messages published by a user in the process of watching live broadcast contents are acquired in real time, and then semantic analysis is carried out, so that the user intention which the user wants to express based on the comment messages is acquired. Because there may be many users watching live content at the same time, intentions which are desired to be expressed by comment messages issued by a plurality of users may be the same or similar, and based on the intentions, clustering can be performed based on the obtained user intentions to obtain one or more intention clustering results, and then the intention clustering results are displayed through a live information interface. Therefore, on one hand, core appeal and idea of the user can be known through semantic analysis and subsequent intention clustering of the comment message, manual processing is not needed, and labor and time cost is saved; on the other hand, due to the fact that the comment messages are obtained in real time and corresponding analysis and clustering are carried out in time, timeliness of the user complaint and the transmission of the idea can be effectively guaranteed, user requirements are met in time and effectively, user experience and anchor experience are improved, and user and anchor stickiness is increased.

And on the user intention display layer, when the comment intention prompt condition is met, the user intention prompt message generated according to the comment message of the user can be displayed on the live broadcast interface. Therefore, the message acquisition end can timely know the user intention capable of expressing the core appeal and idea of the user such as the live anchor and the like and timely process the user intention so as to timely and effectively meet the user requirements, improve the anchor experience of the user and increase the stickiness of the user and the anchor.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the embodiments of the present application, and other drawings can be obtained by those skilled in the art according to the drawings.

Fig. 1 is a schematic diagram of an exemplary system to which a real-time live data processing method according to an embodiment of the present application is applied;

fig. 2A is a flowchart illustrating steps of a method for processing live data according to a first embodiment of the present application;

FIG. 2B is a diagram illustrating an example of a scenario in the embodiment shown in FIG. 2A;

fig. 3A is a flowchart illustrating steps of a method for processing live broadcast data according to a second embodiment of the present application;

FIG. 3B is a diagram of a live information interface in the embodiment shown in FIG. 3A;

FIG. 3C is a diagram of another live information interface in the embodiment shown in FIG. 3A;

fig. 4A is a flowchart illustrating steps of a method for processing live broadcast data according to a third embodiment of the present application;

FIG. 4B is a schematic diagram of a process in the embodiment of FIG. 4A;

fig. 5 is a flowchart illustrating steps of a real-time live data processing method according to a fourth embodiment of the present application;

fig. 6A is a flowchart illustrating steps of a method for processing live broadcast data according to a fifth embodiment of the present application;

FIG. 6B is a diagram illustrating an example of a scenario in the embodiment shown in FIG. 6A;

fig. 7 is a schematic structural diagram of an electronic device according to a sixth embodiment of the present application;

fig. 8 is a schematic structural diagram of another electronic device according to a seventh embodiment of the present application.

Detailed Description

In order to make those skilled in the art better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, but not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application shall fall within the scope of the protection of the embodiments in the present application.

The following further describes specific implementations of embodiments of the present application with reference to the drawings of the embodiments of the present application.

Fig. 1 illustrates an exemplary system to which a real-time live data processing method according to an embodiment of the present application is applied. As shown in fig. 1, the system 100 may include a server 102, a communication network 104, and/or one or more user devices 106, illustrated in fig. 1 as a plurality of user devices. In a live application, the plurality of user devices 106 may include both a user device 1062 at the anchor end and a user device 1064 at the audience end. Although two user devices are shown in fig. 1 at the same time, it should be understood by those skilled in the art that, although only two user devices 1062 at the anchor end and two user devices 1064 at the audience end are shown in fig. 1, the numbers are merely illustrative, and in practical applications, the specific numbers of the user devices 1062 at the anchor end and the user devices 1064 at the audience end can be flexibly set, which is not limited by the embodiment of the present application. In various embodiments of the present application, the scenario is mainly described in the perspective of the user equipment 1062 at the anchor end.

Server 102 may be any suitable server for storing information, data, programs, and/or any other suitable type of content. In some embodiments, server 102 may perform any suitable functions. For example, in some embodiments, the server 102 may be used to process audience user comment messages to obtain intent cluster results that may reflect the user intent of one or more audience users, which are sent to the anchor user device 1062 for presentation by the live anchor user device 1062. As an alternative example, in some embodiments, the server 102 may be configured to perform semantic analysis on the comment messages of the audience users to obtain user intentions, perform clustering based on the user intentions to obtain intention clustering results, and send the intention clustering results to the user device 1062 at the anchor side for presentation through the user device 1062 at the anchor side that is live. As an alternative example, in some embodiments, the server 102 may be used to provide a corresponding interface for the user device 1062 at the anchor or the user device 1064 at the audience user to call to return data needed by the user device 1062 at the anchor or the user device 1064 at the audience user.

In some embodiments, the communication network 104 may be any suitable combination of one or more wired and/or wireless networks. For example, the communication network 104 can include any one or more of the following: the network may include, but is not limited to, the internet, an intranet, a Wide Area Network (WAN), a Local Area Network (LAN), a wireless network, a Digital Subscriber Line (DSL) network, a frame relay network, an Asynchronous Transfer Mode (ATM) network, a Virtual Private Network (VPN), and/or any other suitable communication network. The user device 1062 at the anchor end and the user device 1064 at the audience clients can be connected to the communication network 104 via one or more communication links (e.g., communication link 112), and the communication network 104 can be linked to the server 102 via one or more communication links (e.g., communication link 114). The communication link may be any communication link suitable for transferring data between the user device 1062 at the anchor end/the user device 1064 at the audience user end and the server 102, such as a network link, a dial-up link, a wireless link, a hardwired link, any other suitable communication link, or any suitable combination of such links.

In a live scenario, the user device 1064 at the audience client is mainly used to display live content and receive comment messages issued by the audience user during the live content viewing process, where the comment messages are sent to the server 102.

The user device 1062 of the anchor may include any user device or devices suitable for interacting with an audience user to present corresponding information and interfaces. In some embodiments, the anchor-side user device 1062 may be configured to transmit anchor recorded live content to the server 102 for transmission by the server 102 to the user device 1064 at the audience user's end. On this basis, as an optional example, in some embodiments, the user device 1062 at the anchor may also be configured to receive an intention clustering result sent by the server 102 (where the intention clustering result is obtained by the server 102 by determining a user intention according to a comment message sent by the user device 1064 at the audience client and clustering based on the user intention), so as to be presented to the anchor through the user device 1062 at the anchor.

In some embodiments, the user device 1062 at the anchor end and the user device 1064 at the audience user end may each comprise any suitable type of device. For example, in some embodiments, the user device 1062 at the anchor end and the user device 1064 at the audience user end may each comprise a mobile device, a tablet computer, a laptop computer, a desktop computer, a wearable computer, a game console, a media player, a vehicle entertainment system, and/or any other suitable type of user device.

Although server 102 is illustrated as one device, in some embodiments, any suitable number of devices may be used to perform the functions performed by server 102. For example, in some embodiments, multiple devices may be used to implement the functions performed by the server 102. Alternatively, the functionality of the server 102 may be implemented using a cloud service. It should be noted that if the user device 1062 at the anchor end has high software and hardware performance, part or all of the above-mentioned functions related to comment message processing may be performed instead of the server 102.

Based on the system, the embodiment of the application provides a method for processing live broadcast data in real time, which is described below through a plurality of embodiments.

Example one

Referring to fig. 2A, a flowchart illustrating steps of a real-time live data processing method according to a first embodiment of the present application is shown.

The real-time live broadcast data processing method of the embodiment comprises the following steps:

step S202: and obtaining comment messages sent by the user in the process of watching the live broadcast content in real time.

Most live applications provide settings for audience users to interact with the anchor or other audience users while viewing live content, including at least settings for audience users to post comment messages. Most audience users interact with the anchor in the form of comment messages with respect to live content, objects targeted by the live broadcast, live broadcast behavior of the anchor, and the like. The comment messages are uploaded to the server and then sent to the anchor side by the server. Correspondingly, the anchor end can acquire the comment messages published by the audience users in real time. It should be noted that, in the embodiment of the present application, for simplicity, the "user" means an audience user hereinafter.

In the embodiment of the present application, the form of the comment message includes, but is not limited to: at least one of a text comment message, a voice comment message, an image comment message, an emoticon comment message, and the like. For other types of comment messages than text comment messages, they may be converted to text messages before processing. The expression comment message can be converted into a default text message or a text message capable of expressing expression meanings; the image comment message can be converted into a text message which can reflect the main content of the image. Based on this, in one possible way, the present step can be implemented as: acquiring an original text comment message input by a user in a live content watching process in real time; and/or acquiring a text comment message generated after converting an expression comment message, a voice comment message or an image comment message input by a user in the process of watching the live content in real time. Therefore, the data processing of the embodiment of the application is more flexible, and the live application can provide richer comment forms for the user.

In addition, in order to guarantee the instantaneity of obtaining the comment messages, in a feasible mode, the comment messages sent by the user in the process of watching the live content can be obtained in real time through the message queue. The message queue may take the form of a queue such as a kafka message queue, a TT log queue, or the like. But not limited thereto, other forms of queue and other forms capable of obtaining comment messages in real time can be applied to the solution of the embodiment of the present application.

Through the process, the comment message issued by the user can be acquired in real time.

Step S204: and performing semantic analysis on the comment message within a preset time range, and acquiring a user intention corresponding to the comment message according to a semantic analysis result.

The specific implementation of semantic analysis on the comment message can be implemented by those skilled in the art in a suitable manner according to actual needs, including but not limited to a neural network model such as a CNN model, or by a semantic analysis algorithm, and the embodiment of the present application does not limit this.

Through semantic analysis, the user intention corresponding to each comment message can be obtained. For example, when a casual pants is sold by a live broadcast, a anchor broadcasts comment messages of material, dressing effect and unit price of the consulted pants, and semantic analysis is performed on the comment messages, so that it can be determined that the user intention corresponding to each comment message is respectively the consulted intention for the material, dressing effect and unit price of the pants.

Step S206: and clustering the comment messages according to the user intention to obtain at least one intention clustering result.

Since there is a possibility that a plurality of comment messages express the same or similar user intentions, for example, user a issues a comment message "how much money is in the trousers", user B issues a comment message "how much the asking price", and the two comment messages are subjected to semantic analysis, so that the consulting intentions that the corresponding user intentions are unit prices can be determined. Then, by clustering, the two comment messages can be grouped into one category.

Therefore, after the user intention corresponding to the comment message is obtained, the comment messages can be clustered based on the user intention, so that the comment messages with the same or similar intentions are clustered into one class, the subsequent data processing and displaying efficiency is improved, and the user intention can be obtained more easily by the anchor.

Step S208: and displaying the intention clustering result through a live broadcast information interface of the live broadcast content.

After the intention clustering result is determined, in a feasible scheme, the intention clustering result can be sent to a main broadcasting end, and the main broadcasting end processes or processes the intention clustering result and displays the intention clustering result on a live broadcasting information interface of live broadcasting content; in another feasible scheme, the intention clustering result can be processed into a prompt message form according to a pre-stored template, and the prompt message form is sent to the anchor terminal to be displayed through a live broadcast information interface of the anchor terminal.

The live information interface may be an independent interface that can be displayed based on live content, such as a pop-up window interface, a mask interface, a floating layer, or the like. But the live broadcast information interface can also be an area belonging to the live broadcast interface, and the intention clustering result is displayed through the area.

Hereinafter, the above process is exemplarily illustrated with one scenario example, as shown in fig. 2B.

In fig. 2B, the brief illustration is that there are 10 users (users 1-10, respectively) watching a live X sale casual pants, where user 1 posts a comment message 1: "what material the trousers are made of"; user 2 posts comment message 2: "what model is there is"; user 3 posts comment message 3: "how much money there is a discount"; user 4 posts a comment message 4: "do model, see effect"; user 5 posts comment message 5: what price; user 6 posts a comment message 6: "does the lean person wear it; user 7 posts comment message 7: "how much money, now discounted"; user 8 posts comment message 8: "is the corduroy"; user 9 posts comment message 9: "make an order now, can arrive in stock for several days"; user 10 posts a comment message 10: "what time my Tianjin, who has made an order, can be sent". These comment messages are sent in real time to the server via the data transmission channel and to the live client of the anchor X via the server. On one hand, the live client of the anchor X can display the comment messages in real time after receiving the comment messages, and update the interface display according to the messages sent by the server continuously. On the other hand, after receiving the comment messages, the server periodically processes the comment messages received in real time according to a certain time range, which is assumed to be processed once every minute, and the 10 comment messages are assumed to be all in a processing time range period. Then, semantic analysis is performed on the 10 comment messages, and the user intention corresponding to each comment message is determined, for example, comment message 1-material, comment message 2-model, comment message 3-unit price and discount, comment message 4-loading effect, comment message 5-unit price, comment message 6-loading effect, comment message 7-unit price and discount, comment message 8-material, comment message 9-delivery time, comment message 10-delivery time.

Further, after the user intentions corresponding to the respective comment messages are determined, the user intentions may be clustered to obtain intention clustering results, for example, an interest material class (comment messages 1 and 8), an interest model class (comment message 2), an interest unit price and benefit class (comment messages 3, 5, and 7), an interest loading effect class (comment messages 4 and 6), and an interest delivery time class (comment messages 9 and 10). Based on this, a corresponding prompt message can be generated, for example, the prompt message can be sent to the anchor terminal, such as "the user is interested in the material/model/unit price and the privilege/dressing effect/delivery time, please introduce". And at the anchor end, displaying the prompt information to the anchor through a popup window so as to prompt the anchor to respond to the user intention, namely the content concerned by the user in time through the prompt information, or adjust the live broadcast content or the live broadcast behavior of the anchor, and the like.

Of course, the above is only an exemplary illustration, and in practical applications, the presentation manner of the intent clustering result may be set by those skilled in the art according to actual needs, and is not limited to the manner described in the above examples.

Through the embodiment, the comment messages published in the process of watching the live broadcast content by the user can be acquired in real time, and then semantic analysis is carried out, so that the user intention which the user wants to express based on the comment messages is obtained. Because there may be many users watching live content at the same time, intentions which are desired to be expressed by comment messages issued by a plurality of users may be the same or similar, and based on the intentions, clustering can be performed based on the obtained user intentions to obtain one or more intention clustering results, and then the intention clustering results are displayed through a live information interface. Therefore, on one hand, core appeal and idea of the user can be known through semantic analysis and subsequent intention clustering of the comment message, manual processing is not needed, and labor and time cost is saved; on the other hand, due to the fact that the comment messages are obtained in real time and corresponding analysis and clustering are carried out in time, timeliness of the user complaint and the transmission of the idea can be effectively guaranteed, user requirements are met in time and effectively, user experience and anchor experience are improved, and user and anchor stickiness is increased.

Example two

Referring to fig. 3A, a flowchart illustrating steps of a real-time live data processing method according to a second embodiment of the present application is shown.

The embodiment of the present application takes a display mode of an intention clustering result as an emphasis point, and explains the real-time live broadcast data processing method of the embodiment of the present application.

step S302: and determining a preset time range and a display mode of a live information interface according to the interaction setting.

The interactive setting at least comprises a plurality of option settings for indicating the information display mode.

In this embodiment, an option or an input capable of setting an interaction, that is, an interaction setting, is further provided in the live application. Through the interaction setting, the anchor can set the time period of semantic analysis of comment messages, namely the preset time range, according to the needs of the anchor, and set the display mode of a live broadcast information interface. These interaction settings may be implemented, for example: and the time range input box or the single option, the live broadcast information interface display mode option box or the single option and the like are used for facilitating the selection and setting of the anchor. These interaction settings are uploaded to the server where they are stored and live data processing is performed in real time according to the interaction settings.

In one possible approach, the interaction settings for the preset time range may include, but are not limited to: the preset time range is set as a preset time interval period, or the preset time range is set as a preset time window length.

In another possible approach, the interaction setting for the presentation manner of the live information interface may include, but is not limited to: and displaying the live broadcast information interface in a rolling window mode, or displaying the live broadcast information interface in a sliding window mode.

Through the interactive setting, a convenient and flexible setting means is provided for the anchor, and the actual requirements of the anchor can be met to the greatest extent.

It should be noted that, the effect of one-time interactive setting for subsequent long-term use can be realized through the step, and interactive setting is not required to be executed every time real-time live data processing is carried out. In addition, the interaction setting is not limited to the setting of the preset time range and the setting of the display mode of the live information interface, and other setting items can be added by a person skilled in the art according to actual needs.

Step S304: and obtaining comment messages sent by the user in the process of watching the live broadcast content in real time.

For the specific implementation of this step, reference may be made to the related description in the foregoing first embodiment, and details are not described herein again.

Step S306: and performing semantic analysis on the comment message within a preset time range, and acquiring a user intention corresponding to the comment message according to a semantic analysis result.

This step is performed in relation to the setting of the preset time range in the aforementioned interaction setting. Based on this, in one possible way, in response to the preset time range being the preset time interval period, the step may be implemented as: and performing semantic analysis on the comment message according to a preset time interval period. The time interval period can be set by a person skilled in the art according to actual needs, and if the time interval period is 2 minutes, for example, semantic analysis is performed on the obtained comment messages every 2 minutes to obtain corresponding user intentions. By the method, the intention of the user can be timely acquired in real time, the algorithm is simple to implement, and the implementation cost is low.

In another possible way, in response to the preset time range being the preset time window length, the step may be implemented as: and obtaining the comment messages in the time period corresponding to the time window according to the preset time window length, and performing semantic analysis on the obtained comment messages. Different from the time interval period, the time window is related to the step length, for example, the time window is 2 minutes, and the step length is 1 minute, then in the first time window, the comment messages of 1-2 minutes will be semantically analyzed; in the second time window, semantic analysis is carried out on the comment messages of the 2 nd to 3 rd minutes; in a third time window, the comment messages at minutes 3-4 are semantically analyzed, and so on. That is, the comment messages corresponding to different time windows may partially overlap. By the method, omission of comment messages can be effectively avoided, and the display of subsequent intention clustering results is more continuous.

Step S308: and clustering the comment messages according to the user intention to obtain at least one intention clustering result.

Clustering comment messages based on their corresponding user intentions can be achieved using conventional clustering algorithms, including but not limited to: k-means clustering, gaussian mixture model clustering, etc., which are not limited in the embodiments of the present application.

Step S310: and displaying the intention clustering result through the live broadcast information interface of the live broadcast content according to the set display mode of the live broadcast information interface.

As mentioned above, the interactive setting for the presentation mode of the live information interface includes but is not limited to: and displaying the live broadcast information interface in a rolling window mode, or displaying the live broadcast information interface in a sliding window mode.

Based on this, in a feasible manner, in the case of performing semantic analysis on the comment message according to a preset time interval period, the step may be implemented as: and displaying the intention clustering results in a rolling mode through a rolling window based on a live broadcast interface of live broadcast content and a determined display mode of a live broadcast information interface, wherein the rolling window displays the intention clustering results corresponding to a preset number of comment messages in a time interval period each time. Further optionally, when the intention clustering result is displayed in a rolling mode through a rolling window on the live broadcast interface of the live broadcast content, a new window can be displayed on the live broadcast interface of the live broadcast content, the intention clustering result is updated and displayed through the new window at intervals, and the intention clustering results displayed twice in adjacent are different intention clustering results. In this way, semantic analysis is performed on the comment messages every preset time interval period to obtain the intentions of the users, clustering is performed to obtain intention clustering results, and the intention clustering results are displayed in a rolling mode through a window, wherein a preset number (for example, 5) of intention clustering results are displayed each time, and the intention clustering results displayed this time are different from the intention clustering results displayed last time. Therefore, the periodical automatic updating display of the intention clustering result is realized.

A live information interface in the manner described above is shown in fig. 3B, which illustrates an example of a scrolling window that automatically updates the display at preset time interval periods. In this way, comment messages in a time interval period are clustered based on corresponding user intents, then top N processing is performed according to indexes such as the number and weight of the comment messages corresponding to intention clustering results, after N intention clustering results with high number and high weight are normalized and integrated based on preset rules or templates, an anchor terminal is actively pushed, and a scroll window bar (shown as a dashed line frame part in fig. 3B) is displayed in a schematic diagram in fig. 3B through a scroll window in a live broadcast information interface. Wherein, N can be set by those skilled in the art according to actual requirements, such as 10. For example, a highlight message pop-up for displaying the intent cluster results may be made through a scrolling window above the lower left panel of the live-room interface. In fig. 3B, an example is an upper layer display of user comments, and the key message itself may be subjected to a visual enhancement process, so as to improve the anchor experience.

In another feasible manner, under the condition that comment messages in a time period corresponding to a time window are acquired according to a preset time window length, and the acquired comment messages are subjected to semantic analysis, the method can be implemented as follows: and displaying intention clustering results in a preset time step through sliding of a sliding window based on a live broadcast interface of live broadcast content and a determined display mode of a live broadcast information interface, wherein the sliding window displays intention clustering results corresponding to a preset number of comment messages in the length of the time window each time. Further optionally, when the intention clustering result within the preset time step is displayed in a sliding mode through a sliding window on the live broadcast interface based on the live broadcast content, an input trigger operation can be received, a new window is displayed on the live broadcast interface of the live broadcast content according to the trigger operation, the intention clustering result is updated and displayed through the new window every other preset time step, and the intention clustering results displayed twice in adjacent times are partly overlapped intention clustering results. The trigger operation may be a gesture trigger operation, a voice trigger operation, a trigger operation for an option or a button displayed in the interface, and the like. Therefore, in this way, the anchor can be allowed to perform manual trigger operation, and convenience is provided for the anchor to display and control the intention clustering result. The length of the time window and the preset time step can be properly set by a person skilled in the art according to actual requirements, and the time step is generally smaller than the length of the time window. Partial coincident data exists in the intention clustering results displayed in two adjacent times, so that the display is more continuous, and the overlooking of the anchor can be effectively avoided.

A live information interface in the manner described above is shown in fig. 3C, which illustrates an example of a sliding window that updates the display intention cluster result by a time step. Both the size (time window length) and the slide (time step) of the sliding window can be adjusted, taking the size as 5 minutes and the slide as 15 seconds as an example. The method comprises the steps of acquiring user intentions corresponding to comment messages of the latest 5 minutes every 15 seconds, clustering based on the user intentions to obtain intention clustering results, carrying out top N processing according to indexes such as the number and the weight of the comment messages corresponding to the intention clustering results, carrying out normalization and integration processing on N intention clustering results with high number and high weight, and sending the processed intention clustering results to an anchor terminal and displaying the intention clustering results through a sliding window if it is determined that the anchor actively initiates a trigger operation. Wherein, N can be set by those skilled in the art according to actual requirements, such as 10. Illustratively, the anchor can slide left on the interface of the live broadcast room to trigger a floating layer to appear on the interface of the live broadcast room, wherein three tags tab containing key comments, commodity popularity and active users are shown, and then clicking the key comments can show an intention clustering result of the latest time window length, which is indicated as key messages in the figure. Moreover, the key message can be subjected to visual enhancement processing, and anchor experience is improved.

Based on the two display modes, for the anchor with a large number of comment messages, even if the comment messages QPS are thousands, the core, key and top problems which need to be replied can be extracted from the comment messages by the scheme and displayed through the two modes, so that the anchor can timely and easily control the user intention of the user in the live broadcast room, replies, responses and the like are timely made, and live broadcast experience and live broadcast effects are improved.

In addition, if the anchor is a novice anchor, when the anchor is broadcast, if there is no comment message or no effective comment message in the live broadcast room, a default file can be preset as a key message, such as "parent, there is no key message in the live broadcast room temporarily, please multicast a moment, please ask for an expectation! "and the like.

EXAMPLE III

Referring to fig. 4A, a flowchart illustrating steps of a method for processing live data according to a third embodiment of the present application is shown.

In this embodiment, the processing of data is used as an emphasis to describe the real-time live data processing method in this embodiment.

step S402: and obtaining comment messages sent by the user in the process of watching the live broadcast content in real time.

For example, the comment message may be acquired in real time through a message queue or a TT log, or the like.

Step S404: and performing semantic analysis on the comment message within a preset time range, and acquiring a user intention corresponding to the comment message according to a semantic analysis result.

For the comment message, because the comment message is published by the user, part of the comment message may carry meaningless information, even information violating relevant regulations, such as bad words and the like. In order to improve the efficiency of semantic analysis on comment messages and avoid interference of invalid information on the semantic analysis, in a feasible way, dirty data filtering can be performed on the comment messages within a preset time range; performing semantic analysis on the comment message subjected to the dirty data filtering to obtain a semantic analysis result, wherein the semantic analysis result at least comprises a user intention and an intention category; and re-filtering the comment messages according to the intention categories to obtain the re-filtered comment messages and the user intentions corresponding to the re-filtered comment messages.

In the embodiments of the present application, dirty data means data that is not within a given range or meaningless to an actual service, or has an illegal data format. For example, the comment message may be filtered out of null characters, special characters, test data, malicious attack data, chatty data, and the like through dirty data filtering. In this embodiment, after performing semantic analysis on the comment message after filtering the dirty data, the obtained semantic analysis result includes not only the user intention, but also intention categories, such as user intentions related to live broadcast content (for example, intentions related to electronic commerce may be further subdivided into commodity recommendations, commodity attributes, commodity sizes, hot questions, general questions, coupons, live broadcast room discounts, and the like), user intentions related to target objects targeted by live broadcast, user intentions related to live broadcast anchor, and illegal user intentions, where the illegal user intentions may be intention categories related to yellow gambling poison, bad speech, confusing and disgust violence, discrimination, politics, deceptive expressions, and the like. By means of the intention categories, the comment messages can be effectively distinguished, and accurate bases are provided for subsequent data processing.

Based on the above, the comment message can be filtered again according to the intention category, and the comment message after being filtered again and the user intention corresponding to the comment message after being filtered again can be obtained. For example, by filtering out the comment messages corresponding to the intention categories such as yellow gambling poison, bad words, confusing, disguising, violence, discrimination, political, meaning deceptive words, and the like, the remaining comment messages and the corresponding user intentions thereof will be the comment messages and the user intentions effective for the subsequent data processing.

Step S406: and clustering the comment messages according to the user intention to obtain at least one intention clustering result.

In this embodiment, the obtained intent clustering result may include at least one of: the method comprises the steps of obtaining intention clustering results corresponding to comment messages larger than a preset number, intention clustering results corresponding to comment messages aiming at key contents in live broadcast contents, intention clustering results corresponding to comment messages aiming at key live objects in the live broadcast contents, and intention clustering results corresponding to comment messages aiming at key attributes of the live broadcast objects in the live broadcast contents.

The specific numerical value of the preset number can be set by a person skilled in the art according to actual conditions, and the embodiment of the present application does not limit the specific numerical value. For example, if 30 comment messages that are all the query unit prices and the offers exist in 100 comment messages, the clustering result will include the intended clustering result corresponding to such comment messages.

In the live broadcast, the anchor may play important contents that need to be advertised or promoted, for example, repeatedly played contents or contents that take a long time to play, and correspondingly, the important contents may also pay more attention to and issue comment messages, and the intention clustering results corresponding to the comment messages may also become a part of the obtained intention clustering results for subsequent display.

Similarly, live broadcast objects may also exist in the live broadcast content, such as some or some of the multiple commodities, for which there will be more comment messages, and the intention clustering result corresponding to this part of comment messages may be a part of the obtained intention clustering result for subsequent presentation.

The live broadcast object usually has various attributes, for example, content type, character, scene, material, model, unit price, preference, and the like, in live broadcast such as merchandise promotion, the unit price and the preference can be generally regarded as key attributes, corresponding comment messages are also more, and the corresponding intention clustering result also becomes a part of the obtained intention clustering result for subsequent display. For example, in live broadcast of content promotion, such as live broadcast of video content promotion, content types, actor characters and the like become important concerns of users, corresponding comment messages become more, and corresponding intention clustering results also become a part of obtained intention clustering results for subsequent display.

It should be understood by those skilled in the art that the above-mentioned multiple intent clustering results and the information that may be involved in the multiple intent clustering results are exemplary, and in practical applications, there may be other types of intent clustering results and other applicable information involved.

As can be seen from the above, in practical applications, the obtained intent clustering results may include a plurality of intent clustering results, and in a feasible manner, the plurality of intent clustering results may be sorted according to the number of comment messages corresponding to each intent clustering result and the preset weight; and screening the intention clustering results to be displayed from the plurality of intention clustering results according to the sorting result. Subsequently, when the intention clustering result needs to be displayed on the live interface of the live content, the intention clustering result to be displayed only needs to be displayed on the live interface of the live content. The weights corresponding to different intention clustering results can be set by a person skilled in the art according to the importance degree of intentions relative to live broadcast content, for example, for live broadcast of electronic goods, unit price, preference, effect and the like of the electronic goods are generally concerned much, weight setting can be respectively carried out based on the unit price, the preference, the effect and the like in combination with other possible intention clustering results, furthermore, in the live broadcast process, ranking is carried out based on the number of comment messages corresponding to each intention clustering result and preset weights, and the intention clustering result ranked as top N is used as an intention clustering result to be displayed. Wherein N may be set as appropriate by a person skilled in the art, such as 10, and the like, and this is not limited in the embodiments of the present application.

Further optionally, candidate intention clustering results can be screened from the plurality of intention clustering results according to the sorting result; normalizing the candidate intention clustering result according to a preset mode; and determining an intention clustering result to be displayed based on the result of the normalization processing. The preset mode may be a preset template or a preset data processing rule. For example, the preset template may be a template which is generated based on the intent clustering results and is convenient for the anchor understanding and reading, for example, "parent, audience now pay more attention … …, please introduce," wherein the ellipsis part is a part replaced by the intent clustering results, a key message may be generated according to the template for each intent clustering result, or a key message may be generated according to the template in a form of combining partial intent clustering results. The preset data processing rule can add or modify the file based on the obtained intention clustering result, and can set the display mode of the text, such as font, size, color and the like, so as to achieve the effect of striking display.

In addition, in practical application, for each live broadcast application or live broadcast platform, a super anchor usually exists, for example, live broadcasts attract thousands or even tens of thousands of audience users, and in order to avoid data processing delay and even collapse caused under similar conditions, in a feasible manner, comment messages are clustered according to user intentions, and when at least one intention clustering result is obtained, the comment messages can be clustered according to preset clustering parameters; obtaining a plurality of comment clustering hotspots according to the clustering result in which the number of the comment messages in the clustering result is greater than the preset number; and grouping the comment messages according to the comment clustering hotspots and the number of preset computing task nodes, and obtaining at least one corresponding intention clustering result according to the grouping result and the user intention corresponding to the comment messages in the grouping. The clustering parameters are different in different processing modes, the specific data form and numerical value of the clustering parameters are not limited in the embodiment of the application, and only data hot spots can be removed and the data are averagely dispersed to different task nodes. Therefore, a plurality of task nodes can be balanced, and the abnormal condition or the breakdown caused by the overweight task of a single node can be avoided. Illustratively, taking the case of processing using the BLINK technique as an example, the clustering process of comment messages can be optimized through two clustering parameters, minisbatch and aggregate, so as to balance the computation power.

Step S408: and displaying the intention clustering result through a live broadcast information interface of the live broadcast content according to the determined display mode.

The anchor can know the user intention in time by showing the intention clustering result, and if the user intention aiming at the live broadcast content or the live broadcast behavior of the anchor exists, the anchor can adjust the live broadcast content or the live broadcast behavior of the anchor in time according to the user intentions so as to meet the user requirements of audiences and obtain better live broadcast effect.

The specific determination and display implementation of the display mode of the live information interface may refer to the description of the relevant parts in the foregoing embodiment one or two, and are not described herein again.

Hereinafter, the above-described process is exemplified by an example, as shown in fig. 4B.

In the scene of the example, after the anchor clicks to start playing and push streaming, the audience users can enter the live broadcast room, and then the anchor can interact with the audience users through comment messages, such as chatting, introducing goods, answering questions of the audience users, interacting with the audience users, and the like, so that real-time comment messages are generated.

Based on the comment message generated in real time, the data processing procedure in this example includes: the method comprises five parts of data embedding, data calculation, algorithm model processing, data storage and key message display. Hereinafter, the description will be given separately.

Process 1, data Point burying

Namely, comment messages published by audience users in each live broadcast room are subjected to point burying and then collected into a message queue or TT logs for subsequent real-time data calculation task subscription. Shown schematically in fig. 4B as collected into the TT log.

Procedure 2, data calculation

After subscribing the real-time comment information of the live broadcast room, firstly, a first layer of data cleaning work is carried out, dirty data in the live broadcast room are filtered, and filtering is carried out on idle characters, special characters, test data, malicious attack data, data identified as chatty by a pre-algorithm and the like, for example, in a keyword matching mode.

Process 3, Algorithm model processing

In the process, the comment message after dirty data filtering is used as an input parameter, and an algorithm model of semantic analysis is called in a UDF mode to carry out semantic recognition, so that the user intention and intention category corresponding to the comment message are obtained. Optionally, a weight corresponding to the user intention may also be obtained, and/or a result obtained after the comment message is normalized. Then, according to the intention category, a layer of cleaning is carried out again, namely a second layer of cleaning, so as to filter data such as gambling related poison, bad speech, vicious and disguised violence, discrimination, political relation, meaning deception phrase and the like identified by the algorithm. And finally, outputting the original comment message, the normalized comment message, the user intention, the intention category, the weight and other related information.

Because two display modes of the live information interface exist in the example, the processing of the algorithm model is different.

The first active push mode is a mode of periodically and automatically updating the displayed intention clustering result through a rolling window, and calculation processing is performed based on the setting of the rolling window (wherein the window time is a preset time interval period and can be freely adjusted, such as 1 minute, 5 minutes and the like), for example, comment messages in the window time are subjected to semantic analysis, user intents obtained through the semantic analysis are clustered, then top N processing is performed according to parameters such as the number and weight of the comment messages corresponding to the intention clustering result obtained through clustering, the N intention clustering results with high number and weight are normalized and integrated and then returned, and then the N intention clustering results are written into a downstream log (such as TT, metaq and the like) for a downstream engineering end to subscribe.

For the second trigger sending mode, that is, the sending intention clustering result is triggered to be displayed based on the input of the anchor through the sliding window, based on the sliding window (the size and slide of the sliding window are both adjustable, for example, the size is 5 minutes, the slide is 15 seconds, that is, the comment message processing of the latest 5 minutes will be performed every 15 seconds); semantic analysis is carried out on comment messages in window time to obtain user intentions, clustering is carried out based on the user intentions, top N processing is carried out according to indexes such as the number and the weight of the comment messages corresponding to meaning clustering results obtained through clustering, N intention clustering results with high number and high weight are normalized, integrated and returned, and then the N intention clustering results are written into databases such as downstream OLAP/NoSQL (such as ADB, HOLO, HBASE, Lindorm and the like, including but not limited to the two databases). In one example, when the anchor performs a left-sliding operation through a terminal screen in a live broadcast room, an interface of a related engineering end is called, then a latest key message (i.e., an intention clustering result) in the live broadcast room is queried, and the key message is returned and then displayed in a left-sliding window.

For a live broadcast room of a head anchor (an anchor with a large traffic), when data processing is performed on a real-time comment message, a hot spot problem of a calculation task node may occur because the traffic of a single live broadcast room is larger than the total traffic. For this purpose, it is possible to: (1) first, clustering parameters are set for data processing optimization. For example, BLINK can be used to add miniBatch, aggregate, etc. optimization strategies for the window. However, in different data processing methods, the clustering parameters for performing the optimization may be different, for example, Spark Streaming may be micro batch processing. Because factors influencing the data processing performance mainly comprise the quality of the code and the setting of the batch Duration optimization value, whether the code is configured or adjusted, data hot spots can be removed, the data can be uniformly distributed, and multiple layers of aggregation and deduplication can be performed. (2) And then data hot spots are scattered. For example, a HASH _ CODE discrete function may be used to separate data hotspots, and then a MOD function is used to perform a grouping operation on HASH values obtained by the HASH _ CODE (for example: MOD (HASH _ CODE (X), N), where X represents a field with relatively uniform distribution in the data index, and N is an average number of divided task nodes, such as thread nodes, etc.), so that data can be effectively and evenly scattered onto each task node, data skew caused by data volume accumulation of individual task nodes is avoided, and then group agg is performed according to the dimension. Therefore, the problem of data processing abnormity or crash caused by data hot spots can be effectively avoided.

Procedure 4, data storage

As described above, in this example, because there are two display modes of the live information interface, different data storage modes are set for different display modes. The first active push mode is a mode of periodically and automatically updating the display intention clustering result through a rolling window, a storage medium is set to be of a log and message queue type, downstream user subscription is supported, and when data are written in the upstream, the data are directly pushed to a downstream scene; and the second trigger sending mode is a mode of triggering sending intention clustering results to be displayed based on the input of the anchor through a sliding window, and a storage medium is set to be a relational or non-relational database such as OLAP/NoSQL and the like so as to support the anchor to query and obtain required data according to query conditions.

Process 5, Key message display

As also described above, the live information interface in this example has two display modes.

Aiming at the first active push mode, namely a mode of periodically and automatically updating the display intention clustering result through a rolling window, an engineering end actively subscribes a TT log, actively packages and pushes the TT log to a main broadcasting end after key messages are obtained, the TT log is displayed through the rolling window, the display effect is as shown in figure 3A, key messages are popped up above a lower left corner evaluation area of a live broadcasting room, the display in figure 3A is shown to be on the upper layer of the evaluation area of a user, the key messages can also be subjected to vision enhancement processing, and the main broadcasting experience is improved.

For the second trigger sending mode, that is, the mode of triggering sending intention clustering results to be displayed based on the input of the anchor through a sliding window, the anchor needs to actively initiate a trigger operation to display (that is, passively). Exemplarily, the anchor performs left-sliding operation on a screen of a live broadcast room interface, a floating layer appears, the floating layer comprises three tabs of key messages, commodity popularity and active users, the key messages of the latest N minutes can be displayed by clicking the key messages, and the visual enhancement processing can be performed at the position, so that the anchor experience is improved.

When the novice broadcasts no comment message or no effective comment message in the live broadcast room, a default file can be set, for example, "no important message is left in the parent and live broadcast rooms, please multicast for a moment and give a desire".

Therefore, even if the comment message of one-time live broadcast reaches more than a QPS (quick Path manager) million anchor, the user intention can be quickly obtained, the reply, the response and the like can be timely made, and the live broadcast experience is improved.

Through the embodiment, based on real-time comment messages, message queues or TT logs, data filtering and the like, the delay of the whole real-time live broadcast data processing can be controlled at the second level, so that the anchor can make timely response and reply, and the experience of the anchor and audience users is improved. Based on the two display modes, the anchor can acquire the user intention with higher flexibility and autonomy, and the user experience and the anchor experience are further improved.

Example four

Referring to fig. 5, a flowchart illustrating steps of a real-time live data processing method according to a fourth embodiment of the present application is shown.

The present embodiment explains the real-time live data processing method provided in the embodiment of the present application from the perspective of a live interface of a anchor terminal, and the real-time live data processing method includes the following steps:

step S502: and displaying the live broadcast content and comment messages input by the user in the process of watching the live broadcast content on a live broadcast interface.

Whether the live application or the live platform is provided with a live room, the anchor carries out live operation through the live room, and live content is generated and displayed through a live interface; audience users watch live content and may enter comment messages during the viewing process to interact with the anchor.

Step S504: and when the comment intention prompt condition is determined to be met, displaying the user intention prompt message generated according to the comment message on a live interface.

And a comment intention prompting condition is preset in the live broadcast application or the live broadcast platform, so that when the condition is met, a user intention prompting message generated based on the comment message is displayed to the anchor through a live broadcast interface of the anchor.

The comment intention prompting condition can be set by a person skilled in the art according to actual requirements, for example, the default condition is met, and the user intention prompting message can be displayed in real time or according to the setting of a display mode; of course, a predetermined condition may be set, and when the condition is satisfied, the user intention prompting message may be displayed. In one possible approach, the user intention prompting message may be generated according to the intention clustering result obtained by the method as described in any one of the first to third embodiments.

Through the embodiment, in the aspect of user intention display, when the comment intention prompt condition is met, the user intention prompt message generated according to the comment message of the user can be displayed on the live interface. Therefore, the message acquisition end can timely know the user intention capable of expressing the core appeal and idea of the user such as the live anchor and the like and timely process the user intention so as to timely and effectively meet the user requirements, simultaneously improve the experience of the user and the anchor and increase the viscosity of the user and the anchor.

EXAMPLE five

Referring to fig. 6A, a flowchart illustrating steps of a real-time live data processing method according to a fifth embodiment of the present application is shown.

In this embodiment, the method for processing live broadcast data provided in this embodiment is still described from the perspective of a live broadcast interface of a anchor, and the method for processing live broadcast data includes the following steps:

step S602: and displaying the live broadcast interface.

Step S604: and displaying the live broadcast content and comment messages input by a user in the process of watching the live broadcast content on a live broadcast interface.

Wherein, the form of the comment message includes but is not limited to: text messages, voice messages, image messages, emoticon messages, and the like. The voice message, the image message and the expression message can be converted into text messages and then are processed subsequently.

And step S606, when the comment intention prompting condition is determined to be met, displaying the user intention prompting message generated according to the comment message on a live interface.

In a feasible mode, when the comment intention prompting condition is determined to be met according to the type of the comment intention prompting condition, the user intention prompting message generated according to the comment message can be displayed on the live interface according to a display mode matched with the type. By setting different types for the comment intention prompt conditions, more flexible display mode setting can be realized.

For example, when the comment intention prompting condition is a state that a default condition is satisfied, it is determined that the comment intention prompting condition is satisfied, and a user intention prompting message generated according to the comment message and sent in an active push manner is received, and the user intention prompting message is displayed through a live interface. In the mode, the back end, such as an engineering end, can actively push the user intention prompt message to the live broadcast application, the anchor end can display the user intention prompt message without further judgment after receiving the user intention prompt message, and therefore operation of the anchor end is simplified, operation burden of the anchor is relieved, implementation is simple, and implementation cost is low.

During specific display, real-time display can be performed, but preferably, the display mode based on active push described in the foregoing embodiments, that is, the display mode based on a rolling window, is adopted, and the display is automatically updated at intervals of a preset time interval period. For specific implementation, reference may be made to the description of relevant parts in the foregoing embodiments, which are not described herein again.

For the presentation of the user intention prompting message, that is, when the user intention prompting message is presented through the live interface, optionally, the following may be implemented: the method comprises the following steps of performing visual enhancement processing on a comment message displayed on a live broadcast interface according to a preset format and displaying the comment message, wherein the preset format comprises at least one of the following: font type format, font color format, font size format. Therefore, the emphasis display and the reminding of the user intention prompt message are realized.

In another feasible way, according to the type of the comment intention prompting condition, when the comment intention prompting condition is determined to be satisfied, according to a display way matched with the type, displaying the user intention prompting message generated according to the comment message on the live broadcast interface can be realized as follows: receiving trigger operation input by a host for indicating comment intention prompt; determining that the comment intention prompting condition is met according to the triggering operation, and responding to the triggering operation to obtain a user intention prompting message generated according to the comment message; and displaying the user intention prompt message on the live broadcast interface through the floating layer.

Wherein the triggering operation includes, but is not limited to, one of: gesture trigger operation, voice trigger operation and interface option trigger operation. The gesture trigger operation is gesture operation based on the display screen of the anchor terminal, such as gesture-related operations of left sliding, right sliding, double clicking, multi-clicking and the like; the voice trigger operation is an operation form of sending a trigger instruction through voice, such as through a voice keyword 'user intention' and the like; the interface option triggering operation is based on corresponding options in the live broadcast interface, such as triggering operations of buttons, options hooking and the like. Through the trigger operation, convenience is provided for the anchor to control the display of the user intention prompt message.

Based on this, only after receiving the trigger operation performed by the anchor, the comment intention prompting condition is considered to be satisfied, and then the subsequent display operation is performed.

When the user intention prompting message is displayed on the live broadcast interface through the floating layer, the user intention prompting message can be subjected to visual enhancement processing according to a preset format, and the user intention prompting message is displayed on the live broadcast interface through the floating layer, wherein the preset format comprises at least one of the following formats: font type format, font color format, font size format.

In addition, in order to improve the display efficiency and enrich the dimensionality of data display, in a feasible mode, a floating layer can be rendered and drawn on a live broadcast interface, wherein the floating layer comprises a plurality of tab pages, and the tab pages comprise a tab page for displaying a user intention prompt message, a tab page for displaying live broadcast content heat and a tab page for displaying user activity; and further, displaying the user intention prompting message on a live broadcast interface according to the selection operation of the tab page for displaying the user intention prompting message. One specific example interface is shown in fig. 3B. It should be noted that, in practical applications, the number of tab pages and the displayed content except the user intention prompting message can be set by those skilled in the art according to practical needs. And relatively speaking, the live content heat and the user activity have more reference significance for evaluating the effect of the main broadcasting playing behavior.

For some broadcasters, the broadcasters may not receive sufficient effective comment messages, and in this case, if the effective comment messages input by the user are not received, when it is determined that the comment intention prompt conditions are met, preset default user intention prompt messages are displayed on the live broadcast interface, for example, "no key messages are temporarily left between parents and live broadcasts, please multicast for a moment and give for expectation", so as to improve the live broadcast confidence of the broadcasters and improve the live broadcast experience of the broadcasters.

The above process is exemplified below by taking an e-commerce live broadcast scene as an example, as shown in fig. 6B.

In fig. 6B, it is assumed that anchor X sells a certain brand of mobile phone through a live broadcast room, and since there are many audience users, in order to avoid missing or failing to reply the comment message of the audience users in time, anchor X performs a left-sliding operation on a live broadcast interface through a display screen of an anchor terminal at the beginning of a live broadcast. The live broadcast application determines that the comment intention prompt condition is met based on the left-sliding operation, and performs corresponding response prompt, such as popping up prompt information "the user intention is to prompt you in time", or the prompt message can be played in a voice form to inform the anchor that the trigger operation of the anchor has been responded.

And then, a floating layer is displayed on the upper layer of the live broadcast interface, the user intention prompting messages are updated in the floating layer every one minute, five user intention prompting messages are displayed each time, and the overlapping parts exist in the user intention prompting messages displayed twice. As shown in fig. 6B, in the first minute, the user intention prompting messages 1-5 are displayed to the anchor X through the floating layer, in the second minute, the user intention prompting messages 4-8 are displayed to the anchor X through the floating layer, in the third minute, the user intention prompting messages 7-11 are displayed to the anchor X through the floating layer, and so on until the live broadcast is finished or the anchor X performs right-sliding operation in the live broadcast process. In this example, the right slide operation is set for instructing to stop the display of the user intention prompting message. The anchor X can prompt the user intention prompt message through the continuously updated user intention displayed in the floating layer, so that the audience user can pay attention to the problem and timely reply.

If the anchor X does not perform the left-sliding operation, the live application may display the user intention prompting message in another manner, as indicated by the dashed boxes in the lower half of fig. 6B. That is, the anchor X does not perform any operation after entering the live broadcast room, and in this case, if the manner that the comment intention prompting condition is satisfied by default is used in the live broadcast application, after the broadcast, a small popup window is displayed on the upper layer of the comment message in the live broadcast interface, the displayed user intention prompting message is updated once every certain time interval, for example, one minute, and the user intention prompting messages displayed twice in adjacent times are completely different. Therefore, even if the anchor does not perform any operation, the intention of the audience user can be known in time, and the reply can be made in time. But not limited thereto, if the anchor X does not want to display the user intention prompting message, the function may also be closed through a corresponding operation, such as clicking a close button in a pop-up window, or a slide-up operation or a slide-down operation, etc.

EXAMPLE six

Referring to fig. 7, a schematic structural diagram of an electronic device according to a sixth embodiment of the present application is shown, and the specific embodiment of the present application does not limit a specific implementation of the electronic device.

As shown in fig. 7, the electronic device may include: a processor (processor)702, a receiver 704, a transmitter 706. In addition, in practical applications, the electronic device may further include a communication Interface 708, a memory 710, and a communication bus 712.

Wherein:

the processor 702, receiver 704, transmitter 706, communication interface 708, and memory 710 communicate with each other via a communication bus 712.

The receiver 704 is configured to receive a comment message input by a user during viewing of the live content, and send the comment message to the processor 702.

The processor 702 is configured to execute the program 714, and may specifically execute the relevant steps of the real-time live data processing method according to any one of the first to third embodiments of the method. For example, for obtaining comment messages in real time; semantic analysis is carried out on the comment message within a preset time range, and a user intention corresponding to the comment message is obtained according to the result of the semantic analysis; and clustering the comment messages according to the user intention to obtain at least one intention clustering result.

And the sender 706 is configured to send the intention clustering result to the user equipment of the anchor side, so that the intention clustering result is displayed on a live interface of live content through the user equipment of the anchor side.

A communication interface 708 for communicating with other electronic devices or servers.

In particular, the program 714 may include program code comprising computer operational instructions.

The processor 702 may be a CPU, or an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits configured to implement embodiments of the present application. The intelligent device comprises one or more processors which can be the same type of processor, such as one or more CPUs; or may be different types of processors such as one or more CPUs and one or more ASICs.

And a memory 710 for storing a program 714. Memory 710 may comprise high-speed RAM memory and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The program 714 may be specifically configured to enable the processor 702 to execute an operation corresponding to the real-time live data processing method in any one of the first to third embodiments of the method.

For specific implementation of each step in the program 714, reference may be made to corresponding steps and corresponding descriptions in units in the first to third embodiments of the real-time live data processing method, and corresponding beneficial effects are provided, which are not described herein again. It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described devices and modules may refer to the corresponding process descriptions in the foregoing method embodiments, and are not described herein again.

EXAMPLE seven

Referring to fig. 8, a schematic structural diagram of an electronic device according to a seventh embodiment of the present application is shown, and the specific embodiment of the present application does not limit a specific implementation of the electronic device.

As shown in fig. 8, the electronic device may include: a processor (processor)802, an input device 804, and a display 806. In addition, in practical applications, the electronic device may further include a communication Interface 808, a memory 810, and a communication bus 812.

Wherein:

the processor 802, input device 804, display 806, communication interface 808, and memory 810 communicate with one another via a communication bus 812.

And an input device 804, configured to receive a comment message input by a user during a live content viewing process, and send the comment message to the processor 802.

The processor 802 is configured to execute the program 814, and may specifically execute relevant steps of the real-time live data processing method according to any one of the fourth to fifth embodiments of the foregoing method. For example, for sending live content and comment messages to the display 806, so that the display 806 shows the live content and the comment messages on a live interface; and when the comment intention prompting condition is determined to be met, acquiring a user intention prompting message generated according to the comment message.

The display 806 is also used for displaying a user intention prompting message on the live interface.

A communication interface 808 for communicating with other electronic devices or servers.

In particular, the program 814 may include program code that includes computer operating instructions.

The processor 802 may be a CPU, or an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits configured to implement embodiments of the present application. The intelligent device comprises one or more processors which can be the same type of processor, such as one or more CPUs; or may be different types of processors such as one or more CPUs and one or more ASICs.

A memory 810 for storing a program 814. Memory 810 may comprise high-speed RAM memory and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The program 814 may be specifically configured to enable the processor 802 to execute an operation corresponding to the real-time live data processing method in any one of the fourth to fifth embodiments of the method.

For specific implementation of each step in the program 814, reference may be made to corresponding steps and corresponding descriptions in units in the fourth to fifth embodiments of the real-time live data processing method, and corresponding beneficial effects are provided, which are not described herein again. It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described devices and modules may refer to the corresponding process descriptions in the foregoing method embodiments, and are not described herein again.

An embodiment of the present application further provides a computer program product, which includes a computer instruction, where the computer instruction instructs a computing device to execute an operation corresponding to any one of the real-time live data processing methods in the multiple method embodiments.

It should be noted that, according to the implementation requirement, each component/step described in the embodiment of the present application may be divided into more components/steps, and two or more components/steps or partial operations of the components/steps may also be combined into a new component/step to achieve the purpose of the embodiment of the present application.

The above-described methods according to embodiments of the present application may be implemented in hardware, firmware, or as software or computer code storable in a recording medium such as a CD ROM, a RAM, a floppy disk, a hard disk, or a magneto-optical disk, or as computer code originally stored in a remote recording medium or a non-transitory machine-readable medium downloaded through a network and to be stored in a local recording medium, so that the methods described herein may be stored in such software processes on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware such as an ASIC or FPGA. It will be appreciated that the computer, processor, microprocessor controller or programmable hardware includes memory components (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code that, when accessed and executed by the computer, processor or hardware, implements the real-time live data processing methods described herein. Further, when a general purpose computer accesses code for implementing the live data processing method shown herein, execution of the code transforms the general purpose computer into a special purpose computer for performing the live data processing method shown herein.

Those of ordinary skill in the art will appreciate that the various illustrative elements and method steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the embodiments of the present application.

The above embodiments are only used for illustrating the embodiments of the present application, and not for limiting the embodiments of the present application, and those skilled in the relevant art can make various changes and modifications without departing from the spirit and scope of the embodiments of the present application, so that all equivalent technical solutions also belong to the scope of the embodiments of the present application, and the scope of patent protection of the embodiments of the present application should be defined by the claims.

Claims

1. A real-time live data processing method comprises the following steps:

obtaining comment messages sent by a user in a live content watching process in real time;

semantic analysis is carried out on the comment message within a preset time range, and a user intention corresponding to the comment message is obtained according to the result of the semantic analysis;

clustering the comment messages according to the user intention to obtain at least one intention clustering result;

and displaying the intention clustering result through a live broadcast information interface of the live broadcast content.

2. The method of claim 1, wherein the method further comprises:

and determining the preset time range and the display mode of the live information interface according to interaction settings, wherein the interaction settings at least comprise a plurality of option settings for indicating the information display mode.

3. The method of claim 2, wherein,

responding to the preset time range as a preset time interval period, and performing semantic analysis on the comment messages in the preset time range, wherein the semantic analysis includes: according to a preset time interval period, performing semantic analysis on the comment message;

the displaying the intention clustering result through a live information interface of the live content comprises: displaying the intention clustering results in a rolling mode through a rolling window based on a live broadcast interface of the live broadcast content and the determined display mode of the live broadcast information interface, wherein the rolling window displays intention clustering results corresponding to a preset number of comment messages in the time interval period each time;

wherein the live broadcast interface based on the live broadcast content displays the intention clustering result by rolling a rolling window, and comprises: and displaying a new window based on the live broadcast interface of the live broadcast content, updating and displaying intention clustering results through the new window at intervals of the time interval period, wherein the intention clustering results displayed twice in adjacent are different intention clustering results.

4. The method of claim 2, wherein,

responding to the preset time range as a preset time window length, and performing semantic analysis on the comment messages in the preset time range, wherein the semantic analysis includes: obtaining comment messages in a time period corresponding to a time window according to a preset time window length, and performing semantic analysis on the obtained comment messages;

the displaying the intention clustering result through a live interface of the live content comprises: based on a live broadcast interface of the live broadcast content and a determined display mode of the live broadcast information interface, displaying the intention clustering results in a preset time step in a sliding mode through a sliding window, wherein the sliding window displays the intention clustering results corresponding to a preset number of comment messages in the length of the time window each time;

wherein, the live broadcast interface based on the live broadcast content slides through a sliding window to display the intention clustering result in a preset time step, and the method comprises the following steps: receiving input trigger operation, displaying a new window on a live broadcast interface of live broadcast content according to the trigger operation, updating and displaying intention clustering results through the new window at intervals of preset time step, wherein the intention clustering results displayed twice in adjacent are intention clustering results with partial coincidence.

5. The method of any of claims 1-4, wherein the intent clustering results include at least one of: the method comprises the steps of obtaining intention clustering results corresponding to comment messages larger than a preset number, intention clustering results corresponding to comment messages aiming at key contents in live broadcast contents, intention clustering results corresponding to comment messages aiming at key live objects in the live broadcast contents, and intention clustering results corresponding to comment messages aiming at key attributes of the live broadcast objects in the live broadcast contents.

6. The method of any one of claims 1-4,

when the obtained intention clustering result comprises a plurality of results, the method further comprises: sequencing the plurality of intention clustering results according to the number of comment messages corresponding to each intention clustering result and a preset weight; screening an intention clustering result to be displayed from the intention clustering results according to the sorting result;

the displaying the intention clustering result through a live interface of the live content comprises: and displaying the intention clustering result to be displayed through a live broadcast interface of the live broadcast content.

7. The method according to any one of claims 1 to 4, wherein the semantic analysis of the comment message within the preset time range, and the obtaining of the user intention corresponding to the comment message according to the result of the semantic analysis includes:

filtering dirty data of the comment messages in the preset time range;

performing semantic analysis on the comment message subjected to the dirty data filtering to obtain a semantic analysis result, wherein the semantic analysis result at least comprises user intention and intention categories;

re-filtering the comment message according to the intention category to obtain the re-filtered comment message and the user intention corresponding to the re-filtered comment message;

wherein the clustering the comment messages according to the user intention to obtain at least one intention clustering result comprises: clustering the comment messages according to preset clustering parameters; obtaining a plurality of comment clustering hotspots according to the clustering result in which the number of the comment messages in the clustering result is greater than the preset number; and grouping the comment messages according to the comment clustering hotspots and the number of preset computing task nodes, and obtaining at least one corresponding intention clustering result according to the grouping result and the user intention corresponding to the comment messages in the grouping.

8. A real-time live data processing method comprises the following steps:

displaying live broadcast content and comment information input by a user in the process of watching the live broadcast content on a live broadcast interface;

and the number of the first and second groups,

and when the comment intention prompt condition is determined to be met, displaying a user intention prompt message generated according to the comment message on the live broadcast interface.

9. The method of claim 8, wherein the presenting, in the live interface, a user intent prompt message generated from the comment message upon determining that a comment intent prompt condition is satisfied comprises:

and according to the type of the comment intention prompting condition, when the comment intention prompting condition is determined to be met, displaying the user intention prompting message generated according to the comment message on the live broadcast interface according to a display mode matched with the type.

10. The method of claim 9, wherein the displaying, according to the type of the comment intention prompting condition and in a display manner matched with the type when the comment intention prompting condition is determined to be satisfied, the user intention prompting message generated according to the comment message on the live interface comprises:

when the comment intention prompting condition is a state that a default condition is met, determining that the comment intention prompting condition is met, receiving a user intention prompting message which is sent in an active push mode and is generated according to the comment message, and displaying the user intention prompting message through the live broadcast interface;

wherein the displaying the user intention prompting message through the live interface comprises: and performing visual enhancement processing on the user intention prompting message and displaying the user intention prompting message according to a preset format on the upper layer of the comment message displayed on the live broadcast interface, wherein the preset format comprises at least one of the following formats: font type format, font color format, font size format.

11. The method of claim 9, wherein the displaying, according to the type of the comment intention prompting condition and in a display manner matched with the type when the comment intention prompting condition is determined to be satisfied, the user intention prompting message generated according to the comment message on the live interface comprises:

receiving trigger operation input by a host for indicating comment intention prompt;

determining that a comment intention prompting condition is met according to the triggering operation, and responding to the triggering operation to obtain a user intention prompting message generated according to the comment message;

and displaying the user intention prompt message on the live broadcast interface through a floating layer.

12. The method of claim 11, wherein the presenting the user intent prompt message on the live interface through a floating layer comprises:

performing visual enhancement processing on the user intention prompting message according to a preset format, and displaying the user intention prompting message on the live broadcast interface through a floating layer, wherein the preset format comprises at least one of the following formats: font type format, font color format, font size format; and/or

Rendering a floating layer on the live broadcast interface, wherein the floating layer comprises a plurality of tab pages, and the tab pages comprise a tab page for displaying the user intention prompt message, a tab page for displaying live broadcast content heat and a tab page for displaying user activity; and displaying the user intention prompting message on the live broadcast interface according to the selection operation of the tab page for displaying the user intention prompting message.

13. The method of claim 12, wherein the triggering operation comprises one of: gesture trigger operation, voice trigger operation and interface option trigger operation.

14. A computer storage medium on which a computer program is stored, which program, when executed by a processor, implements the comment data processing method according to any one of claims 1 to 7; or, when executed, implement a real-time live data processing method as claimed in any of claims 8-13.