CN115391644A

CN115391644A - Conversation recommendation method and device, electronic equipment and storage medium

Info

Publication number: CN115391644A
Application number: CN202210908332.4A
Authority: CN
Inventors: 徐泽坤; 岳文浩
Original assignee: Hisense Visual Technology Co Ltd
Current assignee: Hisense Visual Technology Co Ltd
Priority date: 2022-07-29
Filing date: 2022-07-29
Publication date: 2022-11-25

Abstract

The present disclosure relates to a dialog recommendation method, apparatus, device and medium, comprising: acquiring first voice information of a user; inputting the first voice information and the user portrait into a first topic prediction model to obtain a first predicted topic; inputting the historical conversation, the user portrait and the historical topics into a second topic prediction model to obtain second predicted topics, wherein the historical conversation is historical conversation content of the user and a conversation recommendation system; determining preset recommendation media resources according to the first prediction topic, the second prediction topic and recommendation media resource heat information, and determining target recommendation media resources and recommendation scores corresponding to the target recommendation media resources based on historical conversation and interaction data of the preset recommendation media resources; according to the relationship between the recommendation score of the target recommendation media asset and the preset threshold value, the reply content of the dialogue recommendation system is determined, the deep instant demand of the user is continuously mined, and the dialogue recommendation system is more natural and smooth in the dialogue process with the user.

Description

Conversation recommendation method and device, electronic equipment and storage medium

Technical Field

The present disclosure relates to the field of natural language processing technologies, and in particular, to a dialog recommendation method and apparatus, an electronic device, and a storage medium.

Background

The recommendation system is one of the most obvious successful cases of artificial intelligence in practice. Generally, the main task of recommendation systems is to indicate potential topics of interest to a user, which can provide the user with reasonable recommended content in case of information overload.

The traditional recommendation system needs a large amount of user historical data, and under a Cold Start (Cold Start) scene with missing historical data, the existing recommendation system cannot meet the requirement of accurate recommendation. In conjunction with the rapid development of conversation systems in recent years, a Conversation Recommendation System (CRS) has come into play. In a dialogue recommendation scene, a system needs to guide a user to express own requirements through dialogue and realize accurate recommendation for the user by combining user images.

The existing conversation recommendation method only recommends media resources which meet the user preference and the instant demand based on the historical conversation content of the user, but cannot realize switching and transferring among different topics, so that the conversation process between the user and the system is more rigid.

Disclosure of Invention

In order to solve the technical problem or at least partially solve the technical problem, the present disclosure provides a dialog recommendation method, apparatus, electronic device, and storage medium, which improve a user experience.

In a first aspect, an embodiment of the present disclosure provides a dialog recommendation method, including:

acquiring first voice information of a user;

inputting the first voice information and the user portrait to a first topic prediction model to obtain a first predicted topic, wherein the user portrait is determined by a user based on user preference;

inputting historical conversation, the user portrait and historical topics into a second topic prediction model to obtain second predicted topics, wherein the historical conversation is historical conversation content of the user and a conversation recommendation system;

determining preset recommendation media resources according to the first prediction topic, the second prediction topic and recommendation media resource heat information, and determining target recommendation media resources and recommendation scores corresponding to the target recommendation media resources based on historical conversation and interaction data of each preset recommendation media resource;

and determining the reply content of the dialogue recommendation system according to the relationship between the recommendation score of the target recommendation media asset and a preset threshold value.

As an implementable manner, optionally, the determining the reply content of the dialog recommendation system according to the relationship between the recommendation score of the target recommendation media asset and the preset threshold includes:

when the recommendation score of the target recommendation media asset is smaller than a preset threshold value, generating guidance reply content;

and when the recommendation score of the target recommendation media asset is larger than a preset threshold value, generating explanatory reply content.

As an implementable manner, optionally, the determining the preset recommendation media resource according to the first prediction topic, the second prediction topic and the recommendation media resource popularity information includes:

acquiring a first preset recommendation media resource of a first preset quantity comprising the first predicted topic and the second predicted topic from a media resource recommendation library;

acquiring a second preset number of second preset recommended media assets from the media asset recommendation library according to the popularity information of each recommended media asset in the media asset recommendation library;

and determining preset recommended media resources according to the first preset recommended media resources and the second preset recommended media resources.

acquiring a third preset recommendation media asset with a third preset number, including the first predicted topic and the second predicted topic, from a media asset recommendation library;

according to the popularity information of each third preset recommendation media asset, obtaining a fourth preset number of fourth preset recommendation media assets from the third preset recommendation media assets;

and determining the preset recommended media resources according to the fourth preset recommended media resources.

As an implementable manner, optionally, the determining a target recommended medium resource and a recommendation score corresponding to the target recommended medium resource based on the historical dialogue and the interaction data of each preset recommended medium resource includes:

inputting the historical dialogue and the interaction data of each preset recommendation medium resource into a recommendation model, and determining the recommendation score of each preset recommendation medium resource;

and determining target recommended media assets based on a sorting algorithm according to the recommendation scores of the preset recommended media assets.

As an implementation manner, optionally, when the recommendation score of the target recommended medium asset is smaller than the preset threshold, generating the guided reply content includes:

when the recommendation score of the target recommendation media asset is smaller than a preset threshold value, generating guided reply content according to the historical dialogue and the second prediction topic, wherein the historical dialogue and the second prediction topic are obtained by coding based on a GPT2 model;

when the recommendation score of the target recommendation media asset is greater than a preset threshold value, generating an explanatory reply content, including:

and when the recommendation score of the target recommendation media asset is larger than a preset threshold value, generating explanatory reply content according to a preset reply template and the target recommendation media asset.

As an implementation manner, optionally, after the generating the guided reply content when the recommendation score of the target recommended medium asset is smaller than the preset threshold, the method further includes:

acquiring second voice information corresponding to the guiding reply content of the user;

inputting the second speech information and the user representation to a first topic prediction model.

In a second aspect, an embodiment of the present disclosure provides a dialog recommendation device, including:

the first voice information acquisition module is used for acquiring first voice information of a user;

a first predicted topic module, configured to input the first voice information and a user representation to a first topic prediction model, resulting in a first predicted topic, where the user representation is determined based on user preferences;

the second predicted topic module is used for inputting the historical conversation, the user portrait and the historical topic into a second topic prediction model to obtain a second predicted topic, wherein the historical conversation is the historical conversation content between the user and a conversation recommendation system;

the recommendation media asset determining module is used for determining preset recommendation media assets according to the first prediction topic, the second prediction topic and recommendation media asset popularity information, and determining target recommendation media assets and recommendation scores corresponding to the target recommendation media assets based on historical conversations and interaction data of the preset recommendation media assets;

and the reply content determining module is used for determining the reply content of the dialogue recommending system according to the relationship between the recommending score of the target recommending medium resource and a preset threshold value.

In a third aspect, an embodiment of the present disclosure provides an electronic device, including:

one or more processors;

a storage device to store one or more programs,

when executed by the one or more processors, cause the one or more processors to implement a method as in any one of the first aspects.

In a fourth aspect, the embodiments of the present disclosure provide a computer-readable storage medium, on which a computer program is stored, where the program is executed by a processor to implement the method according to any one of the first aspect.

Compared with the prior art, the technical scheme provided by the embodiment of the disclosure has the following advantages:

according to the conversation recommendation method, the conversation recommendation device, the electronic equipment and the storage medium, firstly, a first prediction topic is determined based on acquired user use information and user portrait, then, a second prediction topic is determined based on historical conversation, the user portrait and the historical topic, then, preset recommendation media resources are determined according to the first prediction topic, the second prediction topic and recommendation media resource heat information, and target recommendation media resources and recommendation scores corresponding to the target recommendation media resources are determined based on interactive data of the historical conversation and each preset recommendation media resource; and finally, determining reply content of the conversation recommendation system according to the relationship between the recommendation score of the target recommendation media asset and a preset threshold, namely adding topic prediction in a conversation recommendation scene, wherein the first prediction topic is related to the user intention, the second prediction topic is a conversation recommendation system guide topic, the conversation recommendation system guides the user to transfer among topics based on the first prediction topic and the second prediction topic, and continuously excavates deep instant requirements of the user, so that the conversation recommendation system is more natural and smooth in a conversation process with the user, media assets meeting the user preference and the instant requirements are finally recommended to the user, and the use experience of the user is improved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and together with the description, serve to explain the principles of the disclosure.

In order to more clearly illustrate the embodiments or technical solutions in the prior art of the present disclosure, the drawings used in the description of the embodiments or prior art will be briefly described below, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive exercise.

Fig. 1 is a schematic view of an application scenario of a dialog recommendation method in an embodiment of the present disclosure;

FIG. 2A is a block diagram of a hardware configuration of a computer device according to one or more embodiments of the present disclosure;

FIG. 2B is a software configuration diagram of a computer device according to one or more embodiments of the present disclosure;

FIG. 2C is a schematic illustration of an icon control interface display of an application program included in a smart device in accordance with one or more embodiments of the present disclosure;

fig. 3A is a flowchart illustrating a dialog recommendation method according to an embodiment of the disclosure;

fig. 3B is a schematic structural diagram of a first topic prediction model provided by an embodiment of the present disclosure;

FIG. 3C is a schematic diagram of a second topic prediction model provided by an embodiment of the present disclosure;

FIG. 3D is a flowchart illustrating another dialog recommendation method provided by embodiments of the present disclosure;

fig. 4A is a flowchart illustrating a further dialog recommendation method provided by an embodiment of the present disclosure;

fig. 4B is a flowchart illustrating a dialog recommendation method according to an embodiment of the present disclosure;

fig. 4C is a flowchart illustrating a further dialog recommendation method provided by an embodiment of the present disclosure;

FIG. 5 is a flowchart illustrating a method for recommending a dialog according to an embodiment of the present disclosure;

FIG. 6 is a flowchart illustrating a method for recommending a dialog according to an embodiment of the present disclosure;

fig. 7 is a schematic structural diagram of a dialog recommendation device according to an embodiment of the present disclosure;

fig. 8 is a schematic structural diagram of an electronic device provided in an embodiment of the present disclosure.

Detailed Description

In order that the above objects, features and advantages of the present disclosure may be more clearly understood, aspects of the present disclosure will be further described below. It should be noted that, in the case of no conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other.

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure, but the present disclosure may be practiced in other ways than those described herein; it is to be understood that the embodiments disclosed in the specification are only a few embodiments of the present disclosure, and not all embodiments.

The terms "first" and "second," etc. in this disclosure are used to distinguish between different objects, rather than to describe a particular order of objects. For example, the first topic prediction model and the second topic prediction model and the like are used for distinguishing different topic prediction models, and are not used for describing a specific order of the topic prediction models.

Aiming at the problems in the prior art, the embodiment of the disclosure acquires the first voice information of the user; inputting the first voice information and the user portrait to a first topic prediction model to obtain a first predicted topic, wherein the user portrait is determined by a user based on user preference; inputting the historical conversation, the user portrait and the historical topics into a second topic prediction module to obtain second predicted topics, wherein the historical conversation is historical conversation content of the user and a conversation recommendation system; determining preset recommendation media resources according to the first prediction topic, the second prediction topic and recommendation media resource heat information, and determining target recommendation media resources and recommendation scores corresponding to the target recommendation media resources based on historical conversation and interaction data of the preset recommendation media resources; according to the relationship between the recommendation score of the target recommendation media asset and a preset threshold, determining the reply content of the conversation recommendation system, namely adding topic prediction in a conversation recommendation scene, guiding a user to transfer among topics, continuously mining deep instant requirements of the user, enabling the conversation recommendation system to be more natural and smooth in a conversation process with the user, and finally recommending media assets which meet the user preference and the instant requirements to the user.

Fig. 1 is a schematic view of an application scenario of a dialog recommendation method in an embodiment of the present disclosure. As shown in fig. 1, the dialog recommendation method may be used in a voice interaction scenario between a user and an intelligent terminal, assuming that the intelligent terminal in the scenario includes an intelligent device 100, when the user wants to interact with the intelligent device through voice, a voice instruction needs to be sent first, and after receiving first voice information of the user, the intelligent device needs to perform semantic understanding on the first voice information, determine a topic corresponding to the first voice information, so that a subsequent intelligent device can determine reply content of a dialog recommendation system according to the topic, and meet the requirements of the user.

The dialog recommendation method provided by the embodiment of the disclosure can be implemented based on a computer device, or a functional module or a functional entity in the computer device.

The computer device may be a Personal Computer (PC), a server, a mobile phone, a tablet computer, a notebook computer, a mainframe computer, and the like, which is not specifically limited in this disclosure.

Fig. 2A is a block diagram of a hardware configuration of a computer device according to one or more embodiments of the present disclosure. As shown in fig. 2A, the computer apparatus includes: at least one of the communicator 220, the detector 230, the external device interface 240, the controller 250, the audio output interface 270, the memory, the power supply, the user interface 280, and in some embodiments, a modem and a display are further included, fig. 2A is only an exemplary illustration, and the hardware configuration of the computer apparatus is not particularly limited by the disclosed embodiments. The controller 250 includes a central processing unit, a video processor, an audio processor, a graphic processor, a RAM, a ROM, a first interface to an nth interface for input/output, among others. The display 260 may be at least one of a liquid crystal display, an OLED display, a touch display, and a projection display, and may also be a projection device and a projection screen. The tuner demodulator 210 receives broadcast television signals through wired or wireless reception, and demodulates audio and video signals, such as EPG audio and video data signals, from a plurality of wireless or wired broadcast television signals. The communicator 220 is a component for communicating with an external device or a server according to various communication protocol types. For example: the communicator may include at least one of a Wifi module, a bluetooth module, a wired ethernet module, and other network communication protocol chips or near field communication protocol chips, and an infrared receiver. The computer device may establish transmission and reception of control signals and data signals with the server or the local control device through the communicator 220. The detector 230 is used to collect signals of the external environment or interaction with the outside. The controller 250 and the tuner-demodulator 210 may be located in different separate devices, that is, the tuner-demodulator 210 may also be located in an external device of the main device where the controller 250 is located, such as an external set-top box.

In some embodiments, controller 250 controls the operation of the computer device and responds to user actions through various software control programs stored in memory. The controller 250 controls the overall operation of the computer device. A user may input a user command on a Graphical User Interface (GUI) displayed on the display 260, and the user input interface receives the user input command through the Graphical User Interface (GUI). Alternatively, the user may input the user command by inputting a specific sound or gesture, and the user input interface receives the user input command by recognizing the sound or gesture through the sensor.

Fig. 2B is a schematic software configuration diagram of a computer device according to one or more embodiments of the present disclosure, and as shown in fig. 2B, the system is divided into four layers, which are, from top to bottom, an Application (Applications) layer (referred to as an "Application layer"), an Application Framework (Application Framework) layer (referred to as a "Framework layer"), an Android runtime (Android runtime) and system library layer (referred to as a "system runtime library layer"), and a kernel layer.

Fig. 2C is a schematic diagram illustrating an interface display of an icon control of an application program included in an intelligent terminal (mainly an intelligent playback device, such as an intelligent television, a digital cinema system, or a video server), according to one or more embodiments of the present disclosure, as shown in fig. 2C, an application layer includes at least one application program that can display a corresponding icon control on a display, such as: the system comprises a live television application icon control, a video on demand VOD application icon control, a media center application icon control, an application center icon control, a game application icon control and the like. The live television application program can provide live television through different signal sources. A video on demand VOD application may provide video from different storage sources. Unlike live television applications, video on demand provides video displays from some storage source. The media center application program can provide various applications for playing multimedia contents. The application program center can provide and store various application programs.

For describing the scheme of the dialog recommendation method in more detail, the following description is made in conjunction with fig. 3A in an exemplary manner, and it is understood that the steps involved in fig. 3A may include more steps or fewer steps in actual implementation, and the order between the steps may also be different, so as to enable the dialog recommendation method provided in the embodiment of the present application to be implemented.

Fig. 3A is a flowchart illustrating a dialog recommendation method according to an embodiment of the present disclosure, where the dialog recommendation method may be implemented by a dialog recommendation device, and the dialog recommendation device may be implemented in hardware and/or software and may be configured in a computer device.

As shown in fig. 3A, the method specifically includes the following steps:

and S10, acquiring first voice information of the user.

In the embodiment of the present disclosure, first voice information of a user is obtained through a user side in a dialog recommendation system, where the first voice information may be first voice information with different intentions, such as query information, chat information, or recommendation information (i.e., a question that needs to be recommended by the dialog system).

And S20, inputting the first voice information and the user portrait into the first topic prediction model to obtain a first predicted topic.

Wherein the user representation user is determined based on user preferences.

In a specific embodiment, as shown in fig. 3B, first, the first speech information and the user image are encoded by a pre-trained Bert model, where the Bert model is a first topic prediction model, and a first predicted topic corresponding to the first speech information is determined based on the encoded first speech information and the user image.

Illustratively, before starting, the dialog recommendation system retrieves the user's viewing record and user profile, the dialog recommendation system defaults to open a dialog of "what you are busy recently", the user gives feedback, the dialog recommendation system retrieves the first voice information fed back by the user, and inputs the first voice information fed back by the user and the retrieved user profile into a first topic prediction model, such as: inputting: user portrait + user first speech information [ "i am watching textbook woolen with children about earth knowledge, and have no curiosity in the world" ], at this time, the first topic prediction model outputs a first prediction topic: the "earth".

And S30, inputting the historical conversation, the user portrait and the historical topics into a second topic prediction model to obtain a second predicted topic.

The historical dialogue is historical dialogue content of a user and a dialogue recommendation system.

In a specific embodiment, as shown in fig. 3C, the pre-trained Bert model encodes the historical dialog, the user portrait, and the historical topic, and outputs the second predicted topic based on the pre-trained Bert model, in which case the second topic prediction model is also a Bert model.

The historical topics comprise a first predicted topic output by the first topic prediction model and a second predicted topic output by the second topic prediction model.

At this time, by inputting the historical dialog, the user figure, and the historical topic into the second topic prediction model, i.e., the input of the second topic prediction model: user portrait + historical topic [ "earth" ] + historical conversation [ "i am having children watching textbook woolen with earth knowledge, the world is curious enough ], and the second topic prediction model outputs a second predicted topic: "textbook".

In the present embodiment, the historical topics input to the second topic prediction model include the first predicted topic output by the first topic prediction model.

In a specific embodiment, the first topic prediction model and the second topic prediction model may be the same model, and the difference is that the model may receive two input vectors and three input vectors, and when the first speech information and the user image are input to the topic prediction model, the topic prediction model outputs the first predicted topic, and when the historical conversation, the user image, and the historical topic are input to the topic prediction model, the topic prediction model outputs the second predicted topic.

S40, determining preset recommendation media resources according to the first prediction topic, the second prediction topic and recommendation media resource heat information, and determining target recommendation media resources and recommendation scores corresponding to the target recommendation media resources based on historical conversation and interaction data of the preset recommendation media resources.

The recommendation media asset popularity information is related to the reading amount, the click amount and the like of the recommendation media assets, and the interactive data can be, for example, a watching sequence, a watching duration, a watching sequence and the like.

In this embodiment, the conversation recommendation system first recalls the recommended media assets once according to the first predicted topic, the second predicted topic and the recommended media asset popularity information, and recalls a part of related preset recommended media assets from the huge recommended media asset library. And then, determining the target recommendation media assets and recommendation scores corresponding to the target recommendation media assets based on the historical conversations and the interaction data of the preset recommendation media assets.

Illustratively, a preset amount of preset recommendation media assets are recalled from a media asset library according to the earth, the textbook and the popularity information of each recommendation media asset, for example, the recalled preset recommendation media assets are respectively media asset 1, media asset 2 and media asset 3, the interaction data corresponding to the media asset 1 is interaction data 1, the interaction data corresponding to the media asset 2 is interaction data 2, the interaction data corresponding to the media asset 3 is interaction data 3, and after the preset recommendation media assets are recalled, the target recommendation media assets and the recommendation scores corresponding to the target recommendation media assets are determined based on historical conversations and the interaction data of each preset recommendation media asset. Illustratively, the input: conversation history [ "i am watching textbook with children about earth knowledge, and have no curiosity in the world" ] + [ interactive data 1, interactive data 2, interactive data 3], and at this time, the determined target recommendation media resources are: XXX, corresponding recommendations were: 0.06214.

and S50, determining the reply content of the dialogue recommendation system according to the relationship between the recommendation score of the target recommendation media asset and a preset threshold value.

After the target recommendation media assets and the recommendation scores corresponding to the target recommendation media assets are obtained, determining reply contents of the conversation recommendation system according to the relation between the recommendation scores corresponding to the target recommendation media assets and a preset threshold value.

Specifically, as shown in fig. 3C, the specific implementation of step S50 includes:

and S51, when the recommendation score of the target recommendation media asset is smaller than a preset threshold value, generating guidance reply content.

And S52, when the recommendation score of the target recommendation media asset is larger than a preset threshold value, generating explanatory reply content.

For example, if the target recommended medium asset determined in step S40 is "XXX", the corresponding recommendation score is: after 0.05214, according to the relationship between the recommendation score and the preset threshold, if the recommendation score is smaller than the preset threshold, at this time, the dialog recommendation system generates a reply content according to the second predicted topic determined in step S30, and performs a chat dialog with the user, for example, if the generated reply content is "kayi, see textbook well, open eye for children", if the target recommendation media asset determined in step S40 is "XXX", the corresponding recommendation score is: after 0.06214, if the recommendation score is greater than the preset threshold value according to the relationship between the recommendation score and the preset threshold value, at this time, the dialog recommendation system outputs an explanatory reply content according to the preset reply template and the target recommendation media asset determined in step S40, for example, the generated reply content is "i highly recommend" XXX ", which is yyyyy".

The conversation recommendation method provided by the embodiment of the disclosure includes the steps that first prediction topics are determined based on acquired user use information and user portrait, second prediction topics are determined based on historical conversation, the user portrait and the historical topics, then preset recommendation media resources are determined according to the first prediction topics, the second prediction topics and recommendation media resource popularity information, and target recommendation media resources and recommendation scores corresponding to the target recommendation media resources are determined based on interactive data of the historical conversation and the preset recommendation media resources; and finally, determining reply content of the conversation recommendation system according to the relationship between the recommendation score of the target recommendation media asset and a preset threshold, namely adding topic prediction in a conversation recommendation scene, wherein the first prediction topic is related to the user intention, the second prediction topic is a conversation recommendation system guide topic, the conversation recommendation system guides the user to transfer among topics based on the first prediction topic and the second prediction topic, and continuously excavates deep instant requirements of the user, so that the conversation recommendation system is more natural and smooth in a conversation process with the user, media assets meeting the user preference and the instant requirements are finally recommended to the user, and the use experience of the user is improved.

Fig. 4A is a schematic flow chart of another conversation recommendation method provided by an embodiment of the present disclosure, and in the embodiment of the present disclosure, as shown in fig. 4A, step S40 includes S41 and S42, where S41 determines preset recommendation media assets according to a first predicted topic, a second predicted topic, and recommendation media asset popularity information, and S42 determines target recommendation media assets and recommendation scores corresponding to the target recommendation media assets based on historical conversations and interaction data of each of the preset recommendation media assets, and a specific implementation manner of S41 includes:

s410, obtaining a first preset recommendation medium resource comprising a first prediction topic and a first preset number of second prediction topics from a medium resource recommendation library.

As an implementation manner, the method for determining the preset recommended media resources includes obtaining a first preset quantity of first preset recommended media resources including a first predicted topic and a second predicted topic from a media resource recommendation library, for example, marking each recommended media resource in the media resource recommendation library to identify a topic corresponding to each recommended media resource, then determining the first predicted topic in step S20 and determining the second predicted topic in step S30, and then searching the first preset quantity of first preset recommended media resources including the first predicted topic and the second predicted topic from the media resource recommendation library according to the first predicted topic and the second predicted topic.

S411, according to the popularity information of each recommended medium resource in the medium resource recommendation base, obtaining a second preset number of second preset recommended medium resources from the medium resource recommendation base.

On the other hand, each recommended medium resource in the medium resource recommendation library corresponds to heat information, sorting is carried out according to the heat information corresponding to each recommended medium resource in the medium resource recommendation library from high to low according to the heat information, and second preset recommended medium resources in a second preset number are selected based on sorting results of each recommended medium resource.

S412, determining the preset recommended media assets according to the first preset recommended media assets and the second preset recommended media assets.

And after the first preset recommended medium resource and the second preset recommended medium resource are obtained, the first preset recommended medium resource and the second preset recommended medium resource form a preset recommended medium resource.

It should be noted that, in a specific implementation manner, a first preset number corresponding to the first preset recommended medium resource and a second preset number corresponding to the second preset recommended medium resource may be flexibly set, and this is not specifically limited in the embodiment of the present disclosure.

Fig. 4B is a schematic flowchart of another dialog recommendation method provided in the embodiment of the present disclosure, where on the basis of the foregoing embodiment, as shown in fig. 4B, another specific implementation manner of step S41 includes:

s413, third preset recommended media resources with third preset quantity comprising the first predicted topic and the second predicted topic are obtained from the media resource recommendation library.

As another implementation manner, the method for determining the preset recommended media resources includes first obtaining a third preset quantity of third preset recommended media resources including the first predicted topic and the second predicted topic from the media resource recommendation library, for example, marking each recommended media resource in the media resource recommendation library to identify a topic corresponding to each recommended media resource, then determining the first predicted topic in step S20 and determining the second predicted topic in step S30, and then searching the third preset quantity of third preset recommended media resources including the first predicted topic and the second predicted topic from the media resource recommendation library according to the first predicted topic and the second predicted topic.

And S414, acquiring fourth preset recommended media assets in a fourth preset number from the third preset recommended media assets according to the popularity information of the third preset recommended media assets.

And S415, determining the preset recommended media assets according to the fourth preset recommended media assets.

After the third preset recommendation media assets with the third preset quantity are determined, according to the heat information of each third preset recommendation media asset, selecting a fourth preset recommendation media asset with a fourth preset quantity from the third preset recommendation media assets, namely, sorting each third preset recommendation media asset from high to low according to the heat information corresponding to each third preset recommendation media asset, selecting the fourth preset recommendation media asset with the fourth preset quantity from the sorted third preset recommendation media assets, and taking the selected fourth preset recommendation media asset with the fourth preset quantity as the preset recommendation media assets.

In this embodiment, first, third preset recommendation media assets are determined based on the first prediction topic and the second prediction topic, then, according to the heat information of each third preset recommendation media asset, a fourth preset number of fourth preset recommendation media assets are selected from the third preset recommendation media assets to serve as the preset recommendation media assets, that is, the determined preset recommendation media assets are related to the first prediction topic, the second prediction topic and the heat information of the recommendation media assets, so that the recommended media assets are more suitable for the user intention.

Fig. 4C is a schematic flowchart of another dialog recommendation method provided in an embodiment of the present disclosure, where in the embodiment of the present disclosure, as shown in fig. 4C, a specific implementation manner of step S42 includes:

and S420, inputting the historical dialogue and the interaction data of each preset recommendation medium resource into a recommendation model, and determining the recommendation score of each preset recommendation medium resource.

And S421, determining the target recommended media assets based on a sorting algorithm according to the recommendation scores of the preset recommended media assets.

As a specific implementation manner, the specific process of determining the target recommendation media asset and the recommendation score corresponding to the target recommendation media asset based on the historical dialogue and the interaction data of each preset recommendation media asset is as follows:

the method comprises the steps of firstly inputting historical conversations and determined interaction data of all preset recommendation media assets into a recommendation model, determining recommendation scores of all the preset recommendation media assets according to the historical conversations and the determined interaction data of all the preset recommendation media assets, after the recommendation scores of all the preset recommendation media assets are determined, sequencing the recommendation scores corresponding to all the preset recommendation media assets from high to low based on a sequencing algorithm, and selecting the preset recommendation media asset with the highest recommendation score as a target recommendation media asset. Because the recommendation score of each preset recommendation medium resource and the historical dialogue and the interaction data corresponding to the preset recommendation medium resources are determined, the preset recommendation medium resources with higher recommendation scores have higher relevance to the intention of the user, the deep instant requirement of the user is mined by the dialogue recommendation system, and the natural fluency of the dialogue between the subsequent dialogue recommendation system and the user is further ensured.

Fig. 5 is a schematic flowchart of another dialog recommendation method provided in the embodiment of the present disclosure, where on the basis of the embodiment corresponding to fig. 3D, as shown in fig. 5, a specific implementation manner of step S51 includes:

and S510, when the recommendation score of the target recommendation media asset is smaller than a preset threshold value, generating guided reply content according to the historical conversation and the second predicted topic.

And the historical conversation and the second predicted topic are obtained by encoding based on a GPT2 model.

In a specific implementation manner, if the recommendation score of the target recommendation media asset is smaller than the preset threshold, the guided reply content is generated according to the historical topic and the second prediction topic, and the generated guided reply content is related to the historical dialogue between the user and the dialogue recommendation system and the second prediction topic, so that the reply content generated by the dialogue recommendation system is more natural and smooth.

The specific implementation manner of step S52 includes:

and S520, when the recommendation score of the target recommendation media asset is larger than a preset threshold value, generating explanatory reply content according to a preset reply template and the target recommendation media asset.

In a specific implementation manner, if the recommendation score of the target recommendation media asset is greater than a preset threshold, generating explanatory reply content according to a preset reply template and the target recommendation media asset, wherein the recommendation score corresponding to the determined target recommendation media asset is greater than the preset threshold, that is, the determined target recommendation media asset is the media asset which is expected to be recommended by the dialogue recommendation system by the user, and at this time, generating explanatory reply content related to the target recommendation media asset, and giving feedback corresponding to the first voice information sent by the user.

Fig. 6 is a schematic flowchart of another dialog recommendation method provided in an embodiment of the present disclosure, where the embodiment of the present disclosure is based on the embodiment corresponding to fig. 5, and as shown in fig. 6, the method further includes:

and S60, acquiring second voice information corresponding to the content of the guided reply of the user.

When the recommendation score of the target recommendation media asset determined in the step S50 is smaller than the preset threshold, at this time, guidance reply content is generated, the user replies based on the guidance reply content, that is, the user sends out second voice information, the second voice information and the user image are input to the first topic prediction model again, and the steps S20 to S50 are executed again until the dialog recommendation system generates explanatory reply content, so that the dialog recommendation is completed.

And S70, inputting the second voice information and the user portrait into the first topic prediction model.

Illustratively, the target recommended medium determined in step S40 is "XXX", and the corresponding recommendation score is: after 0.05214, according to the relation between the recommendation score and the preset threshold, if the recommendation score is smaller than the preset threshold, at this time, the conversation recommendation system generates a reply content according to the second predicted topic determined in step S30, and carries out chat conversation with the user, for example, the generated reply content is "kao, see that the textbook is well-laid, open the eyeground for children", at this time, the user carries out a second voice reply based on the guided reply content: "yes, child likes to read but prefers to read animation, you can help me recommend a movie of animation class", that is, "yes" is the second voice message, child likes to read but prefers to read animation, you can help me recommend a movie of animation class ", at this moment, portrait the user + the second voice message [" is, child likes to read but prefers to read animation, you can help me recommend a movie of animation class? "] inputting to the first topic prediction model, obtaining again the first predicted topic: [ "animation" ], then step S30 is executed, user portrait + historical topic [ "earth", "textbook", "animation" ] + historical conversation [ "i am having children watching textbook with earth knowledge, almost nothing in the world", "kayao, seeing textbook well, opening eyes for children", "yes, children like to watch animation, but prefer to watch movie, can you recommend movie of an animation class? Inputting "] into a second topic prediction model, and obtaining a second predicted topic again: and [ "animation" ], then, executing step S40 again, recalling a preset number of preset recommendation media assets from the media asset library according to the animation and the popularity information of each recommendation media asset, for example, the recalled preset recommendation media assets are respectively media asset 4, media asset 5 and media asset 6, the interaction data corresponding to the media asset 4 is interaction data 4, the interaction data corresponding to the media asset 5 is interaction data 5, the interaction data corresponding to the media asset 6 is interaction data 6, and after recalling the preset recommendation media assets, determining target recommendation media assets and recommendation scores corresponding to the target recommendation media assets based on historical conversations and the interaction data of each preset recommendation media asset. Illustratively, the input: history of conversation [ "do i have children watching textbook woolen with earth knowledge, have no curiosity at all around the world", "take a joss, see textbook very well, open the eye for children", "yes, and do children like to see animation even though they like to see book, but rather can help us recommend an animation-like movie? "] + [ interactive data 4, interactive data 5, interactive data 6], at this time, the determined target recommendation media asset is: YYYY, the corresponding recommended scores are: 0.06214, at this time, the recommendation score corresponding to the target recommendation media asset "YYYY" meets the preset threshold, and the explanatory reply content is generated: "I recommend YYYY very much, this part is the memory of our time" ], end the user's conversation with the conversation recommendation system.

Fig. 7 is a schematic structural diagram of a dialog recommendation device according to an embodiment of the present disclosure, and as shown in fig. 7, the dialog recommendation device includes:

the first voice information obtaining module 710 is configured to obtain first voice information of a user.

A first predicted topic module 720, configured to input the first voice information and the user representation to a first topic prediction model, resulting in a first predicted topic, where the user representation is determined based on user preferences.

And the second predicted topic module 730 is configured to input the historical conversation, the user portrait and the historical topic into a second topic prediction model to obtain a second predicted topic, where the historical conversation is the historical conversation content between the user and the conversation recommendation system.

And the recommended media asset determining module 740 is configured to determine preset recommended media assets according to the first predicted topic, the second predicted topic and the recommended media asset popularity information, and determine target recommended media assets and recommendation scores corresponding to the target recommended media assets based on historical conversations and interaction data of each preset recommended media asset.

And the reply content determining module 750 is configured to determine the reply content of the dialog recommendation system according to a relationship between the recommendation score of the target recommendation media asset and a preset threshold.

The conversation recommendation device provided by the embodiment of the disclosure determines a first prediction topic based on the acquired information used by the user and the user portrait, determines a second prediction topic based on the historical conversation, the user portrait and the historical topic, determines preset recommendation media resources according to the first prediction topic, the second prediction topic and the recommendation media resource popularity information, and determines target recommendation media resources and recommendation scores corresponding to the target recommendation media resources based on the historical conversation and interactive data of the preset recommendation media resources; and finally, determining reply content of the conversation recommendation system according to the relationship between the recommendation score of the target recommendation media asset and a preset threshold, namely adding topic prediction under a conversation recommendation scene, wherein the first prediction topic is related to the user intention, the second prediction topic is a conversation recommendation system guide topic, the conversation recommendation system guides the user to transfer among topics based on the first prediction topic and the second prediction topic, and continuously excavates deep instant requirements of the user, so that the conversation recommendation system is more natural and smooth in a conversation process with the user, and finally, the media asset meeting the user preference and the instant requirements is recommended to the user.

As an implementable manner, optionally, the implementation manner of the reply content determining module includes:

As an implementable manner, optionally, the recommended media asset determining module includes a preset recommended media asset determining unit and a target recommended media asset determining unit;

one implementation manner of the preset recommended medium resource determining unit comprises the following steps:

acquiring a first preset recommendation media asset of a first preset quantity comprising a first prediction topic and a second prediction topic from a media asset recommendation library;

acquiring a second preset number of second recommended media assets from the media asset recommendation library according to the popularity information of each recommended media asset in the media asset recommendation library;

and determining the preset recommended media assets according to the first preset recommended media assets and the second preset recommended media assets.

Another possible implementation manner of the preset recommended medium resource determining unit includes:

acquiring a third preset recommendation media asset of a third preset quantity comprising the first predicted topic and the second predicted topic from the media asset recommendation library;

according to the popularity information of each third preset recommendation media asset, a fourth preset recommendation media asset with a fourth preset quantity is obtained from the third preset recommendation media assets;

One possible implementation manner of the target recommendation media resource determining unit comprises the following steps:

and determining the target recommended media assets based on a sorting algorithm according to the recommendation scores of the preset recommended media assets.

As an implementation manner, optionally, the implementation manner of the reply content determining module includes:

As an implementable manner, optionally, further comprising:

acquiring second voice information corresponding to the content replied by the user based on the guidance;

the second speech information and the user representation are input to a first topic prediction model.

The device provided by the embodiment of the invention can execute the method provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

It should be noted that, in the embodiment of the apparatus, the included units and modules are merely divided according to functional logic, but are not limited to the above division as long as the corresponding functions can be implemented; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.

The present disclosure also provides an electronic device, including: a processor for executing a computer program stored in a memory, the computer program, when executed by the processor, implementing the steps of the above-described method embodiments.

Fig. 8 is a schematic structural diagram of an electronic device provided by the present disclosure, and fig. 8 shows a block diagram of an exemplary electronic device suitable for implementing embodiments of the present invention. The electronic device shown in fig. 8 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present invention.

As shown in fig. 8, the electronic device 800 is in the form of a general purpose computing device. The components of the electronic device 800 may include, but are not limited to: one or more processors 810, a system memory 820, and a bus 830 that couples various system components including the system memory 820 and the processors.

Bus 830 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, and a processor or local bus using any of a variety of bus architectures. By way of example, such architectures include, but are not limited to, industry Standard Architecture (ISA) bus, micro-channel architecture (MAC) bus, enhanced ISA bus, video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

Electronic device 800 typically includes a variety of computer system readable media. Such media may be any available media that is accessible by electronic device 800 and includes both volatile and nonvolatile media, removable and non-removable media.

The system memory 820 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM) 840 and/or cache memory 850. The electronic device 800 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 860 may be used to read from and write to non-removable, nonvolatile magnetic media (commonly referred to as a "hard disk drive"). A magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 830 by one or more data media interfaces. System memory 820 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the invention.

A program/utility 880 having a set (at least one) of program modules 870, which may include but are not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination thereof may comprise an implementation of a network environment, may be stored in, for example, system memory 820. Program modules 870 generally perform the functions and/or methodologies of embodiments described herein.

Processor 810 performs various functional applications and information processing, such as implementing method embodiments provided by embodiments of the present invention, by executing at least one program of the plurality of programs stored in system memory 820.

The present disclosure also provides a computer-readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.

Any combination of one or more computer-readable media may be employed. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).

The present disclosure also provides a computer program product which, when run on a computer, causes the computer to perform the steps of implementing the above-described method embodiments.

It is noted that, in this document, relational terms such as "first" and "second," and the like, may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of additional like elements in a process, method, article, or apparatus that comprises the element.

The foregoing are merely exemplary embodiments of the present disclosure, which enable those skilled in the art to understand or practice the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the disclosure. Thus, the present disclosure is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A conversation recommendation method, comprising:

acquiring first voice information of a user;

inputting historical conversation, the user portrait and historical topics into a second topic prediction model to obtain a second predicted topic, wherein the historical conversation is historical conversation content of the user and a conversation recommendation system;

and determining the reply content of the dialogue recommendation system according to the relation between the recommendation score of the target recommendation media asset and a preset threshold value.

2. The method of claim 1, wherein the determining the reply content of the dialog recommendation system according to the relationship between the recommendation score of the target recommendation media asset and a preset threshold comprises:

3. The method of claim 1, wherein the determining a preset recommendation media based on the first predicted topic, the second predicted topic, and recommendation media popularity information comprises:

4. The method of claim 1, wherein the determining a preset recommendation media based on the first predicted topic, the second predicted topic, and recommendation media popularity information comprises:

acquiring a third preset recommendation media resource comprising the first predicted topic and a third preset number of the second predicted topic from a media resource recommendation library;

5. The method of claim 1, wherein determining the target recommended media asset and the recommendation score corresponding to the target recommended media asset based on the historical dialogue and the interaction data of each of the preset recommended media assets comprises:

6. The method of claim 2, wherein generating the guided reply content when the recommendation score of the target recommendation media asset is smaller than a preset threshold comprises:

when the recommendation score of the target recommendation media asset is smaller than a preset threshold value, generating guided reply content according to the historical dialogue and the second prediction topic, wherein the historical dialogue and the second prediction topic are obtained by encoding based on a GPT2 model;

7. The method of claim 2, wherein after generating the guided reply content when the recommendation score of the target recommendation media asset is less than a preset threshold, the method further comprises:

8. A conversation recommendation apparatus, comprising:

the recommendation media asset determination module is used for determining preset recommendation media assets according to the first prediction topic, the second prediction topic and recommendation media asset popularity information, and determining target recommendation media assets and recommendation scores corresponding to the target recommendation media assets based on historical conversations and interaction data of the preset recommendation media assets;

and the reply content determining module is used for determining the reply content of the dialogue recommending system according to the relationship between the recommendation score of the target recommendation media asset and a preset threshold value.

9. An electronic device, comprising:

one or more processors;

a storage device for storing one or more programs,

when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-7.

10. A computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, is adapted to carry out the method of any one of claims 1 to 7.