CN112465587A

CN112465587A - Image interaction information processing method and system based on e-commerce live broadcast

Info

Publication number: CN112465587A
Application number: CN202011328989.0A
Authority: CN
Inventors: 石伟
Original assignee: Individual
Current assignee: Individual
Priority date: 2020-05-10
Filing date: 2020-05-10
Publication date: 2021-03-09
Also published as: CN111626816A; CN111626816B; CN112465588A

Abstract

The embodiment of the disclosure provides an image interaction information processing method and system based on electronic commerce live broadcast, wherein an image interaction area is determined according to a live broadcast graphic interaction label of online interactive live broadcast commodities, then an interaction behavior sequence and interaction type conversion information of each live broadcast audience account corresponding to the image interaction area are obtained, then a first interaction feature vector of each live broadcast audience account and a second interaction feature vector of each live broadcast audience account are extracted, the first interaction feature vector and the second interaction feature vector are fused to obtain a target interaction feature vector, then live broadcast interaction hotspot nodes of the online interactive live broadcast commodities corresponding to each live broadcast audience account are determined, and accordingly live broadcast commodity recommendation information is pushed. Therefore, the interactive behavior characteristics of the audiences can be effectively mined based on the graph interaction condition in the live broadcast process of the online interactive live broadcast commodities of the audiences, so that the live broadcast commodity recommendation information which is possibly interested in the audiences is pushed for the audiences, and the searching cost of the audiences are reduced.

Description

Image interaction information processing method and system based on e-commerce live broadcast

Technical Field

The disclosure relates to the technical field of electronic commerce and image interaction, in particular to an image interaction information processing method and system based on electronic commerce live broadcast.

Background

With the rapid development of the internet and the mobile communication technology, the live internet video broadcast can release contents such as electronic commerce commodities on the internet on site, and the interactive effect of the electronic commerce commodities is enhanced by utilizing the characteristics of intuition, rapidness, good expression form, rich contents, strong interactivity, unlimited regions, divisible audiences and the like of the internet.

At present, in a live broadcast process of online interactive live broadcast goods initiating interaction, as for audiences, corresponding graphic interaction is usually performed in a video live broadcast area, for example, graphic interaction is initiated for a certain area, a certain node and a certain position, and the graphic interaction usually reflects attention points and interest trends of the audiences in the live broadcast process of the online interactive live broadcast goods. Based on the above, how to effectively mine the interactive behavior characteristics of the audience based on the image interaction condition of the audience in the live broadcast process of the online interactive live broadcast commodity so as to push the live broadcast commodity recommendation information which may be interested by the audience, so as to reduce the search cost and the search cost of the audience, and is a technical problem to be solved urgently in the field.

Disclosure of Invention

In order to overcome at least the above disadvantages in the prior art, an object of the present disclosure is to provide an image interaction information processing method and system based on e-commerce live broadcast, which can effectively mine the interaction behavior characteristics of viewers based on the image interaction condition of the viewers in the live broadcast process of online interactive live broadcast goods, so as to push the recommendation information of live broadcast goods that may be interested in the viewers, so as to reduce the search cost and search cost of the viewers.

In a first aspect, the present disclosure provides an image interaction information processing method based on e-commerce live broadcast, which is applied to a cloud computing platform in communication connection with a plurality of video live broadcast terminals, and the method includes:

acquiring online interactive live broadcast commodities of which the video live broadcast terminals initiate interaction, determining an image interaction area according to a live broadcast graphic interaction label of the online interactive live broadcast commodities, and acquiring an interaction behavior sequence and interaction type conversion information of each live broadcast audience account corresponding to the image interaction area;

respectively inputting the interaction behavior sequence and the interaction type conversion information into an information push network obtained through training, extracting a first interaction feature vector of each live audience account through a first feature vector extraction layer of the information push network, and extracting a second interaction feature vector of each live audience account through a second feature vector extraction layer of the information push network;

fusing the first interaction feature vector and the second interaction feature vector through a feature vector fusion layer of the information push network to obtain a target interaction feature vector;

and determining live broadcast interaction hotspot nodes of the live broadcast audience accounts corresponding to the online interactive live broadcast commodities according to the target interaction characteristic vectors, respectively generating live broadcast commodity recommendation information of the corresponding live broadcast audience accounts according to the live broadcast interaction hotspot nodes, and sending the live broadcast commodity recommendation information to the corresponding video live broadcast terminals.

In a possible implementation manner of the first aspect, the interaction type conversion information includes an interaction node, a conversion type, and an interaction image position;

the step of extracting the first interactive feature vector of each live audience account through the first feature vector extraction layer of the information push network and extracting the second interactive feature vector of each live audience account through the second feature vector extraction layer of the information push network comprises the following steps:

inputting the interaction behavior sequence into a first feature vector extraction layer, and performing feature extraction on interaction behaviors in the interaction behavior sequence to obtain corresponding interaction behavior features;

performing feature migration processing on the interaction behavior features by using the first feature vector extraction layer and interaction deviation parameters corresponding to the live broadcast graph interaction tags to obtain interaction behavior features after feature migration processing;

extracting a first interaction feature vector of each live audience account according to the interaction behavior feature after the feature migration processing; and

inputting the interaction type conversion information into a second feature vector extraction layer, and performing feature extraction on the interaction type conversion information to obtain interaction node features, interaction image position features and conversion type features;

performing feature migration processing on the interactive node feature, the interactive image position feature and the conversion type feature by using the second feature vector extraction layer and the interactive bias parameter corresponding to the live image interactive label to obtain an interactive type conversion information array;

and acquiring the interactive behavior characteristics corresponding to the interactive behavior sequence, inputting the interactive behavior characteristics into the interactive type conversion information array for characteristic fusion to obtain a fused characteristic vector array, and extracting a second interactive characteristic vector of each live audience account according to the characteristic vector array.

In a possible implementation manner of the first aspect, the step of fusing the first interaction feature vector and the second interaction feature vector by a feature vector fusion layer of the information push network to obtain a target interaction feature vector includes:

and fusing the feature vector nodes corresponding to the first interaction feature vector and the second interaction feature vector one by one respectively through a feature vector fusion layer of the information push network to obtain a target interaction feature vector.

In a possible implementation manner of the first aspect, the step of determining, according to the target interaction feature vector, a live broadcast interaction hotspot node of each live broadcast audience account corresponding to the online interactive live broadcast commodity includes:

acquiring interactive item feature data corresponding to interactive items participating in interaction of the online interactive live broadcast commodities from the target interactive feature vector, wherein the interactive item feature data are obtained by performing feature embedded expression on interactive item process vectors in the target interactive feature vector by adopting a feature expression form matched with interactive item associated commodities of corresponding interactive items;

performing feature mapping on the interactive item feature data sent by each corresponding interactive item according to a feature mapping mode respectively matched with each feature expression form to obtain a corresponding interactive item process vector;

respectively analyzing the interaction rule of each interaction item process vector, and determining the interest degree of the interaction rule corresponding to each interaction item, wherein the interest degree of the interaction rule is used for reflecting the interest degree of the interaction items participating in the online interaction live broadcast commodities;

screening out the highest interest degree of the interaction rule from the interest degrees of the interaction rules corresponding to the interaction items, and determining the comparison degrees of the interaction rules corresponding to the interaction items according to the comparison values between the interest degrees of the interaction rules corresponding to the interaction items and the highest interest degrees of the interaction rules; the interactive rule comparison degree corresponding to the interactive item is in positive correlation with the corresponding contrast value;

and performing node source tracing on an interactive item process vector of an interactive item with the interactive rule comparison degree larger than the set interactive rule comparison degree to obtain live broadcast interactive hotspot nodes of the online interactive live broadcast commodities corresponding to the live broadcast audience accounts, wherein the live broadcast interactive hotspot nodes are used for representing interactive objects or interactive results in the interactive process.

In a possible implementation manner of the first aspect, the step of performing interaction rule analysis on each interaction item process vector and determining an interest degree of an interaction rule corresponding to each interaction item includes:

dividing each interactive item process vector into unit vector sequences with more than one vector unit, detecting the interactive rule of each unit vector sequence, determining the number of interactive process nodes with interactive frequent times larger than set frequent times in the included unit vector sequences for each path of interactive item process vector, determining the occupation ratio of the interactive process nodes for each path of interactive item process vector according to the number of the interactive process nodes in the interactive item process vector and the total number of the unit vector sequences included in the interactive item process vector, and determining the interest degree of the interactive rule corresponding to each interactive item according to the occupation ratio of the interactive process nodes; or

Dividing each interactive project process vector into unit vector sequences of more than one vector unit, performing interactive rule detection on each unit vector sequence, determining interactive process nodes with interactive frequent times larger than set frequent times in the unit vector sequences, determining interactive continuous quantity corresponding to each interactive process node, and determining interactive rule interestingness corresponding to each interactive project according to the quantity of effective interactive process nodes with interactive continuous quantity larger than or equal to an energy threshold value in the interactive process nodes included in each interactive project process vector; or

Dividing each interactive project process vector into unit vector sequences of more than one vector unit, calculating interest weights of the directed weighted graphs corresponding to the unit vector sequences respectively, carrying out weighted summation on the interest weights of the directed weighted graphs corresponding to the unit vector sequences included in the interactive project process vectors for each path of interactive project process vectors to obtain description vector values corresponding to the interactive project process vectors, and taking the description vector values corresponding to the interactive project process vectors as the interest degrees of the interactive rules corresponding to the interactive projects.

In a possible implementation manner of the first aspect, the step of dividing each interactive item process vector into unit vector sequences of more than one vector unit, and calculating interest weights of the directed weighted graph corresponding to each unit vector sequence includes:

for each interactive project process vector corresponding to each interactive member, dividing the corresponding interactive project process vector into unit vector sequences of more than one vector unit and positioned in a directed space corresponding to a directed weighted graph;

generating interest thermodynamic diagrams corresponding to calculation results of graph nodes in the directed weighted graph of each unit vector sequence, and determining more than one interest thermodynamic unit included in the interest thermodynamic diagrams respectively corresponding to each unit vector sequence;

for each interested thermal unit in each unit vector sequence, determining an interested thermal unit thermal map corresponding to the interested thermal unit based on the thermal value of the interested thermal point included in the interested thermal unit;

for a current interest thermal unit in a current vector sequence processed currently in each unit vector sequence, determining a preset number of associated interest thermal units associated with the current interest thermal unit in the current vector sequence, forming an interest thermal unit sequence by the associated interest thermal units and the current interest thermal unit together, and performing weighted summation processing on an interest thermal unit thermal map of each interest thermal unit in the interest thermal unit sequence according to a weight corresponding to the interest thermal unit sequence to obtain a single-heat-encoding thermal map corresponding to the current interest thermal unit in the current vector sequence;

weighting and summing the one-hot coded thermal map of the past interest thermal unit corresponding to the same interest thermal unit serial number in the previous frame of the current vector sequence and the one-hot coded thermal map of the current interest thermal unit in the current vector sequence to obtain an interest relationship thermal map corresponding to the current interest thermal unit;

screening out a minimum heat value from interest relationship thermal maps corresponding to interest thermal units with the same interest thermal unit serial number in different unit vector sequences as a thermal comparison value corresponding to each interest thermal unit with the corresponding interest thermal unit serial number, and regarding a current interest thermal unit in a current vector sequence currently processed in each unit vector sequence, taking a quotient of the interest relationship thermal map of the current interest thermal unit and the thermal comparison value as a thermal strength ratio corresponding to the current interest thermal unit in the current vector sequence;

when the thermal strength ratio is larger than a preset threshold value, taking a first preset numerical value as an interaction rule reference value corresponding to the current interested thermal unit in the current vector sequence;

when the thermal strength ratio is smaller than or equal to the preset threshold, taking a second preset numerical value as an interaction rule reference value corresponding to the current interested thermal unit in the current vector sequence; the second preset value is smaller than the first preset value;

acquiring the interaction rule density of a past interest thermal unit corresponding to the same interest thermal unit serial number as the current interest thermal unit in a past vector sequence before the current vector sequence, and performing weighted summation processing on the interaction rule density corresponding to the past interest thermal unit and the interaction rule reference value corresponding to the current interest thermal unit to obtain the interaction rule density corresponding to the current interest thermal unit in the current vector sequence;

taking the difference value between the first preset density and the interaction rule density as the reference density corresponding to the corresponding interest thermal unit;

for the current interested thermal unit in the current vector sequence processed currently in each unit vector sequence, acquiring the density estimated value corresponding to the past interested thermal unit with the same interested thermal unit serial number as the current interested thermal unit in the past vector sequence of the current vector sequence, and a first product of the density estimation value corresponding to the past interest thermal unit and the interaction rule density corresponding to the current interest thermal unit in the current vector sequence, summing the second product of the thermal map of the interest thermal unit corresponding to the current interest thermal unit in the current vector sequence and the reference density to obtain a density estimation value corresponding to the current interest thermal unit in the current vector sequence, and determining an interest thermal unit description value corresponding to each interest thermal unit based on the thermal map of the interest thermal unit and the density estimation value;

and calculating interest weights of the directed weighted graphs corresponding to the unit vector sequences respectively according to the interest thermal unit description values corresponding to the interest thermal units included in the unit vector sequences respectively.

In a possible implementation manner of the first aspect, the step of generating live broadcast commodity recommendation information of each corresponding live broadcast audience account according to the live broadcast interaction hotspot node includes:

acquiring to-be-processed hotspot interaction information corresponding to the live broadcast interaction hotspot node, wherein the to-be-processed hotspot interaction information comprises at least one hotspot interaction record;

calculating a recommendation category confidence corresponding to the to-be-processed hotspot interaction information, wherein the recommendation category confidence represents the probability that the to-be-processed hotspot interaction information belongs to each set recommendation category in a to-be-recommended simulation process;

if the recommendation type confidence is greater than or equal to a set confidence threshold, calculating a recommendation range degree set of the hotspot interaction information to be processed in a formal recommendation process, wherein the recommendation range degree set comprises at least one of a target total recommendation degree and a target unit recommendation degree, the target total recommendation degree represents the probability that the hotspot interaction information to be processed belongs to each set recommendation type, and the target unit recommendation degree represents the probability that a hotspot interaction record corresponding to the maximum unit recommendation degree in the hotspot interaction information to be processed belongs to each set recommendation type;

and determining a hotspot interaction analysis result corresponding to the hotspot interaction information to be processed according to the recommendation scope degree set, and generating corresponding live broadcast commodity recommendation information of each live broadcast audience account according to the hotspot interaction analysis result.

In a possible implementation manner of the first aspect, the step of calculating a recommendation category confidence corresponding to the interaction information of the hotspot to be processed includes:

extracting a first interaction relation vector sequence corresponding to the hotspot interaction information to be processed, wherein the first interaction relation vector sequence comprises at least one first interaction relation vector, and each first interaction relation vector corresponds to one hotspot interaction record;

extracting a first interaction relation preference vector value sequence corresponding to the first interaction relation vector sequence, wherein the first interaction relation preference vector value sequence comprises at least one first interaction relation preference vector value, and each first interaction relation preference vector value corresponds to one first interaction relation vector;

generating a second interaction relation vector sequence according to the first interaction relation preference vector value sequence and the first interaction relation vector sequence, wherein the second interaction relation vector sequence comprises at least one second interaction relation vector, and each second interaction relation vector corresponds to a hotspot interaction record;

extracting a third interaction relation vector sequence corresponding to the second interaction relation vector sequence, wherein the third interaction relation vector sequence comprises at least one third interaction relation vector, and each third interaction relation vector corresponds to one second interaction relation vector;

extracting a first feature vector set corresponding to the third interaction relation vector sequence, wherein the first feature vector set comprises at least one first feature vector, and each first feature vector corresponds to one third interaction relation vector;

performing feature fusion on the first feature vector set to obtain a second feature vector;

and calculating a recommendation category confidence corresponding to the second feature vector, wherein the recommendation category confidence represents the probability that the to-be-processed hotspot interaction information belongs to each set recommendation category in the to-be-recommended simulation process.

In a possible implementation manner of the first aspect, the recommendation scope degree set includes the target unit recommendation degree; the step of calculating the recommendation scope degree set of the hotspot interaction information to be processed in the formal recommendation process comprises the following steps: calculating a target unit recommendation degree of the hotspot interaction information to be processed in the formal recommendation process, wherein the target unit recommendation degree is the maximum value in a unit recommendation degree set, the unit recommendation degree set comprises at least one unit recommendation degree, and each unit recommendation degree corresponds to one hotspot interaction record; the step of determining the hotspot interaction analysis result corresponding to the to-be-processed hotspot interaction information according to the recommendation scope set comprises the following steps: if the target unit recommendation degree is greater than or equal to a second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to first type hotspot interaction information; if the target unit recommendation degree is smaller than the second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to second type hotspot interaction information; or

The recommendation scope degree set comprises the target total recommendation degree; the calculating of the recommendation scope degree set of the to-be-processed hotspot interaction information in the formal recommendation process comprises the following steps: acquiring the target total recommendation degree of the hotspot interaction information to be processed in the formal recommendation process; the step of determining the hotspot interaction analysis result corresponding to the to-be-processed hotspot interaction information according to the recommendation scope set comprises the following steps: if the target total recommendation degree is greater than or equal to a second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to first type hotspot interaction information; if the target total recommendation degree is smaller than the second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to a second type of hotspot interaction information; or

The recommendation scope degree set comprises the target unit recommendation degree and the target total recommendation degree; the step of calculating the recommendation scope degree set of the hotspot interaction information to be processed in the formal recommendation process comprises the following steps: acquiring the target unit recommendation degree and the target total recommendation degree of the hotspot interaction information to be processed in the formal recommendation process, wherein the target unit recommendation degree is the maximum value in a unit recommendation degree set, the unit recommendation degree set comprises at least one unit recommendation degree, and each unit recommendation degree corresponds to one hotspot interaction record; the step of determining the hotspot interaction analysis result corresponding to the to-be-processed hotspot interaction information according to the recommendation scope set comprises the following steps: if at least one of the target unit recommendation degree and the target total recommendation degree is greater than or equal to a second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to first type hotspot interaction information; if the target unit recommendation degree and the target total recommendation degree are both smaller than the second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to second type hotspot interaction information;

the second type of hotspot interaction information and the first type of hotspot interaction information belong to different hotspot interaction information, when the to-be-processed hotspot interaction information belongs to the first type of hotspot interaction information, a hotspot interaction analysis result corresponding to the to-be-processed hotspot interaction information is an interaction recommended item associated with the to-be-processed hotspot interaction information, the to-be-processed hotspot interaction information belongs to the second type of hotspot interaction information, and a hotspot interaction analysis result corresponding to the to-be-processed hotspot interaction information is an interaction recommended item associated with the to-be-processed hotspot interaction information and other interaction recommended items associated with the interaction recommended item.

In a possible implementation manner of the first aspect, the information push network is trained by:

acquiring past interaction behavior sequences and past interaction type conversion information of a plurality of users, and generating training data by using the past interaction behavior sequences and the past interaction type conversion information;

acquiring live image interactive labels of a plurality of users, generating training labels by using the live image interactive labels, extracting interactive behavior characteristics of past interactive behavior sequences, and extracting an interactive type conversion information array of past interactive type conversion information;

inputting the interaction behavior characteristics and the interaction type conversion information array into a preset artificial intelligence network to obtain a training result;

and adjusting parameters of the artificial intelligence network and continuing training based on the difference between the training result and the training label until the training condition is met, and obtaining the information push network.

In a second aspect, an embodiment of the present disclosure further provides an image interaction information processing apparatus based on e-commerce live broadcast, which is applied to a cloud computing platform in communication connection with a plurality of video live broadcast terminals, where the apparatus includes:

the acquisition module is used for acquiring online interactive live broadcast commodities of which the video live broadcast terminals initiate interaction, determining an image interaction area according to a live broadcast graphic interaction label of the online interactive live broadcast commodities, and acquiring an interaction behavior sequence and interaction type conversion information of each live broadcast audience account corresponding to the image interaction area;

the extraction module is used for respectively inputting the interaction behavior sequence and the interaction type conversion information into an information push network obtained through training, extracting a first interaction feature vector of each live audience account through a first feature vector extraction layer of the information push network, and extracting a second interaction feature vector of each live audience account through a second feature vector extraction layer of the information push network;

the fusion module is used for fusing the first interaction feature vector and the second interaction feature vector through a feature vector fusion layer of the information push network to obtain a target interaction feature vector;

and the generation module is used for determining live broadcast interaction hotspot nodes of the online interaction live broadcast commodities corresponding to the live broadcast audience accounts according to the target interaction characteristic vectors, respectively generating live broadcast commodity recommendation information of the corresponding live broadcast audience accounts according to the live broadcast interaction hotspot nodes, and sending the live broadcast commodity recommendation information to the corresponding video live broadcast terminals.

In a third aspect, an embodiment of the present disclosure further provides an image interaction information processing system based on e-commerce live broadcast, where the image interaction information processing system based on e-commerce live broadcast includes a cloud computing platform and a plurality of video live broadcast terminals connected to the cloud computing platform in a communication manner;

In a fourth aspect, an embodiment of the present disclosure further provides a cloud computing platform, where the cloud computing platform includes a processor, a machine-readable storage medium, and a network interface, where the machine-readable storage medium, the network interface, and the processor are connected through a bus system, the network interface is used for being communicatively connected with at least one live video terminal, the machine-readable storage medium is used for storing a program, an instruction, or a code, and the processor is used for executing the program, the instruction, or the code in the machine-readable storage medium to execute the method for processing the image interaction information based on live video streaming in any one of the first aspect or any possible design of the first aspect.

In a fifth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, where instructions are stored, and when executed, cause a computer to perform the method for processing image interaction information based on e-commerce live broadcast in the first aspect or any one of the possible designs of the first aspect.

Based on any one of the above aspects, the image interaction area is determined according to the live image interaction tag of the online interactive live broadcast commodity, then the interaction behavior sequence and the interaction type conversion information of each live audience account corresponding to the image interaction area are obtained, then the first interaction feature vector of each live audience account and the second interaction feature vector of each live audience account are extracted, and after the first interaction feature vector and the second interaction feature vector are fused to obtain the target interaction feature vector, the live interaction hotspot node of the online interactive live broadcast commodity corresponding to each live audience account is determined, and therefore the live broadcast commodity recommendation information is pushed. Therefore, the interactive behavior characteristics of the audiences can be effectively mined based on the graph interaction condition in the live broadcast process of the online interactive live broadcast commodities of the audiences, so that the live broadcast commodity recommendation information which is possibly interested in the audiences is pushed for the audiences, and the searching cost of the audiences are reduced.

Drawings

To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present disclosure and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings may be obtained from the drawings without inventive effort.

Fig. 1 is a schematic view of an application scenario of an e-commerce live broadcast-based image interaction information processing system according to an embodiment of the present disclosure;

fig. 2 is a schematic flowchart of an image interaction information processing method based on e-commerce live broadcast according to an embodiment of the present disclosure;

fig. 3 is a schematic functional module diagram of an image interaction information processing apparatus based on e-commerce live broadcast according to an embodiment of the present disclosure;

fig. 4 is a schematic block diagram of a cloud computing platform for implementing the above-mentioned image interaction information processing method based on e-commerce live broadcast according to the embodiment of the present disclosure.

Detailed Description

The present disclosure is described in detail below with reference to the drawings, and the specific operation methods in the method embodiments can also be applied to the device embodiments or the system embodiments.

Fig. 1 is an interaction diagram of an interactive image information processing system 10 based on e-commerce live broadcast according to an embodiment of the present disclosure. The e-commerce live broadcast based image interaction information processing system 10 may include a cloud computing platform 100 and a video live broadcast terminal 200 communicatively connected to the cloud computing platform 100. The live e-commerce based image interaction information processing system 10 shown in fig. 1 is only one possible example, and in other possible embodiments, the live e-commerce based image interaction information processing system 10 may also include only a part of the components shown in fig. 1 or may also include other components.

In this embodiment, the live video terminal 200 may include a mobile device, a tablet computer, a laptop computer, or any combination thereof. In some embodiments, the mobile device may include a smart home device, a wearable device, a smart mobile device, a virtual reality device, an augmented reality device, or the like, or any combination thereof. In some embodiments, the smart home devices may include control devices of smart electrical devices, smart monitoring devices, smart televisions, smart cameras, and the like, or any combination thereof. In some embodiments, the wearable device may include a smart bracelet, a smart lace, smart glass, a smart helmet, a smart watch, a smart garment, a smart backpack, a smart accessory, or the like, or any combination thereof. In some embodiments, the smart mobile device may include a smartphone, a personal digital assistant, a gaming device, and the like, or any combination thereof. In some embodiments, the virtual reality device and/or the augmented reality device may include a virtual reality helmet, virtual reality glass, a virtual reality patch, an augmented reality helmet, augmented reality glass, an augmented reality patch, or the like, or any combination thereof. For example, the virtual reality device and/or augmented reality device may include various virtual reality products and the like.

In this embodiment, the cloud computing platform 100 and the live video terminal 200 in the image interactive information processing system 10 based on e-commerce live broadcast may cooperatively execute the image interactive information processing method based on e-commerce live broadcast described in the following method embodiment, and the specific steps of executing the cloud computing platform 100 and the live video terminal 200 may refer to the detailed description of the following method embodiment.

In order to solve the technical problem in the foregoing background art, fig. 2 is a schematic flowchart of a method for processing image interaction information based on e-commerce live broadcast according to an embodiment of the present disclosure, where the method for processing image interaction information based on e-commerce live broadcast according to the present embodiment may be executed by the cloud computing platform 100 shown in fig. 1, and the method for processing image interaction information based on e-commerce live broadcast is described in detail below.

Step S110, acquiring an online interactive live broadcast product of which the video live broadcast terminal 200 initiates an interaction, determining an image interaction area according to a live broadcast graphic interaction tag of the online interactive live broadcast product, and acquiring an interaction behavior sequence and interaction type conversion information of each live broadcast audience account corresponding to the image interaction area.

And step S120, respectively inputting the interaction behavior sequence and the interaction type conversion information into an information push network obtained through training, extracting a first interaction feature vector of each live audience account through a first feature vector extraction layer of the information push network, and extracting a second interaction feature vector of each live audience account through a second feature vector extraction layer of the information push network.

And step S130, fusing the first interaction feature vector and the second interaction feature vector through a feature vector fusion layer of the information push network to obtain a target interaction feature vector.

Step S140, determining live broadcast interaction hotspot nodes of online interactive live broadcast commodities corresponding to the live broadcast audience accounts according to the target interaction feature vectors, respectively generating live broadcast commodity recommendation information of the corresponding live broadcast audience accounts according to the live broadcast interaction hotspot nodes, and sending the live broadcast commodity recommendation information to the corresponding video live broadcast terminal 200.

In this embodiment, the online interactive live broadcast merchandise may be any merchandise for merchandise promotion and display, such as but not limited to electronic products, agricultural products, infant products, pregnant woman products, and the like.

In this embodiment, the live graphic interaction tag may refer to an interaction control type generated when the viewer initiates a live graphic interaction, for example, may refer to an interaction control type in a certain interaction area of a certain online interactive live commodity, or may also refer to an interaction control type of a certain interaction time node of a certain online interactive live commodity, where the interaction control type may represent a control type in an interaction process, such as "like" or "like".

In this embodiment, the image interaction area may be specifically determined according to a node where the live graphic interaction tag is located, for example, the node where the live graphic interaction tag is located is an interaction area B of the online interactive live commodity a in the live broadcasting process, and then the image interaction area is the interaction area B.

In this embodiment, the interaction behavior sequence may be used to represent a specifically generated interaction behavior (for example, a bullet screen behavior, a gift behavior, and the like), and the interaction type conversion information may be used to represent a forward-backward conversion process of a specifically generated interaction behavior type, for example, information in a process of switching from a bullet screen behavior to a gift behavior.

In this embodiment, the live broadcast interactive hotspot node may be configured to represent point of interest information corresponding to each live broadcast audience account, for example, a related commodity C in a live broadcast process of a certain online interactive live broadcast commodity a, or a temporarily mentioned online interactive live broadcast commodity D released next time, and the like, which is not specifically limited herein.

Based on the above steps, the image interaction area is determined according to the live image interaction tag of the online interactive live broadcast commodity, then the interaction behavior sequence and the interaction type conversion information of each live audience account corresponding to the image interaction area are obtained, then the first interaction feature vector of each live audience account and the second interaction feature vector of each live audience account are extracted, and after the first interaction feature vector and the second interaction feature vector are fused to obtain the target interaction feature vector, the live interaction hotspot node of the online interactive live broadcast commodity corresponding to each live audience account is determined, and therefore the live broadcast commodity recommendation information is pushed. Therefore, the interactive behavior characteristics of the audiences can be effectively mined based on the graph interaction condition in the live broadcast process of the online interactive live broadcast commodities of the audiences, so that the live broadcast commodity recommendation information which is possibly interested in the audiences is pushed for the audiences, and the searching cost of the audiences are reduced.

In a possible implementation manner, for step S110, the interaction type conversion information may specifically include an interaction node, a conversion type, and an interaction image position.

The interactive nodes may be time nodes or area nodes during interactive type conversion, the conversion types may be interactive types before interactive type conversion and interactive types after interactive type conversion, and the interactive image positions may be positions where interactive images are located during interactive type conversion.

On this basis, step S120 may be specifically implemented by the following exemplary sub-steps, which are described in detail below.

And a substep S121, inputting the interaction behavior sequence into a first feature vector extraction layer, and performing feature extraction on the interaction behavior in the interaction behavior sequence to obtain corresponding interaction behavior features.

And a substep S122, performing feature migration processing on the interactive behavior features by using the interactive deviation parameters corresponding to the first feature vector extraction layer and the live broadcast graphic interactive label to obtain the interactive behavior features after the feature migration processing.

And a substep S123 of extracting a first interaction feature vector of each live audience account according to the interaction behavior feature after the feature migration processing.

And a substep S124 of inputting the interactive type conversion information into a second feature vector extraction layer, and performing feature extraction on the interactive type conversion information to obtain interactive node features, interactive image position features and conversion type features.

And a substep S125 of performing feature migration processing on the interactive node feature, the interactive image position feature and the conversion type feature by using the interactive deviation parameters corresponding to the second feature vector extraction layer and the live image interactive label to obtain an interactive type conversion information array.

And a substep S126, acquiring the interactive behavior characteristics corresponding to the interactive behavior sequence, inputting the interactive behavior characteristics into the interactive type conversion information array for characteristic fusion to obtain a fused characteristic vector array, and extracting a second interactive characteristic vector of each live audience account according to the characteristic vector array.

In a possible implementation manner, for step S130, in order to improve the fusion efficiency, in this embodiment, feature vector nodes corresponding to the first interaction feature vector and the second interaction feature vector one to one respectively may be fused through a feature vector fusion layer of the information push network to obtain the target interaction feature vector.

In a possible implementation manner, for step S140, in order to accurately determine the live broadcast interaction hotspot node of each live broadcast audience account corresponding to the online interaction live broadcast commodity, the following exemplary sub-steps may be specifically implemented, and are described in detail as follows.

And a substep S141 of obtaining interactive item feature data corresponding to interactive items participating in the interaction of the online interactive live broadcast commodities from the target interactive feature vector, wherein the interactive item feature data are obtained by performing feature embedded expression on interactive item process vectors in the target interactive feature vector by adopting a feature expression form matched with interactive item associated commodities of corresponding interactive items.

And a substep S142, performing feature mapping on the interactive item feature data sent by the corresponding interactive items according to the feature mapping mode respectively matched with each feature expression form to obtain corresponding interactive item process vectors.

And a substep S143 of analyzing the interaction law of each interaction project process vector respectively and determining the interest degree of the interaction law corresponding to each interaction project. The interest degree of the interaction rule is used for reflecting the interest degree of the interaction items participating in the online interactive live broadcast commodities.

And a substep S144 of screening out the highest interest degree of the interaction rule from the interest degrees of the interaction rules corresponding to the interaction items, and determining the interaction rule comparison degrees corresponding to the interaction items according to the comparison values between the interest degrees of the interaction rules corresponding to the interaction items and the highest interest degrees of the interaction rules. And the interactive rule comparison degree corresponding to the interactive item is in positive correlation with the corresponding contrast value.

And a substep S145 of performing node source tracing on the interactive item process vector of the interactive item with the interactive rule comparison degree greater than the set interactive rule comparison degree to obtain live broadcast interactive hotspot nodes of online interactive live broadcast commodities corresponding to each live broadcast audience account, wherein the live broadcast interactive hotspot nodes are used for representing interactive objects or interactive results in the interactive process.

Exemplarily, the substep S143 can be specifically realized by the following embodiment (1), embodiment (2) or embodiment (3).

(1) Dividing each interactive project process vector into unit vector sequences with more than one vector unit, detecting the interactive rule of each unit vector sequence, determining the number of interactive process nodes with interactive frequent times larger than the set frequent times in the included unit vector sequences for each path of interactive project process vector, determining the proportion of the interactive process nodes for each path of interactive project process vector according to the number of the interactive process nodes in the interactive project process vector and the total number of the unit vector sequences included in the interactive project process vector, and determining the interest degree of the interactive rule corresponding to each interactive project according to the proportion of the interactive process nodes.

(2) Dividing each interactive project process vector into unit vector sequences of more than one vector unit, detecting the interactive rule of each unit vector sequence, determining interactive process nodes with the interactive frequent times larger than the set frequent times in the unit vector sequences, determining the interactive continuous quantity corresponding to each interactive process node, and determining the interactive rule interestingness corresponding to each interactive project according to the quantity of effective interactive process nodes with the interactive continuous quantity larger than or equal to the energy threshold value in the interactive process nodes included in each interactive project process vector.

(3) Dividing each interactive project process vector into unit vector sequences of more than one vector unit, calculating interest weights of the directed weighted graphs corresponding to the unit vector sequences respectively, carrying out weighted summation on the interest weights of the directed weighted graphs corresponding to the unit vector sequences included in the interactive project process vectors for each path of interactive project process vectors to obtain description vector values corresponding to the interactive project process vectors, and taking the description vector values corresponding to the interactive project process vectors as the interactive regular interest degrees corresponding to the interactive projects.

For example, for each interactive project process vector corresponding to each interactive member, the corresponding interactive project process vector may be divided into a unit vector sequence in a directed space corresponding to the directed weighted graph, where the unit vector sequence is more than one vector unit. On the basis, interest thermodynamic diagrams corresponding to the calculation results of the graph nodes in the directed weighted graph of each unit vector sequence can be generated, and more than one interest thermodynamic unit included in the interest thermodynamic diagrams corresponding to each unit vector sequence is determined.

Therefore, for each interested thermal unit in each unit vector sequence, the thermal unit thermal map corresponding to the interested thermal unit is determined based on the thermal force values of the interested thermal units including the interested thermal force points. Then, for the current interest thermal unit in the current vector sequence processed currently in each unit vector sequence, determining a preset number of associated interest thermal units associated with the current interest thermal unit in the current vector sequence, forming an interest thermal unit sequence by the associated interest thermal units and the current interest thermal unit together, and performing weighted summation processing on the interest thermal unit thermal map of each interest thermal unit in the interest thermal unit sequence according to the weight corresponding to the interest thermal unit sequence to obtain a single-heat-encoding thermal map corresponding to the current interest thermal unit in the current vector sequence.

On this basis, the one-hot coded thermal map of the past interest thermal unit corresponding to the same interest thermal unit serial number in the previous frame of the current vector sequence and the one-hot coded thermal map of the current interest thermal unit in the current vector sequence can be subjected to weighted summation processing to obtain the interest relationship thermal map corresponding to the current interest thermal unit in the current vector sequence. Then, screening out the minimum heat value from the interest relationship thermal maps corresponding to the interest thermal units with the same interest thermal unit serial number in different unit vector sequences as the thermal comparison value corresponding to each interest thermal unit with the corresponding interest thermal unit serial number, and regarding the current interest thermal unit in the current vector sequence currently processed in each unit vector sequence, taking the quotient of the interest relationship thermal map of the current interest thermal unit and the thermal comparison value as the thermal strength ratio corresponding to the current interest thermal unit in the current vector sequence.

In this way, when the thermal strength ratio is greater than the preset threshold, the first preset value may be used as the reference value of the interaction rule corresponding to the current interesting thermal unit in the current vector sequence. For another example, when the thermal strength ratio is smaller than or equal to the preset threshold, the second preset value may be used as the reference value of the interaction rule corresponding to the current thermal unit of interest in the current vector sequence. It will be appreciated that the second predetermined value should be less than the first predetermined value.

Then, the interaction law density of the past interest thermal unit corresponding to the same interest thermal unit serial number as the current interest thermal unit in the past vector sequence before the current vector sequence can be obtained, and the interaction law density corresponding to the past interest thermal unit and the interaction law reference value corresponding to the current interest thermal unit are subjected to weighted summation processing to obtain the interaction law density corresponding to the current interest thermal unit in the current vector sequence, so that the difference value between the first preset density and the interaction law density can be used as the reference density corresponding to the corresponding interest thermal unit.

Then, for the current interested thermal unit in the current vector sequence processed currently in each unit vector sequence, obtaining the density estimated value corresponding to the past interested thermal unit with the same interested thermal unit serial number as the current interested thermal unit in the past vector sequence of the current vector sequence, and a first product of the density estimation value corresponding to the past interested thermal unit and the interactive regular density corresponding to the current interested thermal unit in the current vector sequence, and summing the second product of the thermal map of the interest thermal unit corresponding to the current interest thermal unit in the current vector sequence and the reference density to obtain a density estimation value corresponding to the current interest thermal unit in the current vector sequence, and determining the description value of the interest thermal unit corresponding to each interest thermal unit based on the thermal map of the interest thermal unit and the density estimation value. In this way, the interest weights of the directed weighted graph corresponding to each unit vector sequence can be calculated according to the interest thermal unit description values corresponding to the interest thermal units included in each unit vector sequence.

Based on the design, the interest weights of the directed weighted graphs corresponding to the unit vector sequences can be calculated effectively by combining the interest relations, so that the determination of the live interactive hot point nodes of the online interactive live broadcast commodities corresponding to the subsequent live broadcast audience accounts is facilitated.

In a possible implementation manner, still referring to step S140, in the process of respectively generating live broadcast commodity recommendation information of each corresponding live broadcast audience account according to the live broadcast interaction hotspot node, the following sub-steps may be further specifically implemented, and detailed description is as follows.

And a substep S146 of acquiring to-be-processed hotspot interaction information corresponding to the live broadcast interaction hotspot node, wherein the to-be-processed hotspot interaction information comprises at least one hotspot interaction record.

And in the substep S147, calculating a recommendation category confidence corresponding to the to-be-recommended hotspot interaction information, wherein the recommendation category confidence represents the probability that the to-be-recommended hotspot interaction information belongs to each set recommendation category in the to-be-recommended simulation process.

And a substep S148, if the confidence of the recommendation category is greater than or equal to the set confidence threshold, calculating a recommendation range degree set of the hotspot interaction information to be processed in the formal recommendation process, wherein the recommendation range degree set comprises at least one of a target total recommendation degree and a target unit recommendation degree, the target total recommendation degree represents the probability that the hotspot interaction information to be processed belongs to each set recommendation category, and the target unit recommendation degree represents the probability that the hotspot interaction record corresponding to the maximum unit recommendation degree in the hotspot interaction information to be processed belongs to each set recommendation category.

And in the substep S149, determining a hotspot interaction analysis result corresponding to the hotspot interaction information to be processed according to the recommendation scope degree set, and generating corresponding live broadcast commodity recommendation information of each live broadcast audience account according to the hotspot interaction analysis result.

Exemplarily, in the sub-step S147, the following embodiments may be exemplarily implemented.

(1) And extracting a first interaction relation vector sequence corresponding to the hotspot interaction information to be processed, wherein the first interaction relation vector sequence comprises at least one first interaction relation vector, and each first interaction relation vector corresponds to one hotspot interaction record.

(2) And extracting a first interaction relation preference vector value sequence corresponding to the first interaction relation vector sequence, wherein the first interaction relation preference vector value sequence comprises at least one first interaction relation preference vector value, and each first interaction relation preference vector value corresponds to one first interaction relation vector.

(3) And generating a second interaction relation vector sequence according to the first interaction relation preference vector value sequence and the first interaction relation vector sequence, wherein the second interaction relation vector sequence comprises at least one second interaction relation vector, and each second interaction relation vector corresponds to one hotspot interaction record.

(4) And extracting a third interaction relation vector sequence corresponding to the second interaction relation vector sequence, wherein the third interaction relation vector sequence comprises at least one third interaction relation vector, and each third interaction relation vector corresponds to one second interaction relation vector.

(5) And extracting a first feature vector set corresponding to the third interaction relation vector sequence, wherein the first feature vector set comprises at least one first feature vector, and each first feature vector corresponds to one third interaction relation vector.

(6) And performing feature fusion on the first feature vector set to obtain a second feature vector.

(7) And calculating a recommendation category confidence corresponding to the second feature vector, wherein the recommendation category confidence represents the probability that the to-be-processed hotspot interaction information belongs to each set recommendation category in the to-be-recommended simulation process.

In a possible implementation manner, when the recommendation scope degree set includes the target unit recommendation degree, the target unit recommendation degree of the to-be-processed hotspot interaction information in the formal recommendation process may be calculated. The target unit recommendation degree is the maximum value in the unit recommendation degree set, the unit recommendation degree set comprises at least one unit recommendation degree, and each unit recommendation degree corresponds to one hotspot interaction record. Therefore, if the target unit recommendation degree is greater than or equal to the second recommendation degree threshold, it is determined that the hotspot interaction information to be processed belongs to the first type of hotspot interaction information. And if the target unit recommendation degree is smaller than a second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to a second type of hotspot interaction information.

For another example, when the recommendation scope degree set includes the target total recommendation degree, the target total recommendation degree of the hotspot interaction information to be processed in the formal recommendation process may be obtained. Therefore, if the target total recommendation degree is greater than or equal to the second recommendation degree threshold value, it is determined that the hotspot interaction information to be processed belongs to the first type of hotspot interaction information. And if the target total recommendation degree is smaller than a second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to a second type of hotspot interaction information.

For another example, when the recommendation range degree set includes a target unit recommendation degree and a target total recommendation degree, a target unit recommendation degree and a target total recommendation degree of the to-be-processed hotspot interaction information in a formal recommendation process may be obtained, where the target unit recommendation degree is a maximum value in a unit recommendation degree set, the unit recommendation degree set includes at least one unit recommendation degree, and each unit recommendation degree corresponds to one hotspot interaction record. Therefore, if at least one of the target unit recommendation degree and the target total recommendation degree is greater than or equal to the second recommendation degree threshold, it is determined that the hotspot interaction information to be processed belongs to the first type of hotspot interaction information. And if the target unit recommendation degree and the target total recommendation degree are both smaller than a second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to a second type of hotspot interaction information.

In a possible implementation manner, the information push network may be trained by:

(1) and acquiring past interaction behavior sequences and past interaction type conversion information of a plurality of users, and generating training data by using the past interaction behavior sequences and the past interaction type conversion information.

(2) Acquiring live broadcast graphic interactive labels of a plurality of users, generating training labels by using the live broadcast graphic interactive labels, extracting interactive behavior characteristics of past interactive behavior sequences, and extracting an interactive type conversion information array of past interactive type conversion information.

(3) And inputting the interactive behavior characteristics and the interactive type conversion information array into a preset artificial intelligence network to obtain a training result.

(4) And adjusting parameters of the artificial intelligent network and continuing training based on the difference between the training result and the training label until the training condition is met, and obtaining the information push network.

Fig. 3 is a schematic functional module diagram of an image interaction information processing apparatus 300 based on e-commerce live broadcast according to an embodiment of the present disclosure, in this embodiment, functional modules of the image interaction information processing apparatus 300 based on e-commerce live broadcast may be divided according to the method embodiment executed by the cloud computing platform 100, that is, the following functional modules corresponding to the image interaction information processing apparatus 300 based on e-commerce live broadcast may be used to execute the method embodiments executed by the cloud computing platform 100. The image interactive information processing apparatus 300 based on e-commerce live broadcast may include an obtaining module 310, an extracting module 320, a fusing module 330, and a generating module 340, and the functions of the functional modules of the image interactive information processing apparatus 300 based on e-commerce live broadcast are described in detail below.

An obtaining module 310, configured to obtain an online interactive live broadcast product for which the video live broadcast terminal initiates an interaction, determine an image interaction area according to a live broadcast graphical interaction tag of the online interactive live broadcast product, and obtain an interaction behavior sequence and interaction type conversion information of each live broadcast audience account corresponding to the image interaction area. The obtaining module 310 may be configured to perform the step S110, and the detailed implementation of the obtaining module 310 may refer to the detailed description of the step S110.

The extraction module 320 is configured to input the interaction behavior sequence and the interaction type conversion information to an information push network obtained through training, extract a first interaction feature vector of each live audience account through a first feature vector extraction layer of the information push network, and extract a second interaction feature vector of each live audience account through a second feature vector extraction layer of the information push network. The extracting module 320 may be configured to perform the step S120, and the detailed implementation of the extracting module 320 may refer to the detailed description of the step S120.

The fusion module 330 is configured to fuse the first interaction feature vector and the second interaction feature vector through a feature vector fusion layer of the information push network to obtain a target interaction feature vector. The fusion module 330 may be configured to perform the step S130, and the detailed implementation of the fusion module 330 may refer to the detailed description of the step S130.

The generating module 340 is configured to determine, according to the target interaction feature vector, live broadcast interaction hotspot nodes of the online interaction live broadcast commodities corresponding to the live broadcast audience accounts, generate, according to the live broadcast interaction hotspot nodes, live broadcast commodity recommendation information of the corresponding live broadcast audience accounts, and send the live broadcast commodity recommendation information to corresponding video live broadcast terminals. The generating module 340 may be configured to execute the step S140, and the detailed implementation of the generating module 340 may refer to the detailed description of the step S140.

It should be noted that the division of the modules of the above apparatus is only a logical division, and the actual implementation may be wholly or partially integrated into one physical entity, or may be physically separated. And these modules can be realized in the form of software called by processing element; or may be implemented entirely in hardware; and part of the modules can be realized in the form of calling software by the processing element, and part of the modules can be realized in the form of hardware. For example, the obtaining module 310 may be a processing element separately set up, or may be implemented by being integrated into a chip of the apparatus, or may be stored in a memory of the apparatus in the form of program code, and the processing element of the apparatus calls and executes the functions of the obtaining module 310. Other modules are implemented similarly. In addition, all or part of the modules can be integrated together or can be independently realized. The processing element described herein may be an integrated circuit having signal processing capabilities. In implementation, each step of the above method or each module above may be implemented by an integrated logic circuit of hardware in a processor element or an instruction in the form of software.

For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as: one or more Application Specific Integrated Circuits (ASICs), or one or more microprocessors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs), among others. For another example, when some of the above modules are implemented in the form of a processing element scheduler code, the processing element may be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor that can call program code. As another example, these modules may be integrated together, implemented in the form of a system-on-a-chip (SOC).

Fig. 4 illustrates a hardware structure diagram of the cloud computing platform 100 for implementing the control device provided in the embodiment of the present disclosure, and as shown in fig. 4, the cloud computing platform 100 may include a processor 110, a machine-readable storage medium 120, a bus 130, and a transceiver 140.

In a specific implementation process, at least one processor 110 executes computer-executable instructions stored in the machine-readable storage medium 120 (for example, the obtaining module 310, the extracting module 320, the fusing module 330, and the generating module 340 included in the device 300 for processing interactive information based on live video commerce shown in fig. 3), so that the processor 110 can execute the method for processing interactive information based on live video via live video according to the above method embodiment, where the processor 110, the machine-readable storage medium 120, and the transceiver 140 are connected via the bus 130, and the processor 110 can be configured to control the transceiving action of the transceiver 140, so as to perform data transceiving with the video live broadcast terminal 200.

For a specific implementation process of the processor 110, reference may be made to the above-mentioned method embodiments executed by the cloud computing platform 100, and implementation principles and technical effects thereof are similar, and details of this embodiment are not described herein again.

In the embodiment shown in fig. 4, it should be understood that the Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the present invention may be embodied directly in a hardware processor, or in a combination of the hardware and software modules within the processor.

The machine-readable storage medium 120 may comprise high-speed RAM memory and may also include non-volatile storage NVM, such as at least one disk memory.

The bus 130 may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended ISA (EISA) bus, or the like. The bus 130 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, the buses in the figures of the present application are not limited to only one bus or one type of bus.

In addition, the embodiment of the disclosure also provides a readable storage medium, in which computer execution instructions are stored, and when a processor executes the computer execution instructions, the method for processing the image interaction information based on the e-commerce live broadcast is implemented.

The readable storage medium described above may be implemented by any type of volatile or non-volatile memory device or combination thereof, such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disk. Readable storage media can be any available media that can be accessed by a general purpose or special purpose computer.

Finally, it should be noted that: the above embodiments are only used for illustrating the technical solutions of the present disclosure, and not for limiting the same; while the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present disclosure.

Claims

1. An image interaction information processing method based on electronic commerce live broadcast is characterized by being applied to a cloud computing platform in communication connection with a plurality of video live broadcast terminals, and the method comprises the following steps:

determining live broadcast interaction hotspot nodes of the online interactive live broadcast commodities corresponding to the live broadcast audience accounts according to the target interaction feature vectors, respectively generating live broadcast commodity recommendation information of the corresponding live broadcast audience accounts according to the live broadcast interaction hotspot nodes, and sending the live broadcast commodity recommendation information to corresponding video live broadcast terminals;

the online interactive live broadcast commodity refers to a commodity which is randomly advertised and displayed, the live broadcast graph interactive label refers to an interactive control type generated by audiences when the live broadcast graph interaction is initiated, the interactive control type comprises an interactive control type in an interactive area of the online interactive live broadcast commodity or an interactive control type of an interactive time node of the online interactive live broadcast commodity, and the interactive control type represents a control type in an interactive process.

2. The method as claimed in claim 1, wherein the interactive type conversion information includes an interactive node, a conversion type, and an interactive image position, wherein the interactive node is a time node or an area node during the interactive type conversion, the conversion type is an interactive type before the interactive type conversion and an interactive type after the interactive type conversion, and the interactive image position is a position of the interactive image during the interactive type conversion;

3. The method as claimed in claim 1, wherein the step of fusing the first interactive feature vector and the second interactive feature vector to obtain a target interactive feature vector through a feature vector fusion layer of the information push network comprises:

4. The method as claimed in claim 1, wherein the step of determining live broadcast interaction hotspot nodes of live broadcast audience accounts corresponding to the online interactive live broadcast commodities according to the target interaction feature vector comprises:

5. The method as claimed in claim 4, wherein the step of analyzing the interaction rules of the process vectors of the interaction items to determine interest degrees of the interaction rules corresponding to the interaction items comprises:

6. The method as claimed in claim 5, wherein the step of dividing each interactive item process vector into unit vector sequences of more than one vector unit and calculating interest weights of the directed weighted graph corresponding to each unit vector sequence comprises:

7. The method as claimed in claim 1, wherein the step of generating live broadcast commodity recommendation information for each live broadcast audience account according to the live broadcast interaction hotspot node comprises:

8. The method as claimed in claim 7, wherein the method comprises:

the recommendation scope degree set comprises the target unit recommendation degree; the step of calculating the recommendation scope degree set of the hotspot interaction information to be processed in the formal recommendation process comprises the following steps: calculating a target unit recommendation degree of the hotspot interaction information to be processed in the formal recommendation process, wherein the target unit recommendation degree is the maximum value in a unit recommendation degree set, the unit recommendation degree set comprises at least one unit recommendation degree, and each unit recommendation degree corresponds to one hotspot interaction record; the step of determining the hotspot interaction analysis result corresponding to the to-be-processed hotspot interaction information according to the recommendation scope set comprises the following steps: if the target unit recommendation degree is greater than or equal to a second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to first type hotspot interaction information; if the target unit recommendation degree is smaller than the second recommendation degree threshold value, determining that the hotspot interaction information to be processed belongs to second type hotspot interaction information; or

9. The interactive image information processing method based on the e-commerce live broadcast of any one of claims 1-8, wherein the information push network is trained by:

10. An image interaction information processing system based on electronic commerce live broadcast is characterized by comprising a cloud computing platform and a plurality of video live broadcast terminals, wherein the video live broadcast terminals are in communication connection with the cloud computing platform;