CN113609266A

CN113609266A - Resource processing method and device

Info

Publication number: CN113609266A
Application number: CN202110780305.9A
Authority: CN
Inventors: 施晨; 胡于响; 张增明; 邵亮; 姜飞俊
Original assignee: Alibaba Singapore Holdings Pte Ltd
Current assignee: Alibaba Innovation Co
Priority date: 2021-07-09
Filing date: 2021-07-09
Publication date: 2021-11-05

Abstract

An embodiment of the present specification provides a resource processing method and a device, where the resource processing method includes: the method comprises the steps of receiving a resource obtaining request carrying resources to be obtained, determining historical preference information of a user for each resource category under the condition that the resources to be obtained correspond to at least two resource categories, obtaining current resource categories contained in a target conversation turn in a current conversation of the user, and determining the target resource categories of the resources to be obtained based on the resource obtaining request, the historical preference information and the current resource categories.

Description

Resource processing method and device

Technical Field

The embodiment of the specification relates to the technical field of computers, in particular to a resource processing method. One or more embodiments of the present specification also relate to a resource processing apparatus, a computing device, and a computer-readable storage medium.

Background

Along with the development of science and technology, more and more artificial intelligence dialogue equipment comes with it, and the user can carry out the interactive mode of dialogue interaction or action with artificial intelligence dialogue equipment, lets artificial intelligence dialogue equipment carry out the broadcast of video, audio for the user, or carries out weather report etc..

However, with the continuous expansion of service items that can be provided for users in the artificial intelligence dialog devices, the access of third-party applications or related skills in the artificial intelligence dialog devices is increasing, and often, the same type of applications or skills are also increasing gradually in the same artificial intelligence device. Many times, multiple applications or skills may be accommodated for the same request by the user. Therefore, how to select an application or skill more accurately and more suitable for the user requirement for the user to provide a corresponding service for the user according to the request of the user becomes a problem which needs to be solved urgently to improve the user experience.

Disclosure of Invention

In view of this, the embodiments of the present specification provide a resource processing method. One or more embodiments of the present specification relate to a resource processing apparatus, a computing device, and a computer-readable storage medium to solve technical problems in the prior art.

According to a first aspect of embodiments of the present specification, there is provided a resource processing method, including:

receiving a resource acquisition request carrying resources to be acquired, and determining historical preference information of a user on each resource category under the condition that the resources to be acquired correspond to at least two resource categories;

acquiring a current resource category contained in a target conversation turn in a current conversation of a user;

and determining the target resource category of the resource to be acquired based on the resource acquisition request, the historical preference information and the current resource category.

Optionally, the determining a target resource category of the resource to be acquired based on the resource acquisition request, the historical preference information, and the current resource category includes:

inputting the resource acquisition request, the historical preference information and the current resource category into a prediction model for probability prediction to generate a probability prediction result of negative feedback submitted by the user to each resource category;

and determining the target resource category of the resource to be obtained according to the probability prediction result.

Optionally, the performing probability prediction on the resource acquisition request, the historical preference information, and the current resource category input prediction model to generate a probability prediction result of negative feedback submitted by the user for each resource category includes:

taking the resource acquisition request, the historical preference information and the current resource category as an input set, and inputting a vector coding module of a prediction model for coding to generate a coding vector of the input set;

and inputting the coding vector into a probability prediction module of the prediction model for probability prediction to generate a probability prediction result of negative feedback submitted by the user to each resource category.

Optionally, the inputting the resource obtaining request, the historical preference information, and the current resource category as an input set, and performing encoding processing by using a vector encoding module of a prediction model to generate an encoding vector of the input set includes:

inputting the resource acquisition request into a first vector coding module of a prediction model for coding to generate a first coding vector;

inputting the historical preference information into a second vector coding module of the prediction model for coding to generate a second coding vector;

inputting the current resource type and each resource type into a third vector coding module of a prediction model for coding, and generating a corresponding third coding vector;

and using the first encoding vector, the second encoding vector and the third encoding vector as encoding vectors of the resource acquisition request, the historical preference information and the current resource category.

Optionally, the inputting the current resource category and the third vector coding module of each resource category into the prediction model for coding to generate a corresponding third coding vector includes:

inputting the current resource categories into a third vector coding module of the prediction model as an input set for coding, and generating first sub-coding vectors corresponding to the current resource categories;

the resource categories are jointly used as an input set to be input into a third vector coding module of the prediction model for coding, and a second sub-coding vector corresponding to the input set is generated;

inputting the first sub-coding vector and the second sub-coding vector into an attention calculation module for attention calculation to generate a corresponding third sub-coding vector;

inputting the second sub-coding vector into a full convolution network and an average pooling layer of a prediction model for processing to generate a corresponding fourth sub-coding vector;

and taking the third sub-coding vector and the fourth sub-coding vector as the third coding vector corresponding to the current resource category and each resource category.

Optionally, the inputting the coding vector into the probability prediction module of the prediction model for probability prediction to generate a probability prediction result that the user submits negative feedback to each resource category includes:

inputting the first encoding vector, the second encoding vector and the third encoding vector into an attention calculation submodule in a probability prediction module for attention calculation to generate a corresponding attention calculation result;

performing multi-task learning based on the attention calculation result to generate an initial probability prediction result of negative feedback submitted by the user to each task dimension under each resource category;

and determining the probability prediction result of negative feedback submitted by the user to each resource type according to the initial probability prediction result.

Optionally, the performing multi-task learning based on the attention calculation result to generate an initial probability prediction result of negative feedback submitted by the user to each task dimension under each resource category includes:

inputting the attention calculation result into a full convolution network in the probability prediction module for dimensionality reduction processing to generate a dimensionality reduction processing result corresponding to each task dimensionality under each resource category;

and processing the dimension reduction processing result by using an activation function to generate an initial probability prediction result of negative feedback submitted by the user to each task dimension under each resource category.

Optionally, the resource processing method further includes:

inquiring a common sense information coding vector which has a mapping relation with the resource to be acquired in a pre-established common sense information mapping table;

and determining the target resource category of the resource to be acquired based on the resource acquisition request, the historical preference information, the current resource category and the common sense information encoding vector.

Optionally, the resource processing method further includes:

and extracting the resources to be acquired corresponding to the target resource category, and sending the resources to be acquired corresponding to the target resource category to the user to respond to the resource acquisition request.

Optionally, the resource processing method further includes:

receiving feedback information submitted by the user aiming at the resource to be acquired;

and under the condition that the user submits negative feedback for the resources to be acquired corresponding to the target resource category according to the feedback information, adjusting the target resource category according to the historical preference information and the current resource category, and returning the resources to be acquired corresponding to the adjustment result to the user.

According to a second aspect of embodiments of the present specification, there is provided another resource processing method including:

receiving a voice instruction carrying an information resource acquisition request, and determining historical preference information of a user for each resource category under the condition that at least two resource categories corresponding to information resources to be acquired in the information resource acquisition request are determined;

acquiring the current resource category contained in the target conversation turn in the current conversation of the user;

determining a target resource category of the information resource to be acquired based on the voice instruction, the historical preference information and the current resource category;

and extracting the information resource to be acquired corresponding to the target resource category and sending the information resource to the user so as to respond to the voice instruction.

According to a third aspect of embodiments herein, there is provided a resource processing apparatus including:

the system comprises a receiving module, a processing module and a processing module, wherein the receiving module is configured to receive a resource obtaining request carrying resources to be obtained, and determine historical preference information of a user on each resource category under the condition that the resources to be obtained correspond to at least two resource categories;

the acquisition module is configured to acquire a current resource category contained in a target conversation turn in a current conversation of a user;

a determining module configured to determine a target resource category of the resource to be acquired based on the resource acquisition request, the historical preference information, and the current resource category.

According to a fourth aspect of embodiments herein, there is provided a computing device comprising:

a memory and a processor;

the memory is to store computer-executable instructions, and the processor is to execute the computer-executable instructions to:

According to a fifth aspect of embodiments herein, there is provided a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the resource processing method.

In one embodiment of the present description, a resource acquisition request carrying a resource to be acquired is received, historical preference information of a user for each resource category is determined under the condition that it is determined that the resource to be acquired corresponds to at least two resource categories, a current resource category included in a target conversation turn in a current conversation of the user is acquired, and a target resource category of the resource to be acquired is determined based on the resource acquisition request, the historical preference information, and the current resource category.

After receiving a resource acquisition request of a user, the embodiments of the present specification may provide sorting results of different resource categories for the user in an individualized manner according to history preference information of the user for each resource category and resource categories included in a target conversation turn of the user in a current conversation, and combine resources to be acquired in the resource acquisition request to select resources to be acquired of the target resource category that better meets the user requirements according to the sorting results, thereby facilitating to improve accuracy of the resource acquisition results and facilitating to improve service experience of the user.

Drawings

FIG. 1 is a process flow diagram of a resource processing method provided in one embodiment of the present specification;

FIG. 2 is a diagram of a resource handling process provided in one embodiment of the present specification;

FIG. 3 is a flowchart illustrating a processing procedure of a resource processing method according to an embodiment of the present disclosure;

FIG. 4 is a diagram of a resource processing apparatus according to an embodiment of the present disclosure;

FIG. 5 is a process flow diagram of another resource handling method provided by one embodiment of the present description;

FIG. 6 is a schematic diagram of another resource processing apparatus provided in one embodiment of the present specification;

fig. 7 is a block diagram of a computing device according to an embodiment of the present disclosure.

Detailed Description

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present description. This description may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein, as those skilled in the art will be able to make and use the present disclosure without departing from the spirit and scope of the present disclosure.

The terminology used in the description of the one or more embodiments is for the purpose of describing the particular embodiments only and is not intended to be limiting of the description of the one or more embodiments. As used in one or more embodiments of the present specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used in one or more embodiments of the present specification refers to and encompasses any and all possible combinations of one or more of the associated listed items.

It will be understood that, although the terms first, second, etc. may be used herein in one or more embodiments to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, a first can also be referred to as a second and, similarly, a second can also be referred to as a first without departing from the scope of one or more embodiments of the present description. The word "if" as used herein may be interpreted as "at … …" or "when … …" or "in response to a determination", depending on the context.

First, the noun terms to which one or more embodiments of the present specification relate are explained.

Negative feedback: when the user communicates with the artificial intelligent dialogue equipment, the negative feedback action is generated due to unsatisfactory response to the equipment.

Negative feedback rate: in a specific user request set, the ratio of the request number of the negative feedback behaviors of the user to the total request number is included.

The achievement rate is as follows: the number of requests in a particular set of user requests for which the user request is successfully fulfilled is a proportion of the total number of requests.

In the present specification, a resource processing method is provided, and the present specification relates to a resource processing apparatus, a computing device, and a computer-readable storage medium, which are described in detail one by one in the following embodiments.

After receiving a request of a user, the artificial intelligence dialogue device can distribute traffic for the user to call related applications to provide a server for the user, but the current traffic distribution system is relatively static, the distribution sequencing result of the same user request is fixed and cannot change with different users or scenes, and the target adjustment strategy is not directly optimized by taking online user experience as an algorithm, so that the demand of a new generation of intelligent assistants cannot be met. The new generation of intelligent traffic distribution and sequencing model directly takes user experience as an online optimization target and has the characteristics of individuation, self-adaptation and scene. Under the condition that a plurality of fields or skills can be accepted in the same user request, how to develop a set of sequencing model which refers to different user preferences and different context dialog scenes and adaptively adjusts strategies according to the user experience along with the time lapse is the key for further improving the user experience and promoting the development of skill ecology and quality.

Fig. 1 shows a process flow diagram of a resource processing method provided in accordance with one embodiment of the present specification, which includes steps 102 to 106.

Step 102, receiving a resource obtaining request carrying resources to be obtained, and determining historical preference information of a user for each resource category under the condition that the resources to be obtained correspond to at least two resource categories.

The resource processing method provided in the embodiment of the present specification is applied to a client, where the client includes, but is not limited to, a large-scale Audio/video playing device, a game console, a desktop computer, a smart phone, a tablet computer, an MP3(Moving Picture Experts Group Audio Layer III, Moving Picture Experts compression standard Audio Layer 3) player, an MP4(Moving Picture Experts Group Audio Layer IV, Moving Picture Experts compression standard Audio Layer 4) player, a laptop, an e-book reader, and other clients such as a display terminal.

The user can send a resource acquisition request to the client, and after receiving the resource acquisition request of the user, the client can provide resources of corresponding categories for the user according to the resource categories of the resources to be acquired in the resource acquisition request. However, when the resource to be acquired has multiple resource categories, the client needs to determine the historical preference information of the user for each resource category, so as to select a resource of one category from the multiple resource categories according to the historical preference information, and return the resource to the user.

Specifically, the resources to be acquired, that is, the resources requested to be acquired by the user include, but are not limited to, video resources, audio resources, text resources, and the like; the resource category, i.e. the category of the resource to be acquired, includes but is not limited to a video category, an audio category, a text category, and the like.

The client side receives a resource obtaining request of the resource to be obtained and can determine the historical preference information of the user on each resource category under the condition that the resource to be obtained corresponds to at least two resource categories.

Taking the resource obtaining request as "song M1 of the singer S1", for example, since the resource category of the song M1 includes two categories, i.e., audio (song) and video (music short), in this case, the client needs to determine the historical preference information of the user for each resource category in the client to determine which category of resource is provided for the user.

In practical applications, the historical preference information of the user for each resource category includes, but is not limited to, the number of resource acquisition requests sent to the client by the user in the past 3 months/1 month/7 days/1 day, the number of times that each resource category was invoked in the past 3 months/1 months/7 days/1 day, the negative feedback information of each resource category, the number of times that the user submitted resource acquisition requests in the past 3 months/1 month/7 days/1 day under the resource category, the number of times that the user submitted resource acquisition de-requests in the past 3 months/1 month/7 days/1 day, the number of times that the user submitted resource acquisition de-requests in the past 3 months/7 days, the number of times that the user submitted resource acquisition de-requests under the resource categories, the number of times that the user submitted resource acquisition de-requests, the number of times that the user submitted the resource categories, Negative feedback information submitted by the user, etc.

And 104, acquiring the current resource category contained in the target conversation turn in the current conversation of the user.

Specifically, if a user performs a dialogue interaction with a client, that is, the user sends a resource acquisition request to the client by performing a dialogue with the client, the current dialogue is a dialogue carrying the resource acquisition request, and the current dialogue may include other dialogues except the dialogue carrying the resource acquisition request, and the current dialogue may include a plurality of dialogue turns; one conversation turn is a turn of conversation between the user and the client; and the target conversation turn can be 1-3 conversation turns before the conversation turn in which the conversation carrying the resource acquisition request is located.

In addition, the current resource category, that is, the resource category included in the target session turn, is characterized in that the current resource category is referred to in a session interaction process with the client before submitting the resource acquisition request.

The current resource category is equivalent to real-time information of a user corresponding to the historical preference information, and the embodiment of the specification determines the target resource category of the resource to be acquired by combining the historical preference information and the real-time information, so that the accuracy of the determination result is improved.

And 106, determining the target resource type of the resource to be acquired based on the resource acquisition request, the historical preference information and the current resource type.

Specifically, after determining historical preference information of the user for each resource category and acquiring a current resource category included in a target conversation turn in a current conversation of the user, the target resource category of the resource to be acquired can be determined based on the resource acquisition request, the historical preference information and the current resource category.

In specific implementation, the target resource category of the resource to be acquired is determined based on the resource acquisition request, the historical preference information, and the current resource category, and the method can be specifically implemented in the following manner:

Specifically, after determining historical preference information of a user for each resource category and acquiring a current resource category contained in a target conversation turn in a current conversation of the user, the resource acquisition request, the historical preference information and the current resource category are input into a prediction model, the probability that the user submits negative feedback for each resource category is predicted by the prediction model, a corresponding prediction result is generated, and then the resource category with lower probability that the user submits negative feedback is selected as the target resource category of the resource to be acquired according to the prediction result.

In specific implementation, the step of inputting the resource acquisition request, the historical preference information and the current resource category into a prediction model for probability prediction to generate a probability prediction result of negative feedback submitted by the user for each resource category includes:

Further, using the resource acquisition request, the historical preference information, and the current resource category as an input set, and inputting a vector encoding module of a prediction model to perform encoding processing to generate an encoding vector of the input set, including:

Further, inputting the current resource type and each resource type into a third vector coding module of the prediction model for coding, and generating a corresponding third coding vector, including:

Specifically, the prediction model comprises a vector coding module and a probability prediction module, wherein the vector coding module is used for coding the resource acquisition request, the historical preference information and the current resource category; and the probability prediction module is used for performing probability prediction on each resource category according to the coding result output by the vector coding module so as to obtain a probability prediction result of negative feedback submitted by a user on each resource category.

In addition, since there are differences in resource acquisition requests, historical preference information, and information contained in the current resource category, for example: the resource acquisition request and the current resource category mostly contain some semantic information, and the historical preference information mostly contains some numerical information, so that the prediction model adopts different vector coding modules to code different input information to obtain a more accurate coding result.

In an embodiment of the present specification, a first vector encoding module in a prediction model is used to encode a resource acquisition request to generate a first encoded vector, where the first vector encoding module may use an encoder of a transformer model; a third vector coding module of the prediction model is adopted to carry out coding processing on the current resource type and each resource type to generate a third coding vector, and the third vector coding module can also adopt a coder encoder of a transformer model; and a second vector coding module adopting the prediction model carries out coding processing on the historical preference information to generate a second coding vector, wherein the second vector coding module can be realized by adopting a stat coding layer.

After the first encoding vector, the second encoding vector and the third encoding vector are obtained by encoding, the first encoding vector, the second encoding vector and the third encoding vector can be used as encoding vectors of the resource obtaining request, the historical preference information and the current resource type.

Fig. 2 is a schematic diagram of a resource processing procedure provided in an embodiment of the present specification. After the resource acquisition request, the historical preference information and the current resource category are acquired, the resource acquisition request, the historical preference information and the current resource category are input into a prediction model for processing, and the prediction model comprises a vector coding module and a probability prediction module, so that after the resource acquisition request, the historical preference information and the current resource category are input into the prediction model, the resource acquisition request, the historical preference information and the current resource category are firstly coded by the vector coding module to generate a corresponding coding vector, and then the probability prediction module predicts the probability that a user submits negative feedback to each resource category based on a coding result so as to screen a target resource category of the resource to be acquired according to the prediction result.

As shown in fig. 2, the input of the prediction model includes a request text (Query), a resource category (Item), a current resource category (User real-time information), and history preference information corresponding to the resource acquisition request, including User profile (User data), Item profile (resource category data), User Item profile (User resource category data), and User preference (User preference).

Wherein, item is all resource categories contained in the client; a User real-time viewer is User real-time information, specifically, a resource category included in 1 to 3 pairs of dialogs before a dialog turn of a resource acquisition request is included in a current dialog; user profile, namely the number of times of resource acquisition requests of the User for 3 months/1 month/7 days/1 day in the past, submitted negative feedback information and the like; item profile, namely the number of times that the Item has been called for 3 months/1 month/7 days/1 day in the past, negative feedback information submitted by a user for the Item, and the like; UI profile, namely the frequency of resource acquisition requests submitted by the user in the last 3 months/1 month/7 days/1 day, submitted negative feedback information and the like under the item; the User reference refers to the first 10 resource categories preferred by the User in the past 3 months/7 days, and the number of times of resource acquisition requests submitted by the User under the resource categories, submitted negative feedback information and the like.

After the information is input into a prediction model, a text Encoding Layer in a vector Encoding module encodes Query to generate a first Encoding vector; encoding User profile, Item profile, UI profile and User prediction by a data Encoding Layer Stat Encoding Layer to generate a second Encoding vector; coding each resource category contained in a User real-time viewer by a resource category coding Layer Item Encoding Layer to generate a first sub-coding vector corresponding to each current resource category; coding the Item by a resource category coding Layer Item Encoding Layer to generate a corresponding second sub-coding vector; then, an Attention calculation module (text Attention layer Context) carries out Attention calculation on the first sub-coding vector and the second sub-coding vector to generate a corresponding third sub-coding vector; inputting the second sub-coding vector into a full convolution network (FCN Layer) for processing, and inputting a processing result into an average Pooling Layer (Avg Pooling) for Pooling to generate a fourth sub-coding vector; the third sub-coding vector and the fourth sub-coding vector can be used together as a current resource category and a third coding vector corresponding to each resource category.

Or after the average pooling layer performs pooling processing on the result output by the full-convolution network, the generated fourth sub-coding vector may be input to the drop out layer for processing, so as to generate a fifth sub-coding vector, where the third sub-coding vector and the fifth sub-coding vector may be used as the current resource category and the third coding vector corresponding to each resource category.

In addition, the encoding vector is input into a probability prediction module of the prediction model for probability prediction, and a probability prediction result of negative feedback submitted by the user to each resource category is generated, which can be specifically realized by the following method:

Further, performing multi-task learning based on the attention calculation result to generate an initial probability prediction result of negative feedback submitted by the user to each task dimension under each resource category, including:

Specifically, in this embodiment of the present description, a schematic diagram of a probability prediction module in a prediction model is shown in fig. 2, after a vector coding module performs coding processing on a resource acquisition request, historical preference information, and a current resource category to generate a first coding vector, a second coding vector, and a third coding vector, the first coding vector, the second coding vector, and the third coding vector may be input to the probability prediction module, an Attention computation submodule (Target Attention authorization) in the probability prediction module performs Attention computation on the first coding vector, the second coding vector, and the third coding vector to generate corresponding Attention computation results, and then performs multi-task learning based on the Attention computation results to generate probability prediction results that a user submits negative feedback to each resource category.

Because the embodiment of the specification needs to predict the probability of negative feedback submitted by the user on different resource categories, in practical application, whether the user submits negative feedback on the resource categories or not is mainly influenced by 8 different task dimension factors, such as changing intention, errors in feedback given to the user by a client and the like; therefore, in order to ensure the accuracy of the probability prediction result, after the attention calculation result is generated, multi-task learning can be performed based on the attention calculation result, each task dimension is taken as a learning task, negative feedback probabilities of different task dimensions are predicted respectively, and the total probability that the user submits negative feedback to each resource category is determined according to the prediction result.

The specific multi-task learning process is that the sub-coding vectors corresponding to different task dimensions are extracted from the attention calculation result, and then the different sub-coding vectors are respectively input into a full convolution network (FCN Layer) in a probability prediction module for dimension reduction processing to generate a dimension reduction processing result corresponding to each task dimension under each resource category; then, processing the dimension reduction processing result by using an activation function 1 and an activation function 2 to generate an initial probability prediction result of negative feedback submitted by the user to each task dimension under each resource category; wherein, the activation function 1 may be a leakage modified linear unit function leak Relu, and the activation function 2 may be a Sigmoid function; and finally, calculating the probability prediction result of the total negative feedback submitted by the user to each resource category according to the weight occupied by different task dimensions and the probability prediction results corresponding to the different task dimensions.

By means of multi-task learning, accuracy of probability prediction results is improved.

In addition, after receiving a resource acquisition request of a user, querying a common sense information encoding vector having a mapping relation with the resource to be acquired in a pre-established common sense information mapping table, and determining a target resource category of the resource to be acquired based on the resource acquisition request, the historical preference information, the current resource category and the common sense information encoding vector.

Specifically, as shown in fig. 2, in addition to inputting the resource acquisition request, the historical preference information, and the current resource category into the prediction model, a common sense information encoding vector may also be input into the prediction model, but the difference is that an information encoding module in the prediction model needs to encode the resource acquisition request, the historical preference information, and the current resource category, and does not need to encode the common sense information encoding vector; after the vector encoding module performs encoding processing on the resource acquisition request, the historical preference information and the current resource category to generate a first encoding vector, a second encoding vector and a third encoding vector, the first encoding vector, the second encoding vector, the third encoding vector and the common sense information encoding vector can be input to the probability prediction module together for processing.

In practical application, the common sense information encoding vector can be obtained by querying a pre-established common sense information mapping table, and the common sense information mapping table can be updated by a user. Since the common sense information in the embodiments of the present specification relates to 41 categories, such as "Person", "Artist", "Organization", and the like, and typically, three granularities are involved in one resource acquisition request, which are values of a domain, an intention, and an information slot, after acquiring the resource acquisition request, the client may determine the value of the information slot in the resource acquisition request, and then query a common sense information encoding vector corresponding to the value of the information slot in a pre-established common sense information mapping table, where the common sense information encoding vector is 41 dimensions, and the value of each dimension is a score of one common sense information, such as 0.2, 0.6, or 1.0, and this score is used to represent a confidence of each common sense information.

Taking the resource acquisition request as "song M1 playing singer S1" as an example, the value of the information slot may be extracted here as singer S1 or song M1, then the common sense information encoding vector corresponding to singer S1 or song M1 may be queried in a pre-established common sense information mapping table, and the target resource category of the resource to be acquired is determined based on the resource acquisition request, the history preference information, the current resource category and the common sense information encoding vector.

And finally, after the target resource category of the resource to be acquired is determined, the resource to be acquired corresponding to the target resource category can be extracted, and the resource to be acquired corresponding to the target resource category is sent to the user to respond to the resource acquisition request. For example, if the target resource category is audio, a song is played for the user; if the target resource category is music shorts, then a music short (MV) of song M1 is played for the user.

In this embodiment of the present specification, after sending a resource to be acquired corresponding to a target resource category to a user, the user may submit feedback for the resource to be acquired, and may also adjust the resource category of the resource to be acquired for the user when the feedback submitted by the user is negative feedback, which may specifically be implemented in the following manner:

Specifically, under the condition that feedback information of the user for the resource to be acquired corresponding to the target resource category is received, whether the user submits negative feedback or not can be determined according to the feedback information, if the user submits negative feedback, the resource to be acquired of the target resource category does not meet the requirement of the user is represented, therefore, the resource category meeting the requirement of the user can be determined again according to the historical preference information of the user for the resource category and the current resource category and in combination with the feedback information, namely, the target resource category is adjusted, and the resource to be acquired corresponding to the adjustment result is returned to the user.

For example, when the target resource category is audio, a song is played for the user through the player a, but in the playing process of the song, when negative feedback of 'change one player' submitted by the user is received, the player B for playing the song for the user can be re-determined according to the feedback information and the historical preference information of the user and the current resource category, and the song is continuously played for the user through the player B.

The embodiment of the specification provides a set of intelligent traffic distribution and sequencing system which directly takes negative-going behavior feedback of a user as an online optimization target and has individuation, self-adaptation and scene. The system can reduce negative feedback of online users, improve the achievement rate of user requirements, give the sequencing results individually according to users with different preferences, give the sequencing results in a scene mode under different context conversation histories, and adjust the sequencing strategy in a self-adaptive mode along with the time according to user experience. In addition, the input of the scheme has pointwise (pointwise) aiming at the item list under each resource acquisition request of each user, so that large batches of training data which are easy to collect are trained; according to the scheme, the Query and the Item are respectively coded, and the semantic information of the Query can be directly referred to in the downstream sorting process, so that error conduction caused by the analysis error of an upstream natural language understanding module is avoided; according to the scheme, randomly generated negative sampling samples are added to the item coding module, so that semantic mismatching of item dimensionality is avoided to a certain extent.

The following describes the resource processing method further by taking an application of the resource processing method provided in this specification in an actual scene as an example, with reference to fig. 3. Fig. 3 shows a flowchart of a processing procedure of a resource processing method according to an embodiment of the present specification, and specific steps include step 302 to step 332.

Step 302, receiving a resource obtaining request carrying resources to be obtained, and determining historical preference information of a user for each resource category under the condition that the resources to be obtained correspond to at least two resource categories.

Step 304, obtaining the current resource category contained in the target dialog turn in the current dialog of the user.

Step 306, inputting the resource obtaining request into a first vector coding module of the prediction model for coding, and generating a first coding vector.

And 308, inputting the historical preference information into a second vector coding module of the prediction model for coding, and generating a second coding vector.

And 310, inputting the current resource categories into a third vector coding module of the prediction model as an input set for coding, and generating first sub-coding vectors corresponding to the current resource categories.

And step 312, using the resource categories as input sets together to enter a third vector coding module of the prediction model for coding, and generating a second sub-coding vector corresponding to the input set.

Step 314, inputting the first sub-coded vector and the second sub-coded vector into an attention calculation module for attention calculation, and generating a corresponding third sub-coded vector.

And step 316, inputting the second sub-coding vector into a full convolution network and an average pooling layer of a prediction model for processing, and generating a corresponding fourth sub-coding vector.

Step 318, using the third sub-coding vector and the fourth sub-coding vector as the third coding vector corresponding to the current resource type and each resource type.

Step 320, using the first encoding vector, the second encoding vector and the third encoding vector as the encoding vectors of the resource obtaining request, the historical preference information and the current resource category.

Step 322, inquiring a common sense information encoding vector having a mapping relation with the resource to be acquired in a pre-established common sense information mapping table.

Step 324, inputting the first encoded vector, the second encoded vector, the third encoded vector and the common sense information encoded vector into an attention calculation submodule in a probability prediction module for attention calculation, and generating a corresponding attention calculation result.

And 326, inputting the attention calculation result into a full convolution network in the probability prediction module for dimensionality reduction processing, and generating a dimensionality reduction processing result corresponding to each task dimensionality under each resource category.

And 328, processing the dimension reduction processing result by using an activation function, and generating an initial probability prediction result of negative feedback submitted by the user to each task dimension under each resource category.

Step 330, determining the probability prediction result of negative feedback submitted by the user to each resource type according to the initial probability prediction result.

And 332, determining the target resource type of the resource to be acquired according to the probability prediction result.

The embodiment of the specification customizes the coding structure suitable for different characteristics for the input characteristics. Specifically, for Query features and Item features, a Transformer-based text encoder is used for encoding respectively; for a User real-time viewer, firstly splitting the User real-time viewer into independent items, then adding position codes, and passing through dropout after a full connection layer; furthermore, in order to capture the semantic association between these historical item and the current item, we add a text-based attention mechanism between the two in particular.

After all feature codes are finished, the embodiment of the present specification specifically introduces a layer of attention mechanism module to perform weighted combination of features. Firstly, the features are spliced together to be used as target features, and the original feature list and the target features are compared based on an attention mechanism to obtain the weight of each feature to the target features. And finally weighting each feature based on the weight to obtain a finally processed feature vector.

In addition, the embodiment of the present specification adopts multi-task learning for overall user negative feedback and negative feedback of each sub-dimension (such as playing interruption, repeated text transmission, etc.), and takes prediction of overall negative feedback as a main task and predictions of other dimensions as auxiliary tasks. Considering that the importance of each subtask is different in different scenes (for example, in a song-on-demand scene, the importance of playing interruption is significantly higher than that in other scenes), such a multi-task learning mechanism facilitates us to flexibly adjust the strategy according to the scenes.

Corresponding to the above method embodiment, this specification further provides a resource processing apparatus embodiment, and fig. 4 shows a schematic diagram of a resource processing apparatus provided in an embodiment of this specification. As shown in fig. 4, the apparatus includes:

a receiving module 402, configured to receive a resource acquisition request carrying resources to be acquired, and determine historical preference information of a user for each resource category when determining that the resources to be acquired correspond to at least two resource categories;

an obtaining module 404 configured to obtain a current resource category included in a target dialog turn in a current dialog of a user;

a determining module 406 configured to determine a target resource category of the resource to be acquired based on the resource acquisition request, the historical preference information, and the current resource category.

Optionally, the determining module 406 includes:

the forecasting submodule is configured to input the resource obtaining request, the historical preference information and the current resource category into a forecasting model for probability forecasting and generate a probability forecasting result of negative feedback submitted by the user to each resource category;

a determining submodule configured to determine a target resource category of the resource to be acquired according to the probability prediction result.

Optionally, the prediction sub-module comprises:

the encoding unit is configured to take the resource acquisition request, the historical preference information and the current resource category as an input set, input a vector encoding module of a prediction model for encoding, and generate an encoding vector of the input set;

and the prediction unit is configured to input the coding vector into a probability prediction module of the prediction model for probability prediction, and generate a probability prediction result of negative feedback submitted by the user on each resource category.

Optionally, the encoding unit includes:

the first coding subunit is configured to input the resource acquisition request into a first vector coding module of a prediction model for coding processing to generate a first coding vector;

the second coding subunit is configured to input the historical preference information into a second vector coding module of the prediction model for coding processing, and generate a second coding vector;

a third encoding subunit, configured to input the current resource category and each resource category into a third vector encoding module of the prediction model for encoding, and generate a corresponding third encoding vector;

a processing subunit configured to use the first encoding vector, the second encoding vector, and the third encoding vector together as the encoding vector of the resource acquisition request, the historical preference information, and the current resource category.

Optionally, the third encoding subunit is further configured to:

Optionally, the prediction unit includes:

the calculation subunit is configured to input the first encoded vector, the second encoded vector and the third encoded vector to an attention calculation submodule in a probability prediction module for attention calculation, and generate a corresponding attention calculation result;

the generating subunit is configured to perform multi-task learning based on the attention calculation result, and generate an initial probability prediction result of negative feedback submitted by the user to each task dimension under each resource category;

a determining subunit configured to determine a probability prediction result of the user submitting negative feedback for the each resource category according to the initial probability prediction result.

Optionally, the generating subunit is further configured to:

Optionally, the resource processing apparatus further includes a query module configured to:

Optionally, the resource processing apparatus further includes:

and the extracting module is configured to extract the resources to be acquired corresponding to the target resource category, and send the resources to be acquired corresponding to the target resource category to the user so as to respond to the resource acquisition request.

Optionally, the resource processing apparatus further includes an adjusting module configured to:

The above is a schematic scheme of a resource processing apparatus of this embodiment. It should be noted that the technical solution of the resource processing apparatus and the technical solution of the resource processing method belong to the same concept, and details that are not described in detail in the technical solution of the resource processing apparatus can be referred to the description of the technical solution of the resource processing method.

Fig. 5 shows a process flow diagram of another resource processing method provided in accordance with one embodiment of the present specification, including steps 502 through 506.

Step 502, receiving a voice command carrying an information resource acquisition request, and determining historical preference information of a user for each resource category when determining that the information resource to be acquired in the information resource acquisition request corresponds to at least two resource categories.

Step 504, obtaining the current resource category contained in the target dialog turn in the current dialog of the user.

Step 506, determining a target resource category of the information resource to be acquired based on the voice instruction, the historical preference information and the current resource category.

And step 508, extracting the information resource to be acquired corresponding to the target resource type and sending the information resource to the user so as to respond to the voice command.

Specifically, the resource processing method provided in the embodiments of the present specification is applied to a client, where the client may be an intelligent robot for human-computer interaction, a user may perform voice interaction with the client by sending a voice instruction, that is, send a resource acquisition request to the client by sending the voice instruction, and after receiving the voice instruction carrying the information resource acquisition request from the user, the client may provide resources of corresponding categories for the user according to the resource category of the information resource to be acquired in the information resource acquisition request. However, under the condition that the information resource to be acquired has two or more resource categories, the client needs to determine historical preference information of the user for each resource category, acquire the current resource category included in the target conversation turn in the current conversation of the user, select an information resource to be acquired of one target resource category from the two or more resource categories according to the historical preference information, the current resource category and the information resource acquisition request, and send the information resource to the user to respond to the voice instruction.

In practical applications, the information resource to be acquired may be a multimedia resource, such as an audio or video resource. The user can instruct the intelligent robot to play audio or video resources by sending a voice instruction, and after receiving the voice instruction, if two or more players capable of playing audio or video exist, the intelligent robot can select a target player meeting the user requirement from the two or more players according to the historical preference information of the user for each player and the player mentioned in the process of carrying out dialogue interaction between the user and the client before submitting the voice instruction, and play the audio or video by using the target player to respond to the voice instruction.

Alternatively, the information resource to be acquired may be an information resource, such as news information. The user can instruct the intelligent robot to show news information by sending a voice instruction, and after the intelligent robot receives the voice instruction, if two or more applications capable of showing the news information exist, the intelligent robot can select a target application meeting the requirements of the user from the two or more applications according to historical preference information of the user for each application and the application mentioned in the process of carrying out conversation interaction between the user and the client before submitting the voice instruction, and the target application is used for showing the news information to respond to the voice instruction.

Still alternatively, the information resource to be acquired may also be a commodity to be traded, such as a lipstick, a household appliance, and the like. The method comprises the steps that a user can instruct an intelligent robot to recommend commodities to the user in a mode of sending a voice instruction, after the intelligent robot receives the voice instruction, if two or more transaction platforms capable of being used for commodity transaction exist, a target transaction platform meeting the requirements of the user is selected from the two or more transaction platforms according to historical preference information of the user on each transaction platform and the transaction platforms mentioned in a process of carrying out conversation interaction between the user and a client before submitting the voice instruction, and commodity recommendation is carried out by using the target transaction platform so as to respond to the voice instruction; in addition, after the intelligent robot receives the voice instruction, if only one trading platform capable of being used for trading commodities exists, but commodities to be traded of multiple types or brands may exist in the platform, a target commodity type or commodity brand meeting the requirements of the user can be selected from the multiple commodity types or brands according to historical preference information of the user for each type or each brand and the commodity type or brand mentioned in the process of carrying out dialogue interaction between the user and the client before submitting the voice instruction, and commodity recommendation is carried out to respond to the voice instruction.

In addition, the user can interact with the client by sending the voice instruction, and can also send a corresponding instruction by inputting a text, after the client acquires the text input by the user, the information resource to be acquired of the user can be determined by text recognition and semantic recognition, so that the information resource to be acquired of the target resource type is sent to the user to respond to the instruction, the specific implementation process is similar to the processing process of the voice instruction of the user, and is not repeated here.

After receiving a voice instruction carrying an information resource acquisition request, the embodiments of the present specification may provide a sorting result of different resource categories for a user in an individualized manner according to history preference information of the user for each resource category and resource categories included in a target conversation turn of the user in a current conversation, and in combination with information resources to be acquired in the information resource acquisition request, so as to select information resources to be acquired of a target resource category that better meets the user requirements according to the sorting result, thereby facilitating improvement of accuracy of the information resource acquisition result and service experience of the user.

The above is a schematic scheme of another resource processing method of this embodiment. It should be noted that the technical solution of the resource processing method belongs to the same concept as the technical solution of the above-mentioned resource processing method, and details that are not described in detail in the technical solution of the resource processing method can be referred to the description of the technical solution of the above-mentioned resource processing method.

Corresponding to the above method embodiments, this specification further provides resource processing device embodiments, and fig. 6 shows a schematic diagram of another resource processing device provided in one embodiment of this specification. As shown in fig. 6, the apparatus includes:

an instruction receiving module 602, configured to receive a voice instruction carrying an information resource acquisition request, and determine history preference information of a user for each resource category when it is determined that an information resource to be acquired in the information resource acquisition request corresponds to at least two resource categories;

a resource category obtaining module 604, configured to obtain a current resource category included in a target session turn in a current session of the user;

a resource category determining module 606 configured to determine a target resource category of the information resource to be acquired based on the voice instruction, the historical preference information, and the current resource category;

the response module 608 is configured to extract the information resource to be acquired corresponding to the target resource category and send the information resource to the user, so as to respond to the voice instruction.

The above is a schematic scheme of another resource processing apparatus of the present embodiment. It should be noted that the technical solution of the resource processing apparatus and the technical solution of the other resource processing method belong to the same concept, and details that are not described in detail in the technical solution of the resource processing apparatus can be referred to the description of the technical solution of the other resource processing method.

FIG. 7 illustrates a block diagram of a computing device 700 provided in accordance with one embodiment of the present description. The components of the computing device 700 include, but are not limited to, memory 710 and a processor 720. Processor 720 is coupled to memory 710 via bus 730, and database 750 is used to store data.

Computing device 700 also includes access device 740, access device 740 enabling computing device 700 to communicate via one or more networks 760. Examples of such networks include the Public Switched Telephone Network (PSTN), a Local Area Network (LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), or a combination of communication networks such as the internet. Access device 740 may include one or more of any type of network interface, e.g., a Network Interface Card (NIC), wired or wireless, such as an IEEE802.11 Wireless Local Area Network (WLAN) wireless interface, a worldwide interoperability for microwave access (Wi-MAX) interface, an ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a bluetooth interface, a Near Field Communication (NFC) interface, and so forth.

In one embodiment of the present description, the above-described components of computing device 700, as well as other components not shown in FIG. 7, may also be connected to each other, such as by a bus. It should be understood that the block diagram of the computing device architecture shown in FIG. 7 is for purposes of example only and is not limiting as to the scope of the present description. Those skilled in the art may add or replace other components as desired.

Computing device 700 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., tablet, personal digital assistant, laptop, notebook, netbook, etc.), mobile phone (e.g., smartphone), wearable computing device (e.g., smartwatch, smartglasses, etc.), or other type of mobile device, or a stationary computing device such as a desktop computer or PC. Computing device 700 may also be a mobile or stationary server.

Wherein the memory 710 is configured to store computer-executable instructions and the processor 720 is configured to execute the following computer-executable instructions:

The above is an illustrative scheme of a computing device of the present embodiment. It should be noted that the technical solution of the computing device and the technical solution of the resource processing method belong to the same concept, and details that are not described in detail in the technical solution of the computing device can be referred to the description of the technical solution of the resource processing method.

An embodiment of the present specification also provides a computer readable storage medium storing computer instructions, which when executed by a processor, are used for implementing the steps of the resource processing method.

The above is an illustrative scheme of a computer-readable storage medium of the present embodiment. It should be noted that the technical solution of the storage medium belongs to the same concept as the technical solution of the resource processing method, and details that are not described in detail in the technical solution of the storage medium can be referred to the description of the technical solution of the resource processing method.

The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.

The computer instructions comprise computer program code which may be in the form of source code, object code, an executable file or some intermediate form, or the like. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, usb disk, removable hard disk, magnetic disk, optical disk, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution medium, and the like. It should be noted that the computer readable medium may contain content that is subject to appropriate increase or decrease as required by legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer readable media does not include electrical carrier signals and telecommunications signals as is required by legislation and patent practice.

It should be noted that, for the sake of simplicity, the foregoing method embodiments are described as a series of acts, but those skilled in the art should understand that the present embodiment is not limited by the described acts, because some steps may be performed in other sequences or simultaneously according to the present embodiment. Further, those skilled in the art should also appreciate that the embodiments described in this specification are preferred embodiments and that acts and modules referred to are not necessarily required for an embodiment of the specification.

In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

The preferred embodiments of the present specification disclosed above are intended only to aid in the description of the specification. Alternative embodiments are not exhaustive and do not limit the invention to the precise embodiments described. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the embodiments and the practical application, to thereby enable others skilled in the art to best understand and utilize the embodiments. The specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method of resource processing, comprising:

2. The resource processing method according to claim 1, wherein the determining a target resource category of the resource to be acquired based on the resource acquisition request, the historical preference information, and the current resource category comprises:

3. The resource processing method according to claim 2, wherein the probability prediction of the resource acquisition request, the historical preference information, and the current resource category input prediction model to generate a probability prediction result of negative feedback submitted by the user for each resource category comprises:

4. The resource processing method according to claim 3, wherein the encoding processing is performed by a vector encoding module that inputs a prediction model and generates an encoding vector of the input set, using the resource acquisition request, the historical preference information, and the current resource category as input sets, and the method includes:

5. The resource processing method according to claim 4, wherein the encoding the current resource class and the third vector encoding module of each resource class input into the prediction model to generate a corresponding third encoding vector comprises:

6. The resource processing method according to claim 4 or 5, wherein the probability prediction module that inputs the coding vector into the prediction model for probability prediction to generate the probability prediction result that the user submits negative feedback for each resource category comprises:

7. The resource processing method according to claim 6, wherein the performing multi-task learning based on the attention calculation result to generate an initial probability prediction result of negative feedback submitted by the user to each task dimension under each resource category comprises:

8. The resource processing method of claim 1, further comprising:

9. The resource processing method of claim 1, further comprising:

10. The resource processing method of claim 9, further comprising:

11. A method of resource processing, comprising:

12. A resource processing apparatus comprising:

13. A computing device, comprising:

a memory and a processor;

the memory is configured to store computer-executable instructions, and the processor is configured to implement the steps of the resource handling method according to any one of claims 1 to 11 when executing the computer-executable instructions.

14. A computer readable storage medium storing computer instructions which, when executed by a processor, carry out the steps of the resource handling method of any one of claims 1 to 11.