CN112579755A

CN112579755A - Information response method and information interaction platform based on artificial intelligence and cloud computing

Info

Publication number: CN112579755A
Application number: CN202011530960.0A
Authority: CN
Inventors: 冯启鹏
Original assignee: Individual
Current assignee: Individual
Priority date: 2020-12-22
Filing date: 2020-12-22
Publication date: 2021-03-30

Abstract

The embodiment of the application provides an information response method and an information interaction platform based on artificial intelligence and cloud computing, wherein response input problems are sequentially searched from a response index sequence configured at a cloud end, when a certain response input problem is searched from the response index sequence, a response index object can be quickly responded, configuration association operation of target response service is directly executed through a problem updated at the cloud end, and configuration time of the response input problem is greatly reduced. When a certain response input problem is not found in the response index sequence, the response input problem which is not found is updated in the expansion updating sequence of the response index sequence so as to update the response index sequence and then execute the configuration association operation, the updated problem is stored in the response index sequence, the configuration association operation can be directly executed when the next response meets the same response input problem, the configuration association operation of the target response service can be efficiently realized, and the service configuration efficiency is improved.

Description

Information response method and information interaction platform based on artificial intelligence and cloud computing

Technical Field

The application relates to the technical field of artificial intelligence, in particular to an information response method and an information interaction platform based on artificial intelligence and cloud computing.

Background

With the rapid development of artificial intelligence technology, information interaction terminals (such as intelligent robots) begin to enter a rapid growth stage, and various response interaction requirements of users can be met by configuring response services required by various users of the information interaction terminals.

In the related art, local information response is usually performed only on a target response service added data set configured by a request of an information interaction terminal, so that when the data set is large, the computing processing capacity is limited, the configuration efficiency is slow, and uniform maintenance is not convenient.

Disclosure of Invention

In order to overcome at least the above disadvantages in the prior art, an object of the present application is to provide an information response method and an information interaction platform based on artificial intelligence and cloud computing, which obtain at least one response input question and a target response service corresponding to each response input question, sequentially search each response input question from a response index sequence configured at a cloud, and when a certain response input question is found from the response index sequence, can quickly respond to the response index object, directly execute configuration association operation of the target response service through a question updated at the cloud, thereby greatly reducing configuration time of the response input question. When a certain response input problem is not found in the response index sequence, the response input problem which is not found is updated in the expansion updating sequence of the response index sequence to update the response index sequence, the configuration association operation is triggered and executed through the updated response index sequence and the response input problem in the response index sequence, the updated problem is stored in the response index sequence, and the configuration association operation can be directly executed when the next response meets the same response input problem, so that the configuration association operation of the target response service can be efficiently realized on the premise of not additionally adding a data set aiming at the target response service, and the service configuration efficiency is improved.

In a first aspect, the application provides an information response method based on artificial intelligence and cloud computing, which is applied to an information interaction platform, wherein the information interaction platform is in communication connection with a plurality of information interaction terminals, and the method comprises the following steps:

acquiring at least one response input problem and target response services corresponding to the response input problems respectively, and sequentially searching the response input problems from a response index sequence configured at a cloud end, wherein the target response services are realized through cloud computing services requested by the information interaction platform;

when the response input question is found from the response index sequence, determining a response index object of the response input question in the response index sequence;

when the response input problem is not found in the response index sequence, updating the response input problem which is not found in the extension updating sequence of the response index sequence, and determining a response index object of the updated response input problem in the response index sequence;

after the response index objects of the response input questions in the response index sequence are determined, the response input questions are configured and associated with the corresponding target response service according to the response index sequence of the response input questions and the response index objects of the response input questions in the response index sequence.

In a possible implementation manner of the first aspect, the obtaining at least one response input question and a target response service corresponding to each response input question includes:

when detecting the input response input problem, determining that each response input problem respectively corresponds to a response service identifier;

and determining target response services respectively corresponding to the response input questions according to the response service identifiers.

In a possible implementation manner of the first aspect, the sequentially searching for each of the response input questions from a response index sequence configured by a cloud includes:

determining question semantic features and question scene features respectively corresponding to the answer input questions;

determining response indexes respectively corresponding to the response input questions according to the question semantic features and the question scene features;

and sequentially searching each response input problem in a response index sequence configured at the cloud end according to a response index corresponding to each response input problem.

In a possible implementation manner of the first aspect, when the answer input question is not found in the answer index sequence, updating the answer input question that is not found in an extended update sequence of the answer index sequence, and determining an answer index object of the updated answer input question in the answer index sequence, includes:

when the response input problem is not found in the response index sequence, updating the problem deep learning labeling information corresponding to the response input problem which is not found;

according to the problem deep learning labeling information, in an expansion updating sequence of a problem expansion area included in the response index sequence, distributing updating nodes for the response input problems which are not found;

updating the response input question which is not found at the distributed updating node in the response index sequence so as to update the response index sequence;

and determining a response index object of the updated response input question in the updated response index sequence according to the update node.

In a possible implementation manner of the first aspect, the method further includes:

determining question frequency information corresponding to each of the response input questions;

after determining the response index objects of the response input questions in the response index sequence, respectively, according to the response index sequence of the response input questions and the response index objects of the response input questions in the response index sequence, respectively, triggering to associate the configuration of the response input questions in the corresponding target response service, the method includes:

after the response index objects of the response input questions in the response index sequence are determined, according to the response index sequence comprising the response input questions and the response index objects of the response input questions in the response index sequence, the response input questions are configured and associated with corresponding target response services according to the question frequency information corresponding to the response input questions.

In a possible implementation manner of the first aspect, the step of updating the question deep learning labeling information corresponding to the answer input question that is not found includes:

inputting the unsearched response input questions into a question deep learning labeling network, obtaining and updating question deep learning labeling information corresponding to the response input questions, wherein the question deep learning labeling information comprises question classification characteristic information and question response characteristic information of the response input questions;

the problem deep learning labeling network is obtained by the following configuration:

acquiring calibration response problem data for training a problem deep learning labeling network, wherein the calibration response problem data at least comprises response problems and target problem labeling information of the response problems;

determining calibration data vector distribution and deviation correction data vector distribution of calibration response problem data, wherein the deviation correction data vector distribution is formed by correcting the semantic splitting response characteristic component of the calibration response problem, and the characteristic labels of the calibration data vector distribution and the deviation correction data vector distribution are the same;

performing feature aggregation on the calibration data vector distribution and the deviation correction data vector distribution to obtain an aggregate calibration feature, and determining prediction problem marking information of the aggregate calibration feature;

determining a difference function value based on target problem labeling information and the predicted problem labeling information, adjusting network weight information of the problem deep learning labeling network based on the difference function value and continuing iterative training until iteration stopping conditions are met, wherein the target problem labeling information is used for representing confidence coefficients of the response problems in the calibration response problem data under the labeling information of each different problem;

the problem deep learning labeling network obtained through training is used for classifying the labeling information of the response input problems which are not found;

wherein, the deviation correction data vector distribution is obtained by the following method:

performing problem semantic splitting on the calibration response problem data to obtain a plurality of semantic splitting problem data, and determining the influence weight of each semantic splitting problem data vector distribution;

sequencing all the influence weights to obtain a sequencing result, and correcting the vector distribution of the plurality of semantic splitting problem data according to the sequencing characteristics of the sequencing result to obtain corrected and calibrated response problem data consisting of the plurality of semantic splitting problem data;

inputting the deviation correction calibration response problem data into the problem deep learning labeling network for feature extraction to obtain deviation correction data vector distribution, wherein the deviation correction data vector distribution comprises a plurality of semantic splitting problem data vector distributions.

In a possible implementation manner of the first aspect, the step of performing feature aggregation on the calibration data vector distribution and the deviation correction data vector distribution to obtain an aggregated calibration feature includes:

adopting the problem deep learning labeling network to obtain positive calibration characteristics of positive calibration response problem data, negative calibration characteristics of negative calibration response problem data and positive calibration deviation correction data vector distribution, and carrying out first vector fragment aggregation on the positive calibration characteristics and the positive calibration deviation correction data vector distribution to obtain first aggregated calibration characteristics;

adopting the problem deep learning labeling network to obtain negative calibration characteristics and negative calibration deviation correction data vector distribution of negative calibration response problem data, and carrying out second vector fragment polymerization on the negative calibration characteristics and the negative calibration deviation correction data vector distribution to obtain second polymerization calibration characteristics;

determining respective third vector segments in the first aggregate calibration feature and respective fourth vector segments in the second aggregate calibration feature;

for a feature construction node of each vector segment, aggregating a third vector segment of a third comprehensive weight and a fourth vector segment of a fourth comprehensive weight to obtain a first intersection aggregate nominal sub-feature of the feature construction node of the vector segment, aggregating the third vector segment of the fourth comprehensive weight and the fourth vector segment of the third comprehensive weight to obtain a second intersection aggregate nominal sub-feature of the feature construction node of the vector segment, wherein the sum of the third comprehensive weight and the fourth comprehensive weight is 1;

determining a third aggregate calibration feature according to the first aggregate calibration sub-feature of the feature formation nodes of all the vector segments, and determining a fourth aggregate calibration feature according to the second aggregate calibration sub-feature of the feature formation nodes of all the vector segments;

and taking the third aggregate calibration feature and the fourth aggregate calibration feature as the aggregate calibration feature, wherein the vector distribution of the positive calibration deviation rectifying data is formed by rectifying the semantic splitting response feature component of the positive calibration problem, the vector distribution of the negative calibration deviation rectifying data is formed by rectifying the semantic splitting response feature component of the negative calibration problem, the data of the positive calibration response problem is used for representing correct calibration response problem data, and the data of the negative calibration response problem is used for representing wrong calibration response problem data.

For example, in a possible implementation manner of the first aspect, after, when the answer input question is not found in the answer index sequence, updating the answer input question that is not found in an extended update sequence of the answer index sequence and determining that the updated answer input question is an answer index object in the answer index sequence, the method further includes:

determining question semantic features and question scene features corresponding to the answer input questions updated in the answer index sequence;

constructing a response index according to the determined question semantic features and the determined question scene features;

and performing associated storage on the constructed response index and a response index object of the corresponding response input question in the response index sequence.

For example, in a possible implementation manner of the first aspect, after determining the response index objects of the response input questions in the response index sequence, the triggering step of associating the response input question configurations with the corresponding target response services according to the response index sequence of the response input questions and the response index objects of the response input questions in the response index sequence includes:

after the response index objects of the response input questions in the response index sequence are determined, generating a question knowledge graph according to the response index sequence comprising the response input questions;

and triggering and associating each response input question configuration in the corresponding target response service based on the question knowledge graph and a response index object of each response input question in the response index sequence by calling an information response service.

running and configuring each target response service associated with each response input problem, and acquiring past emotion vector distribution and current emotion vector distribution obtained after emotion analysis is carried out on response service statistical data of each target response service on a user of the information interaction terminal, wherein the target response service is realized through cloud computing service requested by the information interaction platform;

determining emotion distinguishing feature information represented by corresponding emotion features in the past emotion vector distribution and the current emotion vector distribution, and determining target emotion feature representation which corresponds to the past emotion vector distribution and the current emotion vector distribution and meets the requirement of response behavior updating and tracking based on the emotion distinguishing feature information represented by the corresponding emotion features;

associating the target emotion feature representation in the current emotion vector distribution based on the target emotion feature representation in the past emotion vector distribution, and performing emotion transfer recognition on the associated emotion vector distribution in the current emotion vector distribution after emotion feature representation association to obtain emotion transfer information;

and determining first response process evaluation information corresponding to the past emotion vector distribution and second response process evaluation information corresponding to the current emotion vector distribution according to the emotion transfer information, updating response content of a target response service through the first response process evaluation information and the second response process evaluation information, and storing the updated response content into a corresponding block chain service.

In a possible implementation manner of the first aspect, the step of associating the target emotion feature representation in the current emotion vector distribution based on the target emotion feature representation in the past emotion vector distribution includes:

expressing each emotion feature representation included in the emotion transfer information in the past emotion vector distribution through a multi-dimensional emotion class bitmap, forming a past emotion feature expression set by expressing each emotion feature represented by the multi-dimensional emotion class bitmap, and performing emotion class vector extraction and emotion class vector association on the past emotion feature expression set to obtain a past emotion class vector map;

expressing each emotion characteristic representation included in the current response behavior updating information in the current emotion vector distribution through a multi-dimensional emotion category bitmap, expressing each emotion characteristic represented by the multi-dimensional emotion category bitmap to form a current emotion characteristic expression set, and extracting emotion category vectors and associating the emotion category vectors of the current emotion characteristic expression set to obtain a current emotion category vector map;

performing emotion transfer migration map extraction on the past response behavior updating information in the past emotion vector distribution based on the past emotion category vector map to obtain a past emotion transfer migration map;

judging whether the information comparison result of the response interaction information of the current emotion category vector map corresponding to each current response behavior updating information in the current emotion vector distribution and preset first response interaction reference information meets the index requirement corresponding to the current response interaction, and extracting emotion transfer migration maps of each current response behavior updating information in the current emotion vector distribution when the information comparison result meets the index requirement corresponding to the current response interaction to obtain a current emotion transfer migration map, wherein the preset first response interaction reference information is a past emotion category vector map corresponding to the past response behavior updating information in the past emotion vector distribution and target response service guide information of a user service survey result counted in advance;

correlating the target emotion feature representation in the current emotion vector distribution through migration map comparison information between the current emotion transfer migration map and the past emotion transfer migration map;

performing emotion transfer migration map extraction on the past response behavior update information in the past emotion vector distribution based on the past emotion category vector map to obtain a past emotion transfer migration map, wherein the emotion transfer migration map extraction includes:

judging whether response interaction information of a past emotion category vector map corresponding to each past response behavior updating information in the past emotion vector distribution meets index requirements corresponding to past response interaction;

adding emotion transfer labels to response interaction information of past emotion category vector maps of past response behavior update information, wherein the past emotion category vector maps meet index requirements corresponding to past response interaction, determining emotion feedback degree information corresponding to response interaction information of past emotion category vector maps of other past response behavior update information, and generating the past emotion transfer migration map according to the response interaction information added with the emotion transfer labels and the emotion feedback degree information corresponding to response interaction information of past emotion category vector maps of other past response behavior update information.

In a second aspect, an embodiment of the present application further provides an information response device based on artificial intelligence and cloud computing, which is applied to an information interaction platform, the information interaction platform is in communication connection with a plurality of information interaction terminals, and the device includes:

the system comprises an acquisition module, a processing module and an information interaction platform, wherein the acquisition module is used for acquiring at least one response input problem and target response services corresponding to the response input problems respectively, and searching the response input problems in sequence from a response index sequence configured at a cloud end, wherein the target response services are realized through cloud computing services requested by the information interaction platform;

a first determining module, configured to determine, when the response input question is found in the response index sequence, a response index object of the response input question in the response index sequence;

a second determining module, configured to update, in an extended update sequence of the response index sequence, the response input question that is not found when the response input question is not found in the response index sequence, and determine a response index object of the updated response input question in the response index sequence;

and the configuration module is used for configuring and associating each response input question in the corresponding target response service according to the response index sequence of each response input question and the response index object of each response input question in the response index sequence after determining the response index object of each response input question in the response index sequence.

In a third aspect, an embodiment of the present application further provides an information response system based on artificial intelligence and cloud computing, where the information response system based on artificial intelligence and cloud computing includes an information interaction platform and a plurality of information interaction terminals in communication connection with the information interaction platform;

the information interaction platform is used for:

In a fourth aspect, an embodiment of the present application further provides an information interaction platform, where the information interaction platform includes a processor, a machine-readable storage medium, and a network interface, 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 to at least one information interaction 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 an artificial intelligence and cloud computing-based information response method in the first aspect or any one of possible implementation manners in the first aspect.

In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium, where instructions are stored in the computer-readable storage medium, and when the instructions are executed, the computer is caused to execute the artificial intelligence and cloud computing based information response method in the first aspect or any one of the possible implementation manners of the first aspect.

Based on any one of the above aspects, the method and the device for processing the response input question sequentially search the response input questions from the response index sequence configured by the cloud by obtaining at least one response input question and the target response service corresponding to each response input question, when a certain response input question is searched from the response index sequence, the response index object can be quickly responded, the configuration association operation of the target response service is directly executed through the question updated by the cloud, and the configuration time of the response input question is greatly reduced. When a certain response input problem is not found in the response index sequence, the response input problem which is not found is updated in the expansion updating sequence of the response index sequence to update the response index sequence, the configuration association operation is triggered and executed through the updated response index sequence and the response input problem in the response index sequence, the updated problem is stored in the response index sequence, and the configuration association operation can be directly executed when the next response meets the same response input problem, so that the configuration association operation of the target response service can be efficiently realized on the premise of not additionally adding a data set aiming at the target response service, and the service configuration efficiency is improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that need to be called in the embodiments are briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.

Fig. 1 is a schematic view of an application scenario of an information response system based on artificial intelligence and cloud computing according to an embodiment of the present application;

fig. 2 is a schematic flowchart of an information response method based on artificial intelligence and cloud computing according to an embodiment of the present application;

fig. 3 is a schematic functional module diagram of an information response device based on artificial intelligence and cloud computing according to an embodiment of the present application;

fig. 4 is a schematic block diagram of structural components of an information interaction platform for implementing the above-described artificial intelligence and cloud computing-based information response method according to the embodiment of the present application.

Detailed Description

The present application will now be described in detail with reference to the drawings, and the specific operations in the method embodiments may also be applied to the apparatus embodiments or the system embodiments.

Fig. 1 is an interaction diagram of an information response system 10 based on artificial intelligence and cloud computing according to an embodiment of the present application. The information response system 10 based on artificial intelligence and cloud computing can comprise an information interaction platform 100 and an information interaction terminal 200 which is in communication connection with the information interaction platform 100. The artificial intelligence and cloud computing based information response system 10 shown in fig. 1 is only one possible example, and in other possible embodiments, the artificial intelligence and cloud computing based information response 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 information interaction platform 100 and the information interaction terminal 200 in the information response system 10 based on artificial intelligence and cloud computing may cooperatively execute the information response method based on artificial intelligence and cloud computing described in the following method embodiment, and the specific steps executed by the information interaction platform 100 and the information interaction 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 an information response method based on artificial intelligence and cloud computing according to an embodiment of the present application, where the information response method based on artificial intelligence and cloud computing according to the present embodiment may be executed by the information interaction platform 100 shown in fig. 1, and the information response method based on artificial intelligence and cloud computing is described in detail below.

Step S110, obtaining at least one response input question and a target response service corresponding to each response input question, and sequentially searching each response input question from a response index sequence configured in the cloud.

The answer input question may be a question that needs to be answered and configured, and specifically may be at least one of a text question to be answered and a graphic question to be answered. The target response service may be understood as a business service corresponding to the response input question, such as an order business service, a live business service, etc., but is not limited thereto.

For example, the information interaction platform 100 may detect a response request, and when the response request is detected, analyze and obtain at least one response input question and a target response service corresponding to each response input question from the response request. The target response service corresponding to the response input problem may be a preset response service or a response service calculated according to a decision algorithm.

For example, when an input response input question is detected, it is determined that each response input question corresponds to a response service identifier, and a target response service corresponding to each response input question is determined according to the response service identifier.

For example, the corresponding target response service may be determined according to a preset corresponding decision rule and a response service identifier corresponding to each response input question. The preset corresponding decision rule may specifically be a corresponding relationship between a preset response service identifier and a target response service, and different response service identifiers may correspond to the same preset corresponding rule or different preset corresponding rules.

Wherein, different response service identifications can correspond to the same preset corresponding rule. For example, the information interaction platform 100 may preset that the service node information of the target response service is the same as or in a certain corresponding relationship with the service node information of the response service identifier (for example, the service node information of the target response service may be calculated by performing weight calculation on the service node information of the response service identifier).

For example, the information interaction platform 100 may preset a decision manner of the target response service corresponding to each different response service identifier, and after the information interaction platform 100 determines the response service identifier for responding to the input question, the target response service may be calculated according to the preset decision manner.

For example, when the number of answer input questions is more than one, the answer input questions may be divided into at least one group according to the association relationship of the answer service identifiers of the answer input questions, for example, the answer input questions associated with the answer service identifiers may be divided into the same group, and the target answer services of the answer input questions in the same group are adjacent. In this case, the target response service corresponding to each response input question may be specifically a target response service corresponding to each group of response input questions. The target response service corresponding to a certain group of response input questions may specifically be a target response service of an initial response input question in the group of response input questions.

For example, the information interaction terminal can generate a response request according to the input response input question and the determined corresponding target response service. And triggering a question response request at the information interaction terminal locally through the response request.

In this embodiment, the answer index sequence is index sequence answer question data, which may also be referred to as lattice answer question data or update answer question data, and the information interaction platform 100 may update various questions on the answer index sequence. For example, the information interaction platform 100 may store the response index sequence in the cloud-configured storage medium, and when the response request is obtained, may traverse the existing problems in the cloud-configured response index sequence to search the response input problems from the cloud-configured response index sequence.

In the process of sequentially searching each response input problem from the response index sequence configured at the cloud, the problem semantic features and the problem scene features respectively corresponding to each response input problem can be determined. The question semantic features are semantic features obtained by performing semantic feature analysis processing on the response input questions and are used for representing the response input questions based on semantic coding features. The question scene feature is information indicating a scene attribute of the response input question.

For example, the information interaction platform 100 may obtain a response request, where the response request carries a response input question and a question scene feature of the response input question, and the information interaction platform 100 may encode the response input question to obtain a question semantic feature.

For example, when the user triggers an answer request, the information interaction platform 100 may determine an answer service identifier corresponding to the input answer input question. The information interaction platform 100 may determine, according to the preset correspondence between different response service identifiers and different question scene characteristics, the question scene characteristics corresponding to the input response input question according to the response service identifier corresponding to the input response input question.

For example, the information interaction platform 100 may preset the problem scene characteristics corresponding to the response service identifier 1 as follows: e-commerce, tape goods, digital products, etc.; the problem scene characteristics corresponding to the response service identifier 2 are as follows: e-commerce, carry-on-cargo, home products, etc. When the response service identifier corresponding to the current response input question is the response service identifier 1, the information interaction platform 100 may determine that the question scene characteristic corresponding to the response input question is e-commerce, delivery, digital product, or the like.

Therefore, response indexes respectively corresponding to the response input questions can be determined according to the question semantic features and the question scene features.

For example, the information interaction platform 100 may fuse the question semantic features and the scene features of the response input question to obtain fusion information, and perform hash operation on the fusion information to obtain a response index of the response input question. It is understood that the content of the answer input question can be determined by the question semantic feature corresponding to the answer input question, and the state of the answer input question can be determined by the question scene feature of the answer input question. The answer input question may be uniquely determined according to the question semantic features and the question scene features of the answer input question, and thus the answer index determined according to the question semantic features and the question scene features of the answer input question may be indexed to the answer input question. It will be appreciated that different answer input questions correspond to different answer indices. In this way, the response input questions are sequentially searched in the response index sequence configured at the cloud end through the response indexes corresponding to the response input questions.

For example, for an updated question, the information interaction platform 100 may combine the question semantic feature corresponding to each question and the corresponding question scene feature to generate a response index, and store the response index in association with the question in the response index sequence for searching the question. The information interaction platform 100 may store the updated response indexes of the questions together in the cloud to form a response index set.

For example, after determining the response index corresponding to each response input question, for each response input question, the information interaction platform 100 may respectively search whether the response index exists in the response index set configured in the cloud according to the corresponding response index, and if so, determine that the response input question exists in the response index sequence configured in the cloud; and if not, judging that the response input problem does not exist in a response index sequence configured by the cloud.

For example, the information interaction platform 100 may traverse the response indexes in the cloud-configured response index set to determine whether the response index of the current response input question exists in the response index set. If yes, judging that the response input problem exists in a response index sequence configured in the cloud end; and if not, judging that the response input problem does not exist in a response index sequence configured by the cloud.

For example, the information interaction platform 100 may store the response index and the corresponding response index object of the question updated in the response index sequence, and the question classification level information and the question frequency information of the updated question in association with each other. When the information interaction platform 100 finds the response input question in the response index sequence, the information such as the corresponding response index object, the question classification level information, the question frequent frequency information and the like can be determined through the response index of the response input question, so that the rapid configuration operation can be realized through the determined information.

In the above embodiment, the response indexes corresponding to the response input questions are determined according to the question semantic features and the question scene features corresponding to the response input questions, and the response input questions can be quickly and accurately searched in the response index sequence configured in the cloud.

Step S120, when the answer input question is found from the answer index sequence, determining an answer index object of the answer input question in the answer index sequence.

Specifically, when the information interaction platform 100 updates a question in the response index sequence, the response index object corresponding to the question is stored in association for subsequent use. When the information interaction platform 100 finds the answer input question from the answer index sequence, the answer input question may be considered as a history updated question, so that the information interaction platform 100 locally stores an answer index object corresponding to the answer input question, and the information interaction platform 100 may directly obtain the answer input question.

For example, the information interaction platform 100 may store a problem with a history update in the response index sequence and analyze the information of the response index object storing the problem in the response index sequence for subsequent use. Therefore, when the same response request is sent next time, the index question marking information of the response input question can be quickly returned, so that the information interaction platform 100 can perform configuration association operation through the previously updated question.

In this embodiment, the information interaction platform 100 may store the problem with the history update in the response index sequence, and associate and store the problem classification level information, the problem frequent frequency information, the response index object, and the like of the problem. The question classification level information is a classification label indicating the form of the question and a level where the label is located. The question frequency information is frequency information between the response input question and a response service of a similar response input question, and the question frequency information corresponding to each response input question can determine the frequency per unit time when a plurality of response input questions are sequentially responded. When the information interaction platform 100 receives a response request, it can search whether there is a current response input problem from a response index sequence of a history record, and if so, it can directly retrieve associated problem classification level information, problem frequent frequency information, a response index object, and the like, and configure the response input problem for a target response service according to the retrieved information.

In this embodiment, the response index object may be understood as output label information for the response input question, and may be understood as solution information of the response input question or an information index tag corresponding to the response information.

Step S130, when the answer input question is not found in the answer index sequence, updating the not found answer input question in the extended update sequence of the answer index sequence, and determining an answer index object of the updated answer input question in the answer index sequence.

Step S140, after determining the response index object of each response input question in the response index sequence, triggering to associate each response input question with the corresponding target response service according to the response index sequence of each response input question and the response index object of each response input question in the response index sequence.

Based on the above steps, in this embodiment, by obtaining at least one response input question and a target response service corresponding to each response input question, each response input question is sequentially searched from a response index sequence configured at the cloud, when a certain response input question is searched from the response index sequence, a response index object can be quickly responded, and configuration association operation of the target response service is directly executed through a question updated at the cloud, so that configuration time of the response input question is greatly reduced. When a certain response input problem is not found in the response index sequence, the response input problem which is not found is updated in the expansion updating sequence of the response index sequence to update the response index sequence, the configuration association operation is triggered and executed through the updated response index sequence and the response input problem in the response index sequence, the updated problem is stored in the response index sequence, and the configuration association operation can be directly executed when the next response meets the same response input problem, so that the configuration association operation of the target response service can be efficiently realized on the premise of not additionally adding a data set aiming at the target response service, and the service configuration efficiency is improved.

In one possible implementation manner, for step S130, for the process of updating the unsearched answer input question in the extended update sequence of the answer index sequence and determining the answer index object of the updated answer input question in the answer index sequence when no answer input question is found in the answer index sequence, the following exemplary sub-steps can be implemented, which are described in detail below.

And a substep S131, when the answer input question is not found in the answer index sequence, updating the question deep learning labeling information corresponding to the answer input question not found.

And a substep S132, according to the problem deep learning labeling information, distributing an update node for the unsearched response input problem in the expansion update sequence of the problem expansion area included in the response index sequence.

And a substep S133, updating the response input question which is not found at the allocated update node in the response index sequence, so as to update the response index sequence.

Substep S134 determines a response index object of the updated response input question in the updated response index sequence according to the update node.

The extended update sequence refers to a spatial region in the response index sequence where the problem is not updated. Specifically, when the information interaction platform 100 does not find the answer input question from the answer index sequence, it may be considered that the answer input question has not occurred in the previous answer operation, and therefore the answer input question needs to be updated to the answer index sequence in order to perform the subsequent configuration association operation.

For example, the information interaction platform 100 may perform standard information distribution on a service data region (i.e., an update node) corresponding to an answer input question according to the question deep learning label information of the answer input question, partition a corresponding service data region in the extended update sequence of the answer index sequence, and update the unsearched answer input question in the partitioned service data region. The information interaction platform 100 may use the service data area allocated for the undetected answer input question as an answer index object of the answer input question in the answer index sequence.

For example, when the information interaction platform 100 updates the response input question that is not found in the response index sequence, the information interaction platform 100 can record and analyze the updated question. The information interaction platform 100 may determine a question semantic feature and a question scene feature corresponding to the response input question updated in the response index sequence, combine the question semantic feature corresponding to the response input question updated in the response index sequence and the corresponding question scene feature to generate a response index, and store the constructed response index and a response index object of the corresponding response input question in the response index sequence in an associated manner.

For example, the information interaction platform 100 may fuse the question semantic features and the scene features to obtain fusion information, and perform hash operation on the fusion information to obtain a response index corresponding to the response input question updated in the response index sequence.

For example, the information interaction platform 100 may store the response index and the corresponding response index object of the question updated in the response index sequence, and the question classification level information and the question frequency information of the updated question in association with each other. When the information interaction platform 100 finds the response input problem in the response index sequence, the corresponding information such as the response index object, the problem classification level information, the problem frequent frequency information and the like can be determined by changing the response index of the response input problem, so that the rapid screen-up operation can be realized through the determined information.

For example, for a question that the information interaction platform 100 has updated in the answer index sequence, when the information interaction platform 100 acquires the answer request for the question (i.e. the next answer input question), the information interaction platform 100 can quickly find the answer input question according to the answer index and quickly return the information such as the answer index object, the question classification level information, and the question frequency information of the answer input question in the answer index sequence, so as to perform the answer operation using the previously updated question.

In the above embodiment, the response index corresponding to the updated response input question and the response index object of the response input question in the response index sequence are stored in association, so that the questions can be continuously accumulated in the response index sequence, and service configuration can be directly performed when the next response encounters the same question.

In a possible implementation manner, the present embodiment may further specifically determine the question frequency information corresponding to each response input question according to the foregoing implementation manner, so that in step S140, after determining the response index object of each response input question in the response index sequence, each response input question may be configured and associated with the corresponding target response service according to the question frequency information corresponding to each response input question according to the response index sequence including each response input question and the response index object of each response input question in the response index sequence.

In another possible implementation, for example, with respect to step S140, after determining the response index object of each response input question in the response index sequence, the question knowledge map may be generated according to the response index sequence including each response input question, so that by invoking the information response service, based on the question knowledge map and the response index object of each response input question in the response index sequence, the configuration of each response input question is triggered to be associated with the corresponding target response service.

In a further possible implementation manner, for step S131, in the process of updating the problem deep learning label information corresponding to the unsearched response input problem, the unsearched response input problem may be input into the problem deep learning label network, and the problem deep learning label information corresponding to the response input problem is obtained and updated. The problem deep learning labeling information comprises problem classification characteristic information and problem response characteristic information of response input problems.

The problem deep learning labeling network is configured and obtained in the following manner, which is described in detail below.

Step S101, obtaining calibration response problem data used for training a problem deep learning labeling network, wherein the calibration response problem data at least comprises response problems and target problem labeling information of the response problems.

Step S102, determining calibration data vector distribution and deviation correction data vector distribution of the calibration response problem data, wherein the deviation correction data vector distribution is formed by correcting the semantic splitting response characteristic component of the calibration response problem, and the characteristic labels of the calibration data vector distribution and the deviation correction data vector distribution are the same.

Step S103, carrying out feature aggregation on the calibration data vector distribution and the deviation correction data vector distribution to obtain aggregate calibration features, and determining prediction problem marking information of the aggregate calibration features.

And step S104, determining a difference function value based on the target problem labeling information and the predicted problem labeling information, adjusting network weight information of the problem deep learning labeling network based on the difference function value, continuing iterative training until an iteration stopping condition is met, and using the target problem labeling information to represent confidence coefficients of the response problems in the calibrated response problem data under the labeling information of each different problem.

The problem deep learning labeling network obtained through training can be used for classifying the labeling information of the unsearched response input problems.

The deviation correction data vector distribution can be obtained through the following method:

(1) and performing problem semantic splitting on the calibrated response problem data to obtain a plurality of semantic splitting problem data, and determining the influence weight of each semantic splitting problem data vector distribution.

(2) And sequencing all the influence weights to obtain a sequencing result, and correcting the vector distribution of the plurality of semantic splitting problem data according to the sequencing characteristics of the sequencing result to obtain correction calibration response problem data consisting of the plurality of semantic splitting problem data.

(3) Inputting the deviation correction calibration response problem data into a problem deep learning labeling network for feature extraction to obtain deviation correction data vector distribution, wherein the deviation correction data vector distribution comprises a plurality of semantic splitting problem data vector distributions.

In this way, in the training process of the problem deep learning labeling network, when the feature extraction is carried out on the calibration response problem data in the training calibration, the calibration data vector distribution of the original calibration response problem data and the deviation correction data vector distribution after deviation correction is carried out on the calibration data vector distribution are considered, and the aggregated calibration feature is obtained by aggregating the response problem data between the calibration data vector distribution and the deviation correction data vector distribution; further, by mapping the aggregated calibration feature to the predicted problem labeling information, which is the prediction result of the model, it is also necessary to determine a difference function value in combination with the target problem labeling information, and update the network weight information by the difference function value.

In the training process, not only calibration data vector distribution of original calibration response problem data but also deviation correction data vector distribution are considered, so that global features in the calibration response problem data can be learned in the training process, and deviation correction features aiming at target objects can be learned, more comprehensive and accurate features can be extracted by the trained problem deep learning labeling network, after more comprehensive features are extracted, more detectable features of the target objects in the response problem data to be detected can be extracted in the model application process, the feature recognition capability of the problem deep learning labeling network is enhanced, and then the labeling information classification result of the response problem data can be determined more accurately. In the training process of the problem deep learning labeling network, data amplification is carried out on the aspect of the calibration response problem data and the aspect of the deviation rectification characteristic, the integral structural information of the original calibration response problem data is damaged, the problem deep learning labeling network pays attention to the deviation rectification information, then the data space is filled through the deviation rectification between the similar characteristic and the heterogeneous characteristic, and the generalization of the network is improved.

In a possible implementation manner, regarding step S103, in the process of performing feature aggregation on the calibration data vector distribution and the deviation correction data vector distribution to obtain an aggregated calibration feature, the following exemplary sub-steps may be implemented, which are described in detail below.

And a substep S1031, adopting a problem deep learning labeling network to obtain the positive calibration characteristics of the positive calibration response problem data, the negative calibration characteristics of the negative calibration response problem data and the vector distribution of the positive calibration deviation correction data, and performing first vector segment aggregation on the positive calibration characteristics and the vector distribution of the positive calibration deviation correction data to obtain first aggregated calibration characteristics.

And a substep S1032 of adopting a problem deep learning labeling network to obtain a negative calibration feature and a negative calibration deviation correction data vector distribution of the negative calibration response problem data, and performing second vector fragment polymerization on the negative calibration feature and the negative calibration deviation correction data vector distribution to obtain a second polymerized calibration feature.

In sub-step S1033, respective third vector segments in the first aggregate alignment feature and respective fourth vector segments in the second aggregate alignment feature are determined.

And a substep S1034, for the feature formation node of each vector segment, aggregating the third vector segment of the third comprehensive weight and the fourth vector segment of the fourth comprehensive weight to obtain a first intersection aggregate calibration sub-feature of the feature formation node of the vector segment, and aggregating the third vector segment of the fourth comprehensive weight and the fourth vector segment of the third comprehensive weight to obtain a second intersection aggregate calibration sub-feature of the feature formation node of the vector segment.

In this embodiment, the sum of the third comprehensive weight and the fourth comprehensive weight is 1.

And a substep S1035 of determining a third aggregate nominal feature according to the first aggregate nominal sub-feature of the feature formation node of all the vector segments, and determining a fourth aggregate nominal feature according to the second aggregate nominal sub-feature of the feature formation node of all the vector segments.

In the substep S1036, the third aggregate calibration feature and the fourth aggregate calibration feature are used as aggregate calibration features.

In this embodiment, the vector distribution of the positive calibration deviation correction data is formed by correcting the semantic splitting response characteristic component of the positive calibration problem, the vector distribution of the negative calibration deviation correction data is formed by correcting the semantic splitting response characteristic component of the negative calibration problem, the positive calibration response problem data is used to represent correct calibration response problem data, and the negative calibration response problem data is used to represent wrong calibration response problem data.

In a possible implementation manner, the method provided by the embodiment of the present application may further include the following steps:

step S150, each target response service associated with each response input problem is operated and configured, and past emotion vector distribution and current emotion vector distribution obtained after emotion analysis is carried out on response service statistical data of each target response service on users of the information interaction terminal are obtained.

For example, the response service statistics may refer to process statistics for interacting with the user of the information interaction terminal during the response service of the target response service. Past emotion vector distribution and current emotion vector distribution may be divided by time period, for example, 12, 15 days of 2020, then past emotion vector distribution may be emotion vector distribution 12, 15 days before 2020, and current emotion vector distribution may be emotion vector distribution 12, 15 days after 2020, that is, past emotion vector distribution and current emotion vector distribution may be relative. Further, the emotion vector distribution is used to record relevant emotion information of the user corresponding to the response service statistics data, such as "happy", "excited", "liked", and "enjoyed", and the like.

For example, the target response service can be realized through the cloud computing service requested by the information interaction platform, so that the information throughput capacity is improved, and the response speed is increased.

And step S160, determining emotion distinguishing characteristic information represented by corresponding emotion characteristics in past emotion vector distribution and current emotion vector distribution, and determining target emotion characteristic representation which corresponds to the past emotion vector distribution and the current emotion vector distribution and meets the requirement of updating and tracking response behaviors based on the emotion distinguishing characteristic information represented by the corresponding emotion characteristics.

For example, the corresponding emotional feature representations in the past emotion vector distribution and the current emotion vector distribution may be understood as emotional feature representations of the same service node, e.g., the emotional feature representation of the "live e-commerce service node" in the past emotion vector distribution and the emotional feature representation of the "live e-commerce service node" in the current emotion vector distribution may be understood as corresponding emotional feature representations. The emotion distinguishing feature information is used for representing change information of past emotion vector distribution and current emotion vector distribution on emotion feature representations of the same object, for example, a favorite emotion vector exists for an A object of a live telecommuter service node in the past emotion vector distribution, and a counterintuitive emotion vector exists for the A object of the live telecommuter service node in the current emotion vector distribution, so that the emotion distinguishing feature information can be a difference comparison result between the favorite emotion vector and the counterintuitive emotion vector. Satisfying the response behavior update tracking requirement may be understood as a condition of satisfying the update tracking requirement, because slight changes in some emotional characteristic representations may not be enough to drive the response behavior update, while target emotional characteristic representations generally correspond to the key needs of the user, and therefore such emotional characteristic representations should be focused on.

Step S170, associating the target emotion feature representation in the current emotion vector distribution based on the target emotion feature representation in the past emotion vector distribution, and performing emotion transfer recognition on the associated emotion vector distribution in the current emotion vector distribution after the emotion feature representation is associated to obtain emotion transfer information.

For example, emotional feature representation association is a retrospective feature that globally integrates a series of relatively isolated emotional feature representations, such that the associated emotional vector distribution reflects the change in emotion from the whole. The emotion transfer information can be used for representing emotion transfer conditions between past emotion vector distribution and current emotion vector distribution, the emotion transfer information can be understood as changes of user consultation intention objects, the reasons of emotion changes can be determined by comprehensively analyzing changes of user requirements and using behavior information, and therefore the emotion transfer information is associated with actual service configuration information, and reliable guide basis can be provided for upgrading and updating of the service configuration information.

Step S180, determining first response process evaluation information corresponding to past emotion vector distribution and second response process evaluation information corresponding to current emotion vector distribution according to emotion transfer information, updating response content of the target response service through the first response process evaluation information and the second response process evaluation information, and storing the updated response content into corresponding block chain service.

For example, the response process evaluation information may be usage behavior information of the user for different target response service elements, the response process evaluation information may include positive evaluation information and negative evaluation information, and the positive and negative evaluation information may provide the developer with the most direct business requirement of the user for the target response service elements, so that the target response service elements can be updated and upgraded based on the different response process evaluation information. For example, for some live e-commerce services, the improvement of the related functions may be performed according to the negative evaluation information in the response process evaluation information, and the maintenance of the related functions may be performed according to the positive evaluation information in the response process evaluation information. In this way, not only can the defects of the target response service element be improved, but also the competitive point of the target response service element can be maintained, so that the software and hardware cost caused by blind product updating is effectively reduced, and the intellectualization and low-cost upgrading of the target response service element are realized.

For example, in the process of updating the response content of the target response service through the first response process evaluation information and the second response process evaluation information, the response content at the response content classification level matched with the first response process evaluation information and the second response process evaluation information may be further selected to be updated, or may be updated in other ways, which is not a technical problem solved by the embodiments of the present application, and reference may be made to the prior art, which is not limited specifically.

Finally, the updated response content is stored in the corresponding block chain service, so that the feature change can be traced further in the following process.

Based on the steps, different emotion vector distributions can be analyzed to determine emotion transfer information, so that first response process evaluation information corresponding to past emotion vector distributions and second response process evaluation information corresponding to current emotion vector distributions can be obtained through the emotion transfer information, emotion change retrospective analysis is achieved, a big data basis is conveniently provided for response content updating of the target response service, the target response service can be enabled to be more capable of matching with intention tendencies of users corresponding to response service objects, and accuracy of service response is improved.

In the following, some alternative embodiments will be described, which should be understood as examples and not as technical features essential for implementing the present solution.

For some possible embodiments, the emotion distinguishing feature information for determining corresponding emotion feature representations in the past emotion vector distribution and the current emotion vector distribution described in step S160 may include the following content described in step S161 and step S162.

Step S161, determining the emotion encoding characteristics represented by each emotion characteristic in the past emotion vector distribution and the emotion encoding characteristics represented by each emotion characteristic in the current emotion vector distribution. For example, the emotion encoding feature may be performed according to a preset classification index, which is not described herein.

Step S162, determining emotion state change data corresponding to emotion feature representation in past emotion vector distribution and current emotion vector distribution based on emotion coding features represented by each emotion feature in past emotion vector distribution and emotion coding features represented by each emotion feature in current emotion vector distribution, wherein emotion distinguishing feature information comprises emotion state change data. For example, emotional state change data may characterize changes in some behavioral habits of a user while conducting business processes.

In this way, based on the above steps S161 and S162, some changes in behavior habits of the user when performing business processing can be taken into account, thereby ensuring that the emotion distinguishing feature information can match the actual situation of the user as much as possible.

For further embodiments, the emotional state change data represented by the corresponding emotional feature in the past emotional vector distribution and the current emotional vector distribution may be determined in a variety of ways, such as one of the three ways shown below.

A first way of determining emotional state change data: and determining service emotion category vectors represented by corresponding emotion characteristics in the past emotion vector distribution and the current emotion vector distribution to determine emotion state change data based on the emotion coding characteristics represented by each emotion characteristic in the past emotion vector distribution and the emotion coding characteristics represented by each emotion characteristic in the current emotion vector distribution. It will be appreciated that this approach takes into account the business emotion category vector to determine the emotional state change data.

Second way of determining emotional state change data: and determining associated habit data of the emotion coding features expressed by the corresponding emotion features in the past emotion vector distribution and the current emotion vector distribution to determine emotion state change data based on the emotion coding features expressed by the emotion features in the past emotion vector distribution and the emotion coding features expressed by the emotion features in the current emotion vector distribution. It will be appreciated that this way the associated habit data is taken into account to determine the emotional state change data.

A third way of determining emotional state change data: determining emotional behavior description information corresponding to emotional characteristic representation in past emotional vector distribution and current emotional vector distribution, and determining emotional state change data based on the determined emotional behavior description information and emotional coding characteristics corresponding to emotional characteristic representation in the past emotional vector distribution and the current emotional vector distribution. It will be appreciated that this approach takes into account emotional behavior description information to determine emotional state change data.

In practical implementation, the manners of determining the emotional state change data may be combined arbitrarily, and are not limited herein, so that the emotional state change data can be determined from multiple layers, thereby providing a data basis for subsequent emotional transition analysis.

For some possible embodiments, the determining of the target emotion feature representation corresponding between the past emotion vector distribution and the current emotion vector distribution and satisfying the response behavior update tracking requirement based on the emotion distinguishing feature information of the corresponding emotion feature representation described in step S160 may include the following.

Firstly, updating the behavior content of the target response service, wherein the corresponding emotion characteristics in the past emotion vector distribution and the current emotion vector distribution represent user behavior information corresponding to emotion distinguishing characteristic information.

Secondly, the determination of the target emotional characteristic representation may be performed in any one or more of the following exemplary embodiments.

(1) And determining target emotion characteristic representation according to the corresponding emotion scene characteristics after updating. For example, the emotional scene characteristics may be derived from a running log of the target response server.

(2) And determining a target emotion characteristic representation according to the corresponding emotion directing characteristics of the updated corresponding emotion characteristic representation. For example, the emotional pointing feature is used to indicate relevant guidance information represented by the emotional feature.

(3) And determining the corresponding emotion characteristic representation of the words with emotion distinguishing characteristic information satisfying the emotion characteristic representation as the target emotion characteristic representation. For example, the emotional characteristic expression words may be set in advance, and are not limited herein.

(4) And performing semantic recognition on each corresponding emotion characteristic representation included in the corresponding emotion characteristic representation of the emotion distinguishing characteristic information unsatisfied emotion characteristic representation words according to a preset semantic recognition strategy, and determining target emotion characteristic representation based on a semantic recognition result. For example, the predetermined semantic recognition policy may be set in advance, and is not limited herein.

(5) And selecting the target emotion characteristic representation based on the target response service tracking information corresponding to the emotion characteristic representation. For example, the target response service tracking information may be information related to after-sales service or the like for user tracking.

In this way, the implementation mode of determining the target emotional characteristic representation can be flexibly selected according to different application scenes or actual situations, and thus the flexible implementation of the whole scheme can be ensured.

Further, in (4), semantic recognition is performed on each corresponding emotional feature representation included in the corresponding emotional feature representation of the emotional feature representation words whose emotional discrimination feature information does not satisfy the emotional feature representation words in order according to a predetermined semantic recognition policy, a target emotional feature representation is determined based on a semantic recognition result, by determining semantic identification strategies corresponding to emotion distinguishing feature information for which the emotion distinguishing feature information does not satisfy the emotion distinguishing feature information for each corresponding emotional feature representation included in the corresponding emotional feature representation of the emotional feature representation word, then semantic recognition is carried out on each corresponding emotion characteristic representation included in the corresponding emotion characteristic representation of the emotion distinguishing characteristic information unsatisfied emotion characteristic representation words according to the determined semantic recognition strategy, and obtaining target emotion characteristic representation according to the corresponding emotion characteristic representation after semantic recognition and the corresponding semantic word clustering result.

Further, in (5), the target emotion feature representation is selected based on the target response service tracking information corresponding to the emotion feature representation, the corresponding emotion feature representation with contextual emotion intensity variation may be selected according to the emotion scene features, and the first emotion update information of the corresponding emotion feature representation with contextual emotion intensity variation is determined, where the number of the corresponding emotion feature representations with contextual emotion intensity variation is a positive integer M.

On the basis, corresponding emotion characteristic representations with non-situational emotion intensity changes can be selected in sequence, second emotion updating information with the corresponding emotion characteristic representations with the non-situational emotion intensity changes is determined, and when it is determined that the updating information matching result of the first emotion updating information and the second emotion updating information meets the emotion determining condition, the corresponding emotion characteristic representations with the situational emotion intensity changes are determined to be target emotion characteristic representations.

In addition, when the fact that the updated information matching results of the first emotion updating information and the second emotion updating information do not meet the emotion determining condition is determined, selecting one more emotion characteristic representation corresponding to the situation emotion intensity change than the previous selecting number is repeatedly executed until emotion dictionary matching information of the emotion characteristic representation corresponding to the situation emotion intensity change selected later and the emotion dictionary matching information of the emotion characteristic representation corresponding to the situation emotion intensity change selected last are selected to be one of preset emotion dictionary matching information, and the corresponding emotion characteristic representation selected last is determined to be the target emotion characteristic representation.

In practical applications, the inventor finds that, in order to implement complete association of the emotion feature representations so as to reflect the actual situation of the user as much as possible globally, in step S170, the target emotion feature representation in the current emotion vector distribution is associated based on the target emotion feature representation in the past emotion vector distribution, and the following description of steps S171 to S175 may be included.

Step S171, each emotion feature representation included in emotion transfer information in past emotion vector distribution is represented through a multi-dimensional emotion class bitmap, each emotion feature representation represented by the multi-dimensional emotion class bitmap forms a past emotion feature representation set, and emotion class vector extraction and emotion class vector association are performed on the past emotion feature representation set to obtain a past emotion class vector map.

And step S172, expressing each emotion characteristic representation included in the current response behavior updating information in the current emotion vector distribution through the multi-dimensional emotion category bitmap, expressing each emotion characteristic represented by the multi-dimensional emotion category bitmap to form a current emotion characteristic expression set, and extracting emotion category vectors and associating the emotion category vectors of the current emotion characteristic expression set to obtain a current emotion category vector map.

In step S173, emotion migration map extraction is performed on the past response behavior update information in the past emotion vector distribution based on the past emotion category vector map to obtain a past emotion migration map.

Step S174, judging whether the response interaction information of the current emotion category vector map corresponding to each current response behavior updating information in the current emotion vector distribution meets the index requirement corresponding to the current response interaction or not, and when the response interaction information meets the index requirement corresponding to the current response interaction, performing emotion transfer migration map extraction on each current response behavior updating information in the current emotion vector distribution to obtain a current emotion transfer migration map, wherein the preset first response interaction reference information is the target response service guide information of the past emotion category vector map corresponding to the past response behavior updating information in the past emotion vector distribution and the user service survey result counted in advance.

And step S175, correlating target emotion characteristic representation in the current emotion vector distribution through the migration diagram comparison information between the current emotion transfer migration diagram and the past emotion transfer migration diagram. For example, the emotional transition diagram may be expressed in any form, and is not limited herein.

Therefore, a past emotion transfer transition diagram of past emotion vector distribution and a current emotion transfer transition diagram of current emotion vector distribution can be respectively determined, so that time sequence continuity in analyzing the past emotion vector distribution and the current emotion vector distribution can be ensured, and further, target emotion characteristic representation can be completely and real-timely correlated through transition diagram comparison information between the past emotion transfer transition diagram and the current emotion transfer transition diagram, so that the actual situation of a user can be reflected as much as possible in a global manner.

For further embodiments, the emotion transfer migration map extraction of the past response behavior update information in the past emotion vector distribution based on the past emotion category vector map described in step S173 to obtain a past emotion transfer migration map includes: and judging whether response interaction information of a past emotion category vector map corresponding to each past response behavior updating information in the past emotion vector distribution meets the index requirement corresponding to the past response interaction. Adding emotion transfer labels to response interaction information of past emotion category vector maps of past response behavior update information, wherein the past emotion category vector maps meet index requirements corresponding to past response interaction, determining emotion feedback degree information corresponding to response interaction information of past emotion category vector maps of other past response behavior update information, and generating a past emotion transfer migration map according to the response interaction information added with the emotion transfer labels and the emotion feedback degree information corresponding to response interaction information of the past emotion category vector maps of other past response behavior update information.

In a possible implementation manner, it is determined in step S174 whether the result of the information comparison between the response interaction information of the current emotion category vector map corresponding to each current response behavior update information in the current emotion vector distribution and the preset first response interaction reference information meets the index requirement corresponding to the current response interaction.

For example, the preset first response interaction reference information is a past emotion category vector map corresponding to the past response behavior update information in the past emotion vector distribution and target response service guidance information of the user service survey result counted in advance, and specifically may be: and judging whether the response interaction information of the current emotion category vector map corresponding to each current response behavior updating information in the current emotion vector distribution meets the index requirement corresponding to the current response interaction or not according to the information comparison result of preset first response interaction reference information, wherein the preset first response interaction reference information is the past emotion category vector map corresponding to the past response behavior updating information in the past emotion vector distribution and target response service guide information of a user service investigation result counted in advance. On the basis, adding emotion transfer labels to the current emotion category vector map of the current response behavior update information, the information comparison result of which meets the index requirement corresponding to the current response interaction, of the current emotion category vector map of the current response behavior update information, determining emotion feedback degree information corresponding to response interaction information of the current emotion category vector map of other current response behavior update information, and generating a current emotion transfer migration map according to the response interaction information added with the emotion transfer labels and emotion feedback degree information corresponding to response interaction information of past emotion category vector maps of other current response behavior update information.

In an actual implementation process, in order to ensure high correlation between emotion transfer information and the target response service, in step S170, emotion transfer recognition is performed on an associated emotion vector distribution in the current emotion vector distribution after emotion feature representation association to obtain emotion transfer information, which may specifically be: and comparing response interaction information which is subjected to emotion characteristic representation after emotion characteristic representation association and is included in the associated emotion vector distribution with response interaction information which is not subjected to emotion characteristic representation before emotion characteristic representation association, and determining the response interaction information of the associated emotion vector distribution by screening response interaction information which meets emotion transfer indexes according to the comparison result of selecting the response interaction information which is subjected to emotion characteristic representation after emotion characteristic representation association and is not subjected to emotion characteristic representation before emotion characteristic representation association and response interaction information which is not subjected to emotion characteristic representation before emotion characteristic representation association, so as to determine emotion transfer information according to the response interaction information of the associated emotion vector distribution.

Therefore, response interaction information related to emotion vector distribution is determined by considering the response interaction information before and after emotion feature representation and the response interaction information related to emotion vector distribution, emotion transfer information can be determined according to the response interaction information related to emotion vector distribution, high correlation between the emotion transfer information and target response service is further ensured, and subsequent targeted upgrade of related target response service is facilitated.

In a possible implementation manner, in order to ensure that the response process evaluation information can comprehensively cover the use process and the use feeling of the product, the user requirement, the consultation source object and the related abnormal information need to be considered. Based on this, determining the first answer process evaluation information corresponding to the past emotion vector distribution and the second answer process evaluation information corresponding to the current emotion vector distribution from the emotion transfer information described in step S180 may include the following steps S181 to S186.

Step S1801, according to the service response record of the response service statistical data, performing intent classification on the target response intention data corresponding to the emotion transfer information into a plurality of intention tags, and according to the target response service element data corresponding to the target response service event corresponding to the intention content of each intention tag, determining the service element interaction information of each target response service event. The service element interaction information comprises element data interacted with the user.

Step S1802, after determining the service element interaction information of each target response service event, perform consultation source object analysis on the service element interaction information of each target response service event, determine a consultation source object result of each service element interaction information, and determine first abnormal behavior generation information of the target response service event corresponding to the target response intention data according to the consultation source object result of each service element interaction information and the service element interaction information of each target response service event.

Step S1803, for each exception category, determining a target response interaction record corresponding to each target response service event according to the association relationship between each target response service event and at least one exception category, and determining first exception information corresponding to each exception category according to the user emotional performance record of each target response service event in the first exception behavior generation information and the target response interaction record corresponding to each target response service event.

Step S1804, determining a target abnormality type corresponding to the target response intention data according to the first abnormality information corresponding to each abnormality type.

Step S1805, determining a target response service evaluation score indicator of each target response service event according to target response service difference information in a target exception category between each target response service event and a first target response service event with a highest target response service evaluation score, target response service difference information in a target exception category between a first target response service event and a second target response service event with a lowest target response service evaluation score, and service element interaction information of each target response service event.

Step S1806, determining a target response service evaluation score index of each intention event in the intention tag corresponding to each target response service event according to the target response service evaluation score index of each target response service event. And classifying response process evaluation information of each intention event in the target response intention data according to the target response service evaluation score index of each intention event to obtain first response process evaluation information corresponding to past emotion vector distribution and second response process evaluation information corresponding to current emotion vector distribution.

Therefore, the target response intention data, the intention label, the target response service element data and the related information generated by the abnormal behavior can be taken into consideration, and then the user demand information, the consultation source object and the related abnormal information are comprehensively considered, so that the response process evaluation information can be ensured to comprehensively cover the use process and the use feeling of the user for the product, and a complete and reliable decision basis is provided for the subsequent upgrading and improvement of the target response service element.

In one possible implementation manner, in step S182, determining first abnormal behavior generation information of the target response service event corresponding to the target response intention data according to the result of consulting the source object of each service element interaction information and the service element interaction information of each target response service event may include: and when the service element interaction information of the target response service event is matched with the consultation source object result, the target response service event is correspondingly recorded without exception in the first abnormal behavior generation information. And when the service element interaction information of the target response service event does not match the consultation source object result, the target response service event corresponds to a target abnormal record in the first abnormal behavior generation information, wherein the target abnormal record and the service element interaction information have time sequence correlation.

Fig. 3 is a schematic diagram of functional modules of an information response device 300 based on artificial intelligence and cloud computing according to an embodiment of the present disclosure, in this embodiment, the information response device 300 based on artificial intelligence and cloud computing may be divided into the functional modules according to a method embodiment executed by the information interaction platform 100, that is, the following functional modules corresponding to the information response device 300 based on artificial intelligence and cloud computing may be used to execute each method embodiment executed by the information interaction platform 100. The information answering device 300 based on artificial intelligence and cloud computing may include an obtaining module 310, a first determining module 320, a second determining module 330, and a configuring module 340, and the functions of the functional modules of the information answering device 300 based on artificial intelligence and cloud computing are described in detail below.

The obtaining module 310 is configured to obtain at least one response input question and a target response service corresponding to each response input question, and sequentially search each response input question from a response index sequence configured in the cloud. 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 first determining module 320 is configured to determine a response index object of the response input question in the response index sequence when the response input question is found in the response index sequence. The first determining module 320 may be configured to perform the step S120, and for a detailed implementation of the first determining module 320, reference may be made to the detailed description of the step S120.

The second determining module 330 is configured to, when no answer input question is found in the answer index sequence, update the not found answer input question in the extended update sequence of the answer index sequence, and determine an answer index object of the updated answer input question in the answer index sequence. The second determining module 330 may be configured to perform the step S130, and the detailed implementation of the second determining module 330 may refer to the detailed description of the step S130.

The configuration module 340 is configured to, after determining the response index objects of the response input questions in the response index sequence, associate the response input questions with the corresponding target response service according to the response index sequence of the response input questions and the response index objects of the response input questions in the response index sequence. The configuration module 340 may be configured to perform the step S140, and the detailed implementation manner of the configuration 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 may all be implemented in software invoked by a 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.

Fig. 4 is a hardware schematic diagram of an information interaction platform 100 for implementing the artificial intelligence and cloud computing-based information response method, provided by the embodiment of the present disclosure, and as shown in fig. 4, the information interaction 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 first determining module 320, the second determining module 330, and the configuring module 340 included in the artificial intelligence and cloud computing based information answering device 300 shown in fig. 3), so that the processor 110 may execute the artificial intelligence and cloud computing based information answering method according to the above method embodiment, where the processor 110, the machine-readable storage medium 120, and the transceiver 140 are connected through the bus 130, and the processor 110 may be configured to control the transceiving action of the transceiver 140, so as to perform data transceiving with the aforementioned information interaction terminal 200.

For a specific implementation process of the processor 110, reference may be made to the above-mentioned method embodiments executed by the information interaction 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 application also provides a readable storage medium, wherein the readable storage medium stores computer execution instructions, and when a processor executes the computer execution instructions, the information response method based on artificial intelligence and cloud computing is realized.

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.

Having thus described the basic concept, it will be apparent to those skilled in the art that the foregoing detailed disclosure is to be regarded as illustrative only and not as limiting the present specification. Various modifications, improvements and adaptations to the present description may occur to those skilled in the art, although not explicitly described herein. Such modifications, improvements and adaptations are proposed in the present specification and still fall within the spirit and scope of the exemplary embodiments of the present specification.

Also, particular push elements are used in this description to describe embodiments of this description. Reference to "one embodiment," "an embodiment," and/or "some embodiments" means that a particular feature, structure, or characteristic described in connection with at least one embodiment of the specification. Therefore, it is emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, some features, structures, or characteristics of one or more embodiments of the specification may be combined as appropriate.

Moreover, those skilled in the art will appreciate that aspects of the present description may be illustrated and described in terms of several patentable species or contexts, including any new and useful combination of processes, machines, manufacture, or materials, or any new and useful improvement thereof. Accordingly, aspects of this description may be performed entirely by hardware, entirely by software (including firmware, resident software, micro-code, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data block," module, "" engine, "" unit, "" component, "or" system. Furthermore, aspects of the present description may be represented as a computer product, including computer readable program code, embodied in one or more computer readable media.

The computer storage medium may comprise a propagated data signal with the computer program code embodied therewith, for example, on baseband or as part of a carrier wave. The propagated signal may take any of a variety of forms, including electromagnetic, optical, etc., or any suitable combination. A computer storage medium may be any computer-readable medium that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, or device. Program code located on a computer storage medium may be propagated over any suitable medium, including radio, cable, fiber optic cable, RF, or the like, or any combination of the preceding.

Computer program code required for the operation of various portions of this specification may be written in any one or more programming languages, including an object oriented programming language such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C + +, C #, VB.NET, Python, and the like, a conventional programming language such as C, VisualBasic, Fortran2003, Perl, COBOL2002, PHP, ABAP, a passive programming language such as Python, Ruby, and Groovy, or other programming languages. The program code may run entirely on the user's computer, as a separate indexing sequence on the user's computer, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any network format, such as a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet), or in a cloud computing environment, or as a service, such as a software as a service (SaaS).

Furthermore, unless explicitly stated in the claims, the order in which the description deals with the problems and sequences, the use of alphanumeric characters, or the use of other names, is not intended to limit the order in which the processes and methods of the description flow. While various presently contemplated embodiments of the invention have been discussed in the foregoing disclosure by way of example, it is to be understood that such detail is solely for that purpose and that the appended claims are not limited to the disclosed embodiments, but, on the contrary, are intended to cover all modifications and equivalent arrangements that are within the spirit and scope of the embodiments herein. For example, although the system components described above may be implemented by hardware devices, they may also be implemented by software-only solutions, such as installing the described system on an existing server or mobile device.

Similarly, it should be noted that in the preceding description of embodiments of the present specification, various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure aiding in the understanding of one or more of the embodiments. This method of disclosure, however, is not intended to imply that more features than are expressly recited in a claim. Indeed, the embodiments may be characterized as having less than all of the features of a single embodiment disclosed above.

Finally, it should be understood that the examples in this specification are only intended to illustrate the principles of the examples in this specification. Other variations are also possible within the scope of this description. Thus, by way of example, and not limitation, alternative configurations of the embodiments of the specification can be considered consistent with the teachings of the specification. Accordingly, the embodiments of the present description are not limited to only those embodiments explicitly described and depicted herein.

Claims

1. An information response method based on artificial intelligence and cloud computing is characterized by being applied to an information interaction platform, wherein the information interaction platform is in communication connection with a plurality of information interaction terminals, and the method comprises the following steps:

2. The information response method based on artificial intelligence and cloud computing according to claim 1, wherein the obtaining of at least one response input question and a target response service corresponding to each response input question respectively comprises:

3. The information response method based on artificial intelligence and cloud computing according to claim 1, wherein the sequentially searching response input questions from a response index sequence configured in a cloud comprises:

4. The information answering method based on artificial intelligence and cloud computing according to claim 1, wherein when the answer input question is not found in the answer index sequence, updating the answer input question not found in the extended update sequence of the answer index sequence and determining an answer index object of the updated answer input question in the answer index sequence comprises:

5. The artificial intelligence and cloud computing based information response method according to claim 1, further comprising:

6. The information response method based on artificial intelligence and cloud computing according to claim 4, wherein the step of updating the question deep learning labeling information corresponding to the response input question that is not found includes:

7. The information response method based on artificial intelligence and cloud computing according to claim 6, wherein the step of performing feature aggregation on the calibration data vector distribution and the rectification data vector distribution to obtain aggregated calibration features comprises:

8. The information answering method based on artificial intelligence and cloud computing according to any one of claims 1-7, wherein the method further comprises:

9. The artificial intelligence and cloud computing based information response method according to claim 1, wherein the step of associating the target emotion feature representation in the current emotion vector distribution based on the target emotion feature representation in the past emotion vector distribution comprises:

10. An information interaction platform, which is characterized by comprising a processor, a machine-readable storage medium and a network interface, wherein the machine-readable storage medium, the network interface and the processor are connected through a bus system, the network interface is used for being in communication connection with at least one information interaction terminal, the machine-readable storage medium is used for storing programs, instructions or codes, and the processor is used for executing the programs, the instructions or the codes in the machine-readable storage medium so as to execute the information response method based on artificial intelligence and cloud computing according to any one of claims 1 to 9.