Question and answer data processing method and device, equipment, storage medium and product

Abstract

Embodiments of the present disclosure provide a question and answer data processing method and device, equipment, storage medium and product. The method comprises: obtaining an online question and answer data set, wherein the online question and answer data set comprises intention data generated by a question and answer robot corresponding to different question and answer fields; performing an aggregation operation on the online question and answer data set according to the intention data and a response evaluation index corresponding to the intention data, to obtain evaluation index data of the question and answer robot corresponding to each intention data; determining a comprehensive score of each question and answer robot corresponding to each intention data according to the evaluation index data; and performing an update operation on a preset question and answer network model by using target question and answer data of the question and answer robot whose comprehensive score under each intention data exceeds a preset threshold. Since the target question and answer data is a high-quality reply in different question and answer fields, the question and answer network model updated by using the target question and answer data can feed back a more high-quality reply according to the question information issued by a user, thereby improving the user experience.

Description

TECHNICAL FIELD

[0001] Embodiments of the present disclosure relate to the field of artificial intelligence, and in particular, to a question and answer data processing method and device, equipment, storage medium and product. BACKGROUND

[0002] An intelligent question and answer robot is based on natural language processing and deep learning, and can accurately locate the question knowledge required by a visitor in real time, and provide personalized information services for the visitor through interaction with the visitor.

[0003] In the question and answer model training process of the question and answer robot, a specific domain question and answer data is generally used to train the model, so that the model can answer questions in the specific domain.

[0004] In the implementation of the present disclosure, the inventors have found that at least the following problems exist in the prior art: the question and answer data in a specific domain usually has a small amount of data, and the quality of the question and answer data may be poor. Using the above data to train the model can cause the question and answer robot to be unable to give a high-quality reply according to a user's question, and thus the user experience is poor. SUMMARY

[0005] Embodiments of the present disclosure provide a question and answer data processing method, device, equipment, storage medium and product to solve the technical problem that the existing question and answer data processing method has a small amount of data and poor data quality, and thus the question and answer robot cannot provide high-quality services.

[0006] In a first aspect, embodiments of the present disclosure provide a question and answer data processing method, comprising:

[0007] obtaining a pre-stored online question and answer data set, wherein the online question and answer data set includes intention data generated by a question and answer robot corresponding to different question and answer domains;

[0008] performing an aggregation operation on the online question and answer data set according to the intention data and a reply evaluation index corresponding to the intention data, to obtain evaluation index data of the question and answer robot corresponding to each intention data;

[0009] determining a comprehensive score of each question and answer robot corresponding to each intention data according to the evaluation index data;

[0010] performing an update operation on a question and answer network model corresponding to each question and answer robot using target question and answer data of the question and answer robot whose comprehensive score under each intention data exceeds a preset threshold.

[0011] In a second aspect, embodiments of the present disclosure provide a question and answer data processing device, comprising:

[0012] The acquisition module is configured to acquire a pre-stored online question and answer data set, wherein the online question and answer data set comprises intention data generated by question and answer robots corresponding to different question and answer fields;

[0013] The aggregation module is configured to perform an aggregation operation on the online question and answer data set according to the intention data and the response evaluation index corresponding to the intention data, to obtain evaluation index data of the question and answer robots corresponding to each intention data;

[0014] The determination module is configured to determine a comprehensive score of each question and answer robot corresponding to each intention data according to the evaluation index data.

[0015] The update module is configured to perform an update operation on a question and answer network model corresponding to each question and answer robot by using target question and answer data of the question and answer robot whose comprehensive score under each intention data exceeds a preset threshold.

[0016] In a third aspect, an electronic device is provided, comprising a memory and a processor.

[0017] The memory is configured to store instructions executable by the processor.

[0018] The processor is configured to invoke the program instructions in the memory to execute the question and answer data processing method in the first aspect.

[0019] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the question and answer data processing method in the first aspect.

[0020] In a fifth aspect, a computer program product is provided, comprising a computer program, and the computer program is executed by a processor to implement the question and answer data processing method in the first aspect.

[0021] The question and answer data processing method, device, equipment, storage medium and product provided by the embodiments of the present disclosure can obtain evaluation index data of question and answer robots corresponding to each intention data by performing an aggregation operation on intention data generated by question and answer robots corresponding to different question and answer fields, and determine a comprehensive score of each question and answer robot corresponding to each intention data according to the evaluation index data. Thus, high-quality target question and answer data can be obtained in the online question and answer data set of different question and answer fields according to the comprehensive score, and the target question and answer data can be used to update the question and answer network model of the question and answer robot. Since the target question and answer data is high-quality reply in different question and answer fields, the question and answer network model updated by using the target question and answer data can feed back more high-quality reply according to the question information issued by the user, thereby improving the user experience. BRIEF DESCRIPTION OF DRAWINGS

[0022] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0023] Figure 1 This is a schematic diagram of the system architecture upon which this disclosure is based;

[0024] Figure 2 This is a flowchart illustrating the online question-and-answer data processing method provided in Embodiment 1 of this disclosure;

[0025] Figure 3 This is a schematic diagram illustrating an application scenario provided by an embodiment of this disclosure;

[0026] Figure 4 This is a flowchart illustrating the question-and-answer data processing method provided in Embodiment 2 of this disclosure;

[0027] Figure 5 This is a flowchart illustrating the question-and-answer data processing method provided in Embodiment 3 of this disclosure;

[0028] Figure 6 This is a flowchart illustrating the question-and-answer data processing method provided in Embodiment 4 of this disclosure;

[0029] Figure 7 This is a schematic diagram of the question-and-answer data processing device provided in Embodiment 5 of this disclosure;

[0030] Figure 8 This is a schematic diagram of the structure of an electronic device provided in Embodiment Six of this disclosure.

[0031] The accompanying drawings have illustrated specific embodiments of this disclosure, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this disclosure to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0032] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0033] In response to the aforementioned technical problems that existing question-and-answer data processing methods suffer from small data volume and poor data quality, resulting in question-and-answer robots being unable to provide high-quality services, this disclosure provides a question-and-answer data processing method, apparatus, device, storage medium, and product.

[0034] It should be noted that the question and answer data processing method, device, equipment, storage medium and product provided by the present disclosure can be applied to the scene of establishing an online question and answer platform, online question and answer robot training data set, and model updating.

[0035] In order to provide more real-time and accurate question and answer services for users, online question and answer robots are used in various fields to answer online questions of users. Specifically, the online question and answer robot can sort the basic conditions of itself and common questions and their answers accumulated for many years in the field into a standard question and answer data set to support intelligent question and answer of various forms of questions. However, since the existing question and answer data set is limited to a fixed field, the data amount may be small, and some question and answer data may not accurately answer the user's question, so the support of intelligent question and answer using the question and answer data set may not accurately solve the user's question, resulting in poor user experience.

[0036] In the process of solving the above technical problems, the inventors found through research that there may be question and answer data of the same intention in the process of question and answer robots in different fields processing questions. In order to improve the reply accuracy of the online question and answer robot and improve the user experience, the online question and answer data set can be established based on the question and answer data in different fields, and the high-quality reply in the online question and answer data set can be selected as the support of the online question and answer robot.

[0037] The question and answer data processing method provided by the embodiments of the present disclosure aims to solve the above technical problems of the prior art.

[0038] The technical solutions of the present disclosure and how the technical solutions of the present disclosure solve the above technical problems will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0039] Figure 1 The system architecture based on the present disclosure is shown in FIG. 1, which at least includes a server 1 and a database 2. The server 1 is provided with a question and answer data processing device. The question and answer data processing device is written in C / C++, Java, Shell or Python language; and the database 2 can be a cloud server or a server cluster, which stores a large amount of question and answer data. Figure 1

[0040] Figure 2 The flowchart of the online question and answer data processing method provided by the first embodiment of the present disclosure is shown in FIG. 2, which includes the following steps. Figure 2

[0041] ​​In step 201, a pre-stored online Q&A data set is obtained, wherein the online Q&A data set comprises intention data generated by Q&A robots corresponding to different Q&A fields.

[0042] The execution subject of the embodiment is an online Q&A data processing device, which can be coupled to a server. The server can be in communication connection with a database, so as to obtain an online Q&A data set from the database.

[0043] In the embodiment, since there can be high-quality Q&A data of the same intention in the process of handling questions by Q&A robots of different fields, in order to realize data support for the question answering of the Q&A robots, a pre-stored online Q&A data set can be obtained, wherein the online Q&A data set comprises intention data generated by Q&A robots corresponding to different Q&A fields. Optionally, the online Q&A data set can comprise all intention data generated by Q&A robots corresponding to different Q&A fields.

[0044] In step 202, the online Q&A data set is aggregated according to the intention data and response evaluation indexes corresponding to the intention data, to obtain evaluation index data of the Q&A robots corresponding to each intention data.

[0045] In the embodiment, each message of each session of each Q&A robot has corresponding intention, and each intention is associated with a response evaluation index. The response evaluation index includes but is not limited to user request transfer to human rate after response / response satisfaction / response solution rate, consultation conversion rate / consultation loss rate, etc.

[0046] The response evaluation indexes can be accurately associated with specific intentions, or each intention can share the response evaluation indexes. Therefore, after obtaining the online Q&A data set, the online Q&A data set can be aggregated according to the intention data and the response evaluation indexes corresponding to the intention data, to obtain the evaluation index data of the Q&A robots corresponding to each intention data.

[0047] In step 203, the comprehensive score of each Q&A robot corresponding to each intention data is determined according to the evaluation index data.

[0048] In the embodiment, after determining the evaluation index data of the Q&A robots corresponding to each intention data, the comprehensive score of each Q&A robot corresponding to each intention data can be determined according to the evaluation index data. The higher the comprehensive score is, the better the response ability of the Q&A robot to the intention data is.

[0049] In step 204, target Q&A data of Q&A robots whose comprehensive scores under each intention data exceed a preset threshold value is used to update Q&A network models corresponding to each Q&A robot.

[0050] In the embodiment, after determining the comprehensive scores of the question and answer robots corresponding to each intention data, the target question and answer data can be selected according to the comprehensive scores. Specifically, the question and answer data of the question and answer robot corresponding to the intention data of each intention data, whose comprehensive score exceeds a preset threshold, can be taken as the target question and answer data. Therefore, the target question and answer data can be used as the support of the online question and answer robot. Alternatively, the target question and answer data can be used to update the question and answer network model corresponding to the question and answer robot.

[0051] As an implementable manner, the question and answer robots can be ranked according to the comprehensive scores, and the question and answer data corresponding to the question and answer robots in the top preset number of the ranking can be selected as the target question and answer data.

[0052] Figure 3 The application scenario diagram provided by the embodiment of the present disclosure is shown in FIG. 1. Figure 3 As shown in FIG. 1, the high-quality target question and answer data 32 can be determined in the intention data 31 generated by the question and answer robots corresponding to different question and answer fields, and the target question and answer data 32 can be used as the support of the online question and answer robot 33.

[0053] The question and answer data processing method provided by the embodiment can obtain the evaluation index data of the question and answer robots corresponding to each intention data by performing the aggregation operation on the intention data generated by the question and answer robots corresponding to different question and answer fields, and determine the comprehensive scores of the question and answer robots corresponding to each intention data according to the evaluation index data. Therefore, the high-quality target question and answer data can be obtained in the online question and answer data set of different question and answer fields according to the comprehensive scores, and the target question and answer data can be used to update the question and answer network model of the question and answer robot. Since the target question and answer data is the high-quality reply in different question and answer fields, the question and answer network model updated by using the target question and answer data can feed back the more high-quality reply according to the question information issued by the user, thereby improving the user experience.

[0054] Figure 4 The flowchart of the question and answer data processing method provided by the second embodiment of the present disclosure is shown in FIG. 2. Figure 2 As shown in FIG. 2, step 202 specifically includes:

[0055] Step 401: For each intention data, the five-tuple data corresponding to each question and answer robot is generated according to at least one answer evaluation index corresponding to the intention data and the identifier of each question and answer robot.

[0056] Step 402: The aggregation operation is performed on the five-tuple data to obtain the to-be-processed matrix of the question and answer robots corresponding to each intention data.

[0057] In step 403, the uniform dimension processing is performed on the to-be-processed matrix to obtain the evaluation index matrix of the question and answer robot corresponding to the each intention data.

[0058] In this embodiment, for each intention data, each question and answer robot can generate the conversation data corresponding to the intention data, and thus, the quintuple data corresponding to each question and answer robot can be generated according to the at least one response evaluation index corresponding to the intention data and the identifier of each question and answer robot.

[0059] For example, the response evaluation index of the conversation corresponding to each intention of each question and answer robot is a quintuple <I, r, i, l, k>, where I is the intention code, r is the robot code, i is the conversation code, l is the index code, and k is the index specific value.

[0060] Further, since the response evaluation index can be accurately associated with a specific intention or each intention corresponding response evaluation index can be collectively allocated, the quintuple data can be aggregated to obtain the to-be-processed matrix of the question and answer robot corresponding to each intention data. Further, in order to improve the subsequent data processing efficiency, the uniform dimension processing can be performed on the to-be-processed matrix to obtain the evaluation index matrix of the question and answer robot corresponding to each intention data.

[0061] Specifically, on the basis of any of the above embodiments, step 403 comprises:

[0062] The to-be-processed matrix is subjected to one or more of normalization, standardization, and rank operation to obtain the evaluation index matrix of the question and answer robot corresponding to each intention data.

[0063] In this embodiment, one or more of normalization, standardization, and rank operation can be specifically used to realize the aggregation operation of the quintuple data.

[0064] For example, the uniform dimension processing can be specifically normalization operation. For example, formula 1 can be used to realize the uniform dimension processing on the to-be-processed matrix I jl to obtain the evaluation index matrix P jl of the question and answer robot corresponding to each intention data.

[0065] P jl = [I jl -min(I l )] / [max(I l )-min(I l )] (1)

[0066] max(I l ) is the maximum index specific value corresponding to the index l, and min(I l) is the minimum index specific value corresponding to the index 1.

[0067] The question and answer data processing method provided in this embodiment can obtain the evaluation index matrix of the question and answer robot corresponding to each intention data by aggregating and performing uniform dimension operation on the online question and answer data set, so that the subsequent calculation of the comprehensive score of the robot can be realized according to the evaluation index matrix, and then the target question and answer data with high quality can be obtained in the online question and answer data set in different question and answer fields according to the comprehensive score.

[0068] Figure 5 The flowchart of the question and answer data processing method provided in Embodiment Three of the present disclosure is shown in FIG. 3, which is based on any of the above embodiments and specifically includes the following steps. Figure 5

[0069] Step 501: determining the weight information corresponding to each evaluation index matrix by using an unsupervised learning method.

[0070] Step 502: determining the comprehensive score of each question and answer robot corresponding to each intention data according to the evaluation index matrix and the weight information corresponding to the evaluation index matrix.

[0071] In this embodiment, the weight information corresponding to each evaluation index matrix can be determined by using an unsupervised method. According to the weight information and the evaluation index matrix, the comprehensive score of each question and answer robot corresponding to each intention data is determined. Since the weight information corresponding to each evaluation index matrix is determined first in the process of calculating the comprehensive score, the accuracy of the comprehensive score of each question and answer robot can be improved.

[0072] Further, on the basis of any of the above embodiments, step 501 specifically includes:

[0073] determining the weight information corresponding to each evaluation index matrix by using a preset weight information determination algorithm;

[0074] The preset weight information determination algorithm includes any one or more of the following algorithms:

[0075] principal component algorithm, independent cost decomposition algorithm, or singular value decomposition algorithm, rough set algorithm.

[0076] Specifically, any one or more of the principal component algorithm, the independent cost decomposition algorithm, or the singular value decomposition algorithm, the rough set algorithm can be used to calculate the weight information corresponding to each evaluation index matrix, and the present disclosure does not limit this.

[0077] Further, on the basis of any of the above embodiments, step 203 specifically includes:

[0078] ​The evaluation index matrix and the weight information corresponding to the evaluation index matrix are subjected to a weighted summation operation.

[0079] The result of the weighted summation is determined as the comprehensive score of each question and answer robot corresponding to the intent data.

[0080] In the embodiment, after determining the weight corresponding to each evaluation index matrix, the comprehensive score of each question and answer robot can be calculated by using a weighted summation method.

[0081] The question and answer data processing method provided in the embodiment determines the comprehensive score of each question and answer robot corresponding to each intent data by respectively determining the weight corresponding to each evaluation index matrix and comprehensively determining the comprehensive score of each question and answer robot according to the weight information and the evaluation index matrix. Therefore, the accuracy of the comprehensive score of each question and answer robot can be improved.

[0082] Figure 6 The flowchart of the question and answer data processing method provided in Embodiment Four of the present disclosure is shown in FIG. 8, which is based on any of the above embodiments. Figure 6 As shown in FIG. 8, step 204 specifically includes:

[0083] Step 601: Obtain a question and answer data set of a question and answer robot whose comprehensive score under each intent data exceeds a preset threshold.

[0084] Step 602: Perform a preset data processing operation on the data in the question and answer data set to obtain the target question and answer data.

[0085] Step 603: Input the question data in the target question and answer data into the question and answer network model corresponding to the question and answer robot to obtain a predicted answer output by the question and answer network model.

[0086] Step 604: Calculate a loss value corresponding to the question and answer network model according to the predicted answer and the standard answer in the target question and answer data.

[0087] Step 605: Adjust the parameters of the question and answer network model according to the loss value.

[0088] The preset data processing operation includes any one or more of the following operations:

[0089] Clustering operation, entity recognition operation, keyword extraction operation, and semantic understanding operation.

[0090] In this embodiment, after obtaining the question and answer data set of the question and answer robot whose comprehensive score exceeds the preset threshold under the intention data, the question and answer data set can be subjected to a preset data processing operation to obtain target question and answer data. Specifically, the preset data processing operation includes any one or more of the following operations: clustering operation, entity recognition operation, keyword extraction operation, and semantic understanding operation. Thus, the target question and answer data can be used as support for the online question and answer robot in the subsequent process. Optionally, the target question and answer data can be used to update the question and answer network model corresponding to the question and answer robot.

[0091] Specifically, the question data in the target question and answer data can be used as input of the question and answer network model, and the question and answer network model can perform prediction processing on the question data to output a predicted answer corresponding to the question and answer data. According to the predicted answer and the standard answer in the target question and answer data, a loss value corresponding to the question and answer network model is calculated. Thus, the parameters of the question and answer network model can be adjusted according to the loss value.

[0092] Optionally, after adjustment, the step of using the question data in the target question and answer data as input of the question and answer network model can be performed again, the loss value of the question and answer network model in the adjustment parameters is calculated according to the predicted answer output by the question and answer network model after adjustment and the standard answer, and it is determined whether the loss value meets the preset convergence condition. When the convergence condition is met, the updated question and answer network model is obtained. Otherwise, the question and answer network model can be retrained using the target question and answer data until the question and answer network model converges.

[0093] The question and answer data processing method provided in this embodiment first performs a preset data processing operation on the question and answer data set, uses the target question and answer data processed by the data processing as support for the online question and answer robot, and thus the question and answer network model updated by using the target question and answer data can feed back a more high-quality reply according to the question information issued by the user, thereby improving the user experience.

[0094] Figure 7 The structure diagram of the question and answer data processing device provided in this embodiment is shown in Figure 7As shown, the apparatus comprises: an acquisition module 71, an aggregation module 72, a determination module 73, and an update module 74. The acquisition module 71 is configured to acquire a pre-stored online question and answer data set, wherein the online question and answer data set comprises intention data generated by question and answer robots corresponding to different question and answer fields. The aggregation module 72 is configured to perform an aggregation operation on the online question and answer data set according to the intention data and a response evaluation index corresponding to the intention data, to obtain evaluation index data of the question and answer robots corresponding to each intention data. The determination module 73 is configured to determine a comprehensive score of each question and answer robot corresponding to each intention data according to the evaluation index data. The update module 74 is configured to update a question and answer network model corresponding to each question and answer robot by using target question and answer data of the question and answer robot whose comprehensive score under each intention data exceeds a pre-set threshold.

[0095] The question and answer data processing apparatus provided in the embodiment performs an aggregation operation on intention data generated by question and answer robots corresponding to different question and answer fields, to obtain evaluation index data of the question and answer robots corresponding to each intention data, and determines a comprehensive score of each question and answer robot corresponding to each intention data according to the evaluation index data. Therefore, high-quality target question and answer data can be obtained in the online question and answer data set of different question and answer fields according to the comprehensive score, and the target question and answer data can be used to update a question and answer network model of the question and answer robot. Since the target question and answer data is a high-quality reply in different question and answer fields, the question and answer network model updated by using the target question and answer data can feed back a more high-quality reply according to question information issued by a user, thereby improving the user experience.

[0096] Further, on the basis of Embodiment Five, the evaluation index data is an evaluation index matrix;

[0097] The aggregation module is configured to: for each intention data, generate five-tuple data corresponding to each question and answer robot according to at least one response evaluation index corresponding to the intention data and an identifier of each question and answer robot; perform an aggregation operation on the five-tuple data, to obtain a to-be-processed matrix of the question and answer robots corresponding to each intention data; and perform a unified dimension processing on the to-be-processed matrix, to obtain an evaluation index matrix of the question and answer robots corresponding to each intention data.

[0098] Further, on the basis of Embodiment Five, the aggregation module is configured to: perform one or more operations of normalization, standardization, and rank operation on the to-be-processed matrix, to obtain the evaluation index matrix of the question and answer robots corresponding to each intention data.

[0099] Further, on the basis of any of the above embodiments, the determination module is configured to determine the weight information corresponding to each evaluation index matrix by using an unsupervised device. According to the evaluation index matrix and the weight information corresponding to the evaluation index matrix, the comprehensive score of each question and answer robot corresponding to each intention data is determined.

[0100] Further, on the basis of any of the above embodiments, the determination module is configured to determine the weight information corresponding to each evaluation index matrix by using a preset weight information determination algorithm. The preset weight information determination algorithm includes any one or more of the following algorithms: principal component algorithm, independent cost decomposition algorithm, or singular value decomposition algorithm, rough set algorithm.

[0101] Further, on the basis of any of the above embodiments, the determination module is configured to perform a weighted summation operation on the evaluation index matrix and the weight information corresponding to the evaluation index matrix. The result of the weighted summation is determined as the comprehensive score of each question and answer robot corresponding to each intention data.

[0102] Further, on the basis of any of the above embodiments, the update module is configured to obtain a question and answer data set of a question and answer robot whose comprehensive score under each intention data exceeds a preset threshold. A preset data processing operation is performed on the data in the question and answer data set to obtain the target question and answer data. The question data in the target question and answer data is input into the question and answer network model corresponding to the question and answer robot to obtain the predicted answer output by the question and answer network model; according to the predicted answer and the standard answer in the target question and answer data, the loss value corresponding to the question and answer network model is calculated; and the parameters of the question and answer network model are adjusted according to the loss value. The preset data processing operation includes any one or more of the following operations: clustering operation, entity recognition operation, keyword extraction operation, and semantic understanding operation.

[0103] Figure 8 is a structural schematic diagram of an electronic device provided for the sixth embodiment of the present disclosure. The device can be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a personal digital assistant, etc.

[0104] The apparatus 800 can include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input / output (I / O) interface 812, a sensor component 814, and a communication component 816.

[0105] The processing component 802 generally controls the overall operations of the device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 can include one or more processors 820 to execute instructions delivered from the memory 804 to complete all or part of the steps of the methods described above. In addition, the processing component 802 can include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 can include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0106] The memory 804 is configured to store various types of data to support the operations of the device 800. Examples of these data include instructions for any application or method operating on the device 800, contact data, phonebook data, messages, pictures, videos, and so on. The memory 804 can be implemented by any type of volatile or non-volatile storage devices or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0107] The power component 806 provides power to the various components of the device 800. The power component 806 can include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device 800.

[0108] The multimedia component 808 includes a screen providing an output interface between the device 800 and a user. In some embodiments, the screen can include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive an input signal from a user. The touch panel includes one or more touch sensors to sense a touch, a slide, and a gesture on the touch panel. The touch sensors can not only sense a boundary of a touching or sliding action, but also detect duration and pressure related to the touching or sliding action. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. The front camera and / or the rear camera can receive external multimedia data when the device 800 is in an operation mode, such as a shooting mode or a video mode. Each of the front and rear cameras can be a fixed optical lens system or have a focal length and optical zoom capability.

[0109] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC) that is configured to receive an external audio signal when the device 800 is in an operation mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0110] The I / O interface 812 provides an interface between the processing component 802 and peripheral interface modules, which can include a keypad, a click wheel, buttons, and so on. The buttons can include, but are not limited to, a home button, a volume button, a start button, and a lock button.

[0111] The sensor component 814 includes one or more sensors for providing status assessments of various aspects of the device 800. For example, the sensor component 814 can detect an open / closed position of the device 800, relative positioning of components, such as a display and a keypad of the device 800, a change of position of the device 800 or a component of the device 800, presence or absence of user contact with the device 800, changes in orientation or acceleration / deceleration

[0112] The communication component 816 is configured to facilitate wired or wireless communication between the device 800 and other devices. The device 800 can access a wireless network based on a corresponding communication standard, such as WiFi, 2G, or 3G, or a combination thereof. In an example embodiment, the communication component 816 receives broadcast signals or broadcast-related information from external broadcast management systems via a broadcast channel. In an example embodiment, the communication component 816 also includes a Near Field Communication (NFC) module to promote short-range communication. For example, the NFC module can be implemented based on Radio Frequency Identification (RFID) techniques, infrared data association (IrDA) techniques, ultra-wideband (UWB) techniques, Bluetooth (BT) techniques, and other techniques.

[0113] In exemplary embodiments, the apparatus 800 can be implemented using one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors or other electronic devices, to perform the above methods.

[0114] In exemplary embodiments, a non-transitory computer readable storage medium including instructions, such as the memory 804 including instructions, is also provided, which can be executed by the processor 820 of the apparatus 800 to complete the above methods. For example, the non-transitory computer readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disc, and an optical data storage device, etc.

[0115] Yet another embodiment of the present disclosure further provides a computer readable storage medium, in which computer execution instructions are stored, which are executed by a processor to implement the question and answer data processing method according to any one of the above embodiments.

[0116] Yet another embodiment of the present disclosure further provides a computer program product, which includes a computer program, which, when executed by a processor, implements the question and answer data processing method according to any one of the above embodiments.

[0117] Other embodiments of the present disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the features of the present disclosure disclosed herein. It is intended that the present disclosure cover any and all variations of the present disclosure including those variations which are now known or which are now considered to be obvious. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0118] It should be understood that the present disclosure is not limited to the precise structures herein described and illustrated in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the claims which follow.

Claims

1. A question-and-answer data processing method, characterized in that, include: Obtain a pre-stored online question-and-answer dataset, which includes intent data generated by question-and-answer robots corresponding to different question-and-answer domains; For each intent data, based on the multiple response evaluation metrics corresponding to the intent data and the identifiers of each question-answering robot, a quintuple data corresponding to each question-answering robot is generated; the quintuple data consists of intent code, robot code, conversation code, metric code, and specific metric value. The five-tuple data are aggregated to obtain the processing matrix of the question-answering robot corresponding to each intent data. The matrix to be processed is subjected to unified dimensional processing to obtain the evaluation index matrix of the question answering robot corresponding to each intent data. The response evaluation index includes: the rate of user request to transfer to human agent after response, response satisfaction, response resolution rate, consultation conversion rate, and consultation churn rate. Based on the evaluation index matrix, determine the comprehensive score of each question-answering robot corresponding to each intent data; The question-answering network model corresponding to each question-answering robot is updated by using the target question-answering data of the question-answering robots whose comprehensive scores under each intent data exceed a preset threshold.

2. The method according to claim 1, characterized in that, The step of performing unified dimensional processing on the matrix to be processed to obtain the evaluation index matrix of the question-answering robot corresponding to each intent data includes: The matrix to be processed is subjected to one or more of the following operations: normalization, standardization, and rank-finding, to obtain the evaluation index matrix of the question-answering robot corresponding to each intent data.

3. The method according to claim 1, characterized in that, The step of determining the comprehensive score of each question-answering robot corresponding to each intent data based on the evaluation index matrix includes: Unsupervised learning methods are used to determine the weight information corresponding to each evaluation index matrix; Based on the evaluation index matrix and the weight information corresponding to the evaluation index matrix, the comprehensive score of each question-answering robot corresponding to each intent data is determined.

4. The method according to claim 3, characterized in that, The method of determining the weight information corresponding to each evaluation index matrix using unsupervised learning includes: The weight information corresponding to each evaluation index matrix is ​​determined by a preset weight information determination algorithm; The preset weight information determination algorithm includes any one or more of the following algorithms: Principal Component Algorithm, Independent Cost Decomposition Algorithm, Singular Value Decomposition Algorithm, Rough Set Algorithm.

5. The method according to claim 3, characterized in that, The step of determining the comprehensive score of each question-answering robot corresponding to each intent data based on the evaluation index matrix and the weight information corresponding to the evaluation index matrix includes: The evaluation index matrix and the corresponding weight information of the evaluation index matrix are weighted and summed. The weighted summation result is determined as the comprehensive score of each question-answering robot corresponding to each intent data.

6. The method according to any one of claims 1-5, characterized in that, The step of updating the question-answering network model corresponding to each question-answering robot by using the target question-answering data of the question-answering robots whose comprehensive scores under each intent data exceed a preset threshold includes: Acquire the question-and-answer data set of the question-and-answer robot whose comprehensive score under each intent data exceeds a preset threshold; Perform preset data processing operations on the data in the question-and-answer dataset to obtain the target question-and-answer data; The question data in the target question-answering data is input into the question-answering network model corresponding to the question-answering robot to obtain the predicted answer output by the question-answering network model; Calculate the loss value corresponding to the question-answering network model based on the predicted answer and the standard answer in the target question-answering data; The parameters of the question-answering network model are adjusted based on the loss value. The preset data processing operations include any one or more of the following: Clustering operations, entity recognition operations, keyword extraction operations, and semantic understanding operations.

7. A question-and-answer data processing device, characterized in that, include: The acquisition module is used to acquire a pre-stored online question-and-answer dataset, which includes intent data generated by question-and-answer robots corresponding to different question-and-answer domains; The aggregation module is used to perform aggregation operations on the online question-and-answer dataset based on the intent data and the response evaluation indicators corresponding to the intent data, so as to obtain the evaluation indicator data of the question-and-answer robot corresponding to each intent data. The determination module is used to determine the comprehensive score of each question-answering robot corresponding to each intent data based on the evaluation index data; The update module is used to update the question-answering network model corresponding to each question-answering robot by using the target question-answering data of the question-answering robot whose comprehensive score under each intent data exceeds a preset threshold. The evaluation index data is an evaluation index matrix; The aggregation module is used to: for each intent data, generate a quintuple of data corresponding to each question-answering robot based on at least one response evaluation index corresponding to the intent data and the identifier of each question-answering robot; the quintuple of data consists of intent code, robot code, conversation code, index code, and specific index value; The response evaluation indicators include: the rate at which users request to be transferred to human assistance after a response, response satisfaction, response resolution rate, consultation conversion rate, and consultation churn rate. The five-tuple data are aggregated to obtain the processing matrix of the question-answering robot corresponding to each intent data. The matrix to be processed is subjected to unified dimensional processing to obtain the evaluation index matrix of the question-answering robot corresponding to each intent data.

8. The apparatus according to claim 7, characterized in that, The aggregation module is used for: The matrix to be processed is subjected to one or more of the following operations: normalization, standardization, and rank-finding, to obtain the evaluation index matrix of the question-answering robot corresponding to each intent data.

9. The apparatus according to claim 7, characterized in that, The determining module is used for: An unsupervised method is used to determine the weight information corresponding to each evaluation index matrix; Based on the evaluation index matrix and the weight information corresponding to the evaluation index matrix, the comprehensive score of each question-answering robot corresponding to each intent data is determined.

10. The apparatus according to claim 9, characterized in that, The determining module is used for: The weight information corresponding to each evaluation index matrix is ​​determined by a preset weight information determination algorithm; The preset weight information determination algorithm includes any one or more of the following algorithms: Principal Component Algorithm, Independent Cost Decomposition Algorithm, Singular Value Decomposition Algorithm, Rough Set Algorithm.

11. The apparatus according to claim 9, characterized in that, The determining module is used for: The evaluation index matrix and the corresponding weight information of the evaluation index matrix are weighted and summed. The weighted summation result is determined as the comprehensive score of each question-answering robot corresponding to each intent data.

12. The apparatus according to any one of claims 7-11, characterized in that, The update module is used for: Acquire the question-and-answer data set of the question-and-answer robot whose comprehensive score under each intent data exceeds a preset threshold; Perform preset data processing operations on the data in the question-and-answer dataset to obtain the target question-and-answer data; The question data in the target question-answering data is input into the question-answering network model corresponding to the question-answering robot to obtain the predicted answer output by the question-answering network model; Calculate the loss value corresponding to the question-answering network model based on the predicted answer and the standard answer in the target question-answering data; The parameters of the question-answering network model are adjusted based on the loss value. The preset data processing operations include any one or more of the following: Clustering operations, entity recognition operations, keyword extraction operations, and semantic understanding operations.

13. An electronic device, comprising: Memory, processor; Memory; Memory used to store the processor's executable instructions; The processor is used to invoke program instructions in the memory to execute the question-and-answer data processing method as described in any one of claims 1-6.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the question-and-answer data processing method as described in any one of claims 1-6.

15. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the question-and-answer data processing method according to any one of claims 1-6.

Images