Method and apparatus for human-machine conversation, device, medium, and program product

Abstract

According to the embodiments of the present disclosure, provided are a method and apparatus for human-machine conversation, a device, a medium, and a program product. In the method, conversation records between a user and a conversation machine are acquired, the conversation records comprising input information of the user and a first response of the conversation machine to the input information; the conversation records are inputted into a trained target machine learning model to acquire an analysis result for the first response outputted by the target machine learning model; on the basis of the analysis result, the validity of the first response of the conversation machine is determined; and, on the basis of the validity of the first response, a request requirement for a manual platform is determined, the request requirement indicating whether a request for responding to the input information of the user is to be initiated to the manual platform.

Description

Method, device, equipment, medium and program product for human-machine conversation

[0001] The present application claims priority to the Chinese patent application No. 202410870246.8, filed on June 28, 2024, entitled "Method, device, equipment, medium and program product for human-machine conversation", the whole content of which is incorporated herein by reference. TECHNICAL FIELD

[0002] The example embodiments of the present disclosure generally relate to the field of computer, and in particular, to a method, device, equipment, computer readable storage medium and computer program product for human-machine conversation. BACKGROUND

[0003] Human-Machine Conversation refers to a technology that enables a machine to understand and use natural language to realize human-machine communication. Through human-machine conversation interaction, a user can query information, chat with a machine, and also obtain specific services. At present, more and more service providing platforms use human-machine conversation technology to realize customer service functions to meet the needs of users for information acquisition of specific services. SUMMARY

[0004] In a first aspect of the present disclosure, a method for human-machine conversation is provided. The method comprises: obtaining a conversation record of a user and a conversation machine, the conversation record comprising input information of the user and a first reply of the conversation machine to the input information; inputting the conversation record into a trained target machine learning model to obtain an analysis result of the first reply output by the target machine learning model; determining validity of the first reply of the conversation machine based on the analysis result; and determining a request demand for an artificial platform based on the validity of the first reply, the request demand indicating whether to initiate a request for the artificial platform to reply to the input information of the user.

[0005] In a second aspect of the present disclosure, a device for human-machine conversation is provided. The device comprises: a conversation record obtaining module configured to obtain a conversation record of a user and a conversation machine, the conversation record comprising input information of the user and a first reply of the conversation machine to the input information; a conversation record inputting module configured to input the conversation record into a trained target machine learning model to obtain an analysis result of the first reply output by the target machine learning model; a validity determining module configured to determine validity of the first reply of the conversation machine based on the analysis result; and a demand determining module configured to determine a request demand for an artificial platform based on the validity of the first reply, the request demand indicating whether to initiate a request for the artificial platform to reply to the input information of the user.

[0006] In a third aspect of the disclosure, an electronic device is provided. The device includes at least one processor; and at least one memory coupled to the at least one processor and storing instructions for execution by the at least one processor. The instructions, when executed by the at least one processor, cause the device to perform the method of the first aspect.

[0007] In a fourth aspect of the disclosure, a computer-readable storage medium is provided. The computer-readable storage medium has stored thereon a computer program, the computer program being executable by a processor to implement the method of the first aspect.

[0008] In a fifth aspect of the disclosure, a computer program product is provided. The computer program product is tangibly stored in a computer storage medium and includes computer- executable instructions that, when executed by a device, cause the device to perform the method of the first aspect.

[0009] It is to be understood that the details set forth herein are not intended to limit the key features or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0010] The above and other features, advantages and aspects of embodiments of the present disclosure will become more apparent by describing in detail some embodiments thereof with reference to the annexed drawings in which:

[0011] FIG. 1 shows a schematic diagram of an example environment in which embodiments of the present disclosure can be implemented;

[0012] FIG. 2 shows a flowchart of a process for human-machine conversation according to some embodiments of the present disclosure;

[0013] FIG. 3 shows a flowchart of a use case for a target machine learning model according to some embodiments of the present disclosure;

[0014] FIG. 4 shows a flowchart of a process for human-machine conversation according to some embodiments of the present disclosure;

[0015] FIG. 5 shows a schematic diagram of a processing workflow for human-machine conversation according to some embodiments of the present disclosure;

[0016] FIG. 6 shows a block diagram of an apparatus for human-machine conversation according to some embodiments of the present disclosure; and

[0017] FIG. 7 shows a block diagram of an electronic device capable of implementing one or more embodiments of the present disclosure. DETAILED DESCRIPTION

[0018] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While certain embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein, but rather, the embodiments are provided so that the present disclosure can be more thoroughly and completely understood. It should be understood that the drawings and embodiments of the present disclosure are only for illustrative purposes and are not intended to limit the scope of protection of the present disclosure.

[0019] In the description of embodiments of the present disclosure, the term "comprising" and its conjugations should be understood to encompass the meanings of "consisting of" and "consisting essentially of", i.e., "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions can also be included below.

[0020] In this document, unless explicitly stated otherwise, performing a step "in response to" an event means that the step can be performed immediately in response to the event, but can also include one or more intermediate steps.

[0021] It can be understood that the data involved in the technical solutions of the present disclosure (including but not limited to the data itself, the obtaining or use of the data) should comply with the requirements of the relevant laws and regulations and the relevant provisions.

[0022] As used herein, the term "model" can learn the association between the corresponding input and output from the training data, so that after the training is completed, the corresponding output can be generated for a given input. The generation of the model can be based on machine learning techniques. Deep learning is a machine learning algorithm that processes input and provides corresponding output by using multiple layers of processing units. In this document, "model" can also be referred to as "machine learning model", "machine learning network" or "network", which are used interchangeably in this document. A model can also include different types of processing units or networks.

[0023] It can be understood that before using the technical solutions disclosed in the embodiments of the present disclosure, the type of personal information involved in the present disclosure, the scope of use, the use scenario, etc. should be informed to the user and the authorization of the user should be obtained through appropriate means according to the relevant laws and regulations.

[0024] For example, in response to receiving a user's active request, a prompt message is sent to the user to explicitly prompt the user that the operation requested to be performed will require obtaining and using the user's personal information, so that the user can voluntarily choose whether to provide personal information to the software or hardware such as electronic devices, application programs, servers or storage media that perform the operation of the technical solutions of the present disclosure according to the prompt message.

[0025] As an optional but non-limiting implementation, in response to receiving the active request of the user, the manner of sending the prompt information to the user can be, for example, a pop-up window manner, in which the prompt information can be presented in the form of text. In addition, the pop-up window can also carry a selection control for the user to select "agree" or "disagree" to provide personal information to the electronic device.

[0026] It can be understood that the above notification and user authorization obtaining process is only illustrative and does not limit the implementation of the present disclosure, and other manners meeting the relevant laws and regulations can also be applied to the implementation of the present disclosure.

[0027] As used herein, the term "model" can learn the association between the corresponding input and output from the training data, so that after the training is completed, the corresponding output can be generated for a given input. The generation of the model can be based on machine learning techniques. Deep learning is a machine learning algorithm that uses multiple layers of processing units to process inputs and provide corresponding outputs. The neural network model is an example of a model based on deep learning. In this paper, "model" can also be referred to as "machine learning model", "learning model", "machine learning network" or "learning network", which are used interchangeably in this paper.

[0028] FIG. 1 shows a schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented. In the example environment 100, a terminal device 110 has a target application 120 installed therein. In some embodiments, the target application 120 can be an application capable of providing customer service functions. For example, the target application 120 can be an application for online shopping, or an application for providing financial services, etc. It should be understood that the target application 120 can also be any other appropriate application, at least including providing customer service functions.

[0029] In some embodiments, the user 140 can interact with the target application 120 via the terminal device 110 and / or an attached device of the terminal device 110. Further, the user 140 can input information in the interaction window of the customer service function of the target application 120, and then the target application 120 calls the dialogue machine 145 (also referred to as intelligent robot) to reply. In some embodiments, the artificial platform 150 can also be called by the target application 120 to reply through artificial customer service.

[0030] In some embodiments, the terminal device 110 communicates with the service-side device 130 to implement the provisioning of the service of the target application 120. In the environment 100, a target machine learning model 135 communicates with the service-side device 130. In some examples, the service-side device 130 can assist in implementing the provisioning of the service of the target application 120 by invoking the target machine learning model 135. The target machine learning model 135 can run on other devices / systems different from the service-side device 130 and the terminal device 110. Although only a single target machine learning model 135 is shown in FIG. 1, it can be understood that there can be more models depending on the specific application needs, and different models can be configured to determine different human-computer dialogue operations in different ways, or provide other functions in the target application 120, and the like.

[0031] In some embodiments, the target machine learning model 135 can have content generation capability. In some embodiments, the target machine learning model 135 can include a language model, such as a large language model. In some embodiments, the target machine learning model 135 can understand language text content and process various natural language tasks, such as question answering, text summarization, and the like. The target machine learning model 135 can be pre-trained from a large amount of data, so as to be able to understand semantic information of text modalities and other modalities. A suitable machine learning model can be selected according to actual needs. In the human-computer dialogue scenario, the target machine learning model 135 can be used to understand text input related to the human-computer dialogue of the user 140 and the dialogue machine 145, and determine whether the reply of the dialogue machine 145 can meet the needs of the user 140 from the text input.

[0032] The target machine learning model 135 can run locally on the terminal device 110 or the service-side device 130 or be deployed in a remote device, a cloud environment, and the like. In the case of local running, the terminal device 110 or the service-side device 130 can directly provide the input of the model to the target machine learning model 135 installed locally, and obtain the model output generated by the target machine learning model 135. In the case of remote running, the terminal device 110 or the service-side device 130 provides data to other devices through a communication connection between the devices. Other devices determine the model input based on the obtained data and provide the model input to the target machine learning model 135. After obtaining the model output of the target machine learning model 135, other electronic devices provide the model output through a communication connection between the terminal device 110 or the service-side device 130.

[0033] In some embodiments, the terminal device 110 can be any type of mobile terminal, fixed terminal, or portable terminal including a mobile handset, a tablet computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio broadcast receiver, an electronic book device, a game device, or any combination thereof, including accessories and peripherals of these devices, or any combination thereof. In some embodiments, the terminal device 110 can also support any type of interface to the user (such as "wearable" circuitry, etc.). The server device 130 can be various types of computing systems / servers capable of providing computing power, including but not limited to mainframes, edge computing nodes, computing devices in a cloud environment, etc.

[0034] It should be appreciated that the structure and function of the various elements in the environment 100 are described for illustrative purposes only and do not imply any limitation on the scope of the present disclosure.

[0035] In the field of traditional network customer service, when operated in a man-machine cooperation mode, a time sequence two-stage service mode of first intelligent robot (i.e., dialogue machine) and then manual is adopted. In the first stage, the intelligent robot matches according to an algorithm model (for example, Bert model), obtains a result from a trained knowledge set to replace manual customer service to answer questions. If the matching fails or the customer (i.e., user) triggers a rule of transferring to manual, the second stage needs to be transferred to manual service (this process is irreversible), and the intelligent robot gives answers to manual customer service as assistance.

[0036] In such a case, the increase of manual cost is caused. In the scene of customer service coping with multiple customer reception, the record of the intelligent robot and the customer needs to be read, and then the customer question that needs to be continuously answered is summarized to reply. This process also causes the decline of the efficiency of manual customer service.

[0037] Therefore, in the field of network customer service, due to the upper limit of the reception capacity of manual customer service, the time sequence two-stage service mode of first intelligent robot and then manual exists bottleneck of efficiency and experience.

[0038] According to an embodiment of the present disclosure, an improved solution for human-machine conversation is proposed. According to the solution of the embodiment of the present disclosure, a conversation record of a user and a conversation machine is acquired, the conversation record including input information of the user and a first reply of the conversation machine to the input information; the conversation record is input to a trained target machine learning model to acquire an analysis result of the first reply output by the target machine learning model; based on the analysis result, validity of the first reply of the conversation machine is determined; and based on the validity of the first reply, a request demand for an artificial platform is determined, the request demand indicating whether a request for replying to the input information of the user is to be initiated to the artificial platform. In this way, by this way, the artificial intervention can be reduced while ensuring the accurate reply to the user's question, and the artificial reply efficiency is improved, the service potential of the conversation machine and the artificial customer service is maximized, so that the reception efficiency of the user is improved.

[0039] Some example embodiments of the present disclosure will be described below with continuous reference to the drawings.

[0040] FIG. 2 shows a flowchart of a process 200 for human-machine conversation according to some embodiments of the present disclosure. For ease of discussion, reference will be made to the environment 100 of FIG. 1 to describe these embodiments. These embodiments can be implemented in the server device 130 of FIG. 1. In some other embodiments, these embodiments can also be implemented in the terminal device 110, or through the cooperation of the terminal device 110 and the server device 130. For the purpose of discussion, the following specific embodiments are taken as an example of being implemented in the server device 130.

[0041] At block 210, the server device 130 can acquire a conversation record of the user 140 and the conversation machine 145, the conversation record including input information of the user 140 and a first reply of the conversation machine 145 to the input information. In some embodiments, the user 140 can input information through an interactive window of a customer service function of the target application 120. The input information of the user 140 can include text information and voice information, etc. The input information of the user 140 may, for example, be a question of the user 140 about the service provided by the target application 120. The following specific embodiments take the example of the user 140 asking "I purchased a commodity using financial product A, but because the return logistics has not arrived, the refund has not been credited, but the financial product A will be due tomorrow, can I wait for the arrival of the logistics to automatically repay the loan without overpayment?" to describe the example process of the present disclosure.

[0042] In such a case, the target application 120 can invoke the dialogue machine 145 to reply to the question of the user 140. Assuming that the dialogue machine 145 matches the answer: “If your current bill exceeds the minimum repayment amount of financial product A on X day of X month, it is considered overdue, and you will be charged XX fees per day, which will also affect your personal credit. It is recommended that you repay on time”, which is the first reply of the dialogue machine 145 to the input information. Thus, the interaction window of the target application 120 presents the dialogue record of the user 140 and the dialogue machine 145, which can be obtained by the server device 130.

[0043] At block 220, the server device 130 can input the dialogue record into the trained target machine learning model 135 to obtain the analysis result of the first reply output by the target machine learning model 135. In some embodiments, when inputting the dialogue record into the target machine learning model 135, the server device 130 can perform content conversion on the dialogue record according to a predetermined dialogue template, and input the converted dialogue record into the target machine learning model 135. In some embodiments, the target machine learning model 135 can be a language model.

[0044] FIG. 3 shows a flowchart of a use example 300 of the target machine learning model 135 according to some embodiments of the present disclosure. The example 300 can be implemented in the environment 100 of FIG. 1. The example 300 can be shown as an embodiment of part of the process 200. In the example 300, for the input information 310 of the user 140 and the first reply 320 of the dialogue machine 145, the server device 130 needs to perform content conversion (330) according to a predetermined dialogue template to convert the content that is easier for the target machine learning model 135 to understand. It should be understood that the predetermined dialogue template supports dynamic configuration. Exemplarily, the content conversion can be implemented by using prompt word engineering.

[0045] For example, the above example dialogue record can be converted to “If you are a customer service, there is a consultation dialogue with a customer as follows: Customer: I purchased goods using financial product A, but because the return logistics has not arrived, the refund has not been credited, but financial product A will reach the repayment period tomorrow, can this situation wait for the arrival of the logistics to automatically repay without overdue? Machine customer service: If your current bill exceeds the minimum repayment amount of financial product A on X day of X month, it is considered overdue, and you will be charged XX fees per day, which will also affect your personal credit. It is recommended that you repay on time”, and then the server device 130 can input the content into the target machine learning model 135.

[0046] Next, the target machine learning model 135 can determine (340) the intent of the input information 319 of the user 140 and the first reply 320 of the dialogue machine 145 from the above content. For example, the analysis result of the target machine learning model 135 is: “the reply of the machine customer service does not answer the question of the customer, the answer only emphasizes the consequences of overdue repayment, and does not answer the question of the customer about whether the automatic repayment after waiting for the arrival of the logistics can be overdue”. Further, if the target machine learning model 135 outputs the analysis result, the following flow (350) is continued. If the target machine learning model 135 does not output the analysis result, the artificial intervention flow (360) is directly entered.

[0047] In some embodiments, the target machine learning model 135 can be configured to split the input information into a plurality of sub-questions, and generate the analysis result of the first reply based on the splitting result. Specifically, the target machine learning model 135 can analyze that the question of the input information of the user 140 contains several sub-questions, and perform sub-question splitting. Then, the target machine learning model 135 analyzes the first reply of the dialogue machine 145 according to the split sub-questions, to determine which answers in the first reply of the dialogue machine 145 are correct and which answers are incorrect. That is, to determine which sub-questions in the question of the user 140 are answered by the first reply of the dialogue machine 145.

[0048] In some embodiments, the target machine learning model 135 can also output the total number of sub-questions in the input information of the user 140, and the number of sub-questions that have been accurately answered. In some embodiments, the target machine learning model 135 can also output the proportion of sub-questions that have been accurately answered according to the total number and the number of sub-questions that have been accurately answered.

[0049] Therefore, by analyzing the dialogue record by means of the target machine learning model 135, the customer service can be converted from the customer dimension to the question dimension, which helps to improve the analysis efficiency of the input information of the user 140.

[0050] With reference to FIG. 2, at block 230, the service end device 130 can determine the effectiveness of the first reply of the dialogue machine 145 based on the analysis result. Specifically, the service end device 130 can determine the effectiveness of the first reply of the dialogue machine 145 according to the total number of sub-questions in the input information of the user 140 and the number of sub-questions that have been accurately answered.

[0051] In some embodiments, in determining the validity of the first reply of the dialogue machine 145 based on the analysis result, the server-side device 130 can determine, based on the analysis result, whether the first reply of the dialogue machine 145 is totally valid or totally invalid. Specifically, if the number of the sub-questions that have been accurately answered is the same as the total number of the sub-questions in the input information of the user 140, it is determined that the first reply of the dialogue machine 145 is totally valid. If the number of the sub-questions that have been accurately answered is zero, it is determined that the first reply of the dialogue machine 145 is totally invalid.

[0052] In some embodiments, in determining the validity of the first reply of the dialogue machine 145 based on the analysis result, the server-side device 130 can also determine, based on the analysis result, whether the first reply of the dialogue machine 145 is partially valid. Accordingly, if the number of the sub-questions that have been accurately answered is greater than zero and less than the total number of the sub-questions in the input information of the user 140, it is determined that the first reply of the dialogue machine 145 is partially valid.

[0053] In block 240, based on the validity of the first reply, the server-side device 130 can determine a request demand for the human platform 150, the request demand indicating whether a request for the human platform 150 to reply to the input information of the user 140 is to be initiated. In some embodiments, in determining the request demand for the human platform 150 based on the validity of the first reply, if it is determined that the first reply is totally valid, the server-side device 130 can determine that the human platform 150 does not need to be requested to reply to the input information of the user 140, and determine that the dialogue machine 145 is to reply to the next input information of the user 140. That is, if it is determined that the first reply is totally valid, it indicates that the dialogue machine 145 has a high accuracy of answering, and thus the human platform 150 can not be requested to reply to the input information of the user 140, and the dialogue machine 145 can be allowed to continue to reply to the next input information of the user 140.

[0054] In another embodiment, the server-side device 130 can compare the ratio of the number of the sub-questions that have been accurately answered to a preset proportion threshold, and if the ratio is greater than the preset proportion threshold, the human platform 150 does not need to be requested to reply to the input information of the user 140, and the dialogue machine 145 can be allowed to continue to reply to the next input information of the user 140.

[0055] In some embodiments, in determining the request demand for the human platform 150 based on the validity of the first reply, if it is determined that the first reply is totally invalid, the server-side device 130 can determine that the human platform 150 is to be requested to reply to the input information of the user 140. That is, if it is determined that the first reply is totally invalid, it indicates that the dialogue machine 145 has a low accuracy of answering, and thus the human platform 150 is requested to inform a human customer service to reply to the input information of the user 140.

[0056] Thus, by determining whether to request the human platform 150 to reply to the input information of the user 140 based on the effectiveness of the first reply, the number of times the human platform 150 is requested can be reduced, thereby reducing the human intervention rate to save manpower.

[0057] In some embodiments, when determining the request demand for the human platform 150 based on the effectiveness of the first reply, the server device 130 can determine to request the human platform 150 to reply to the input information of the user 140 in response to determining that the first reply is partially effective. In some embodiments, if it is determined that the first reply is partially effective, the server device 130 can send at least one sub-question corresponding to the ineffective part of the first reply to the human platform 150 and request the human platform 150 to reply to the at least one sub-question.

[0058] In order to achieve both improving the accuracy of the answer to the input information of the user 140 and reducing the human intervention rate, the sub-questions corresponding to the ineffective part of the first reply can be sent to the human platform 150 and then the human platform 150 is requested to reply to only these sub-questions. Thus, compared with the human platform 150 replying to all the input information of the user 140, this embodiment effectively saves the time of the human platform 150 and improves the efficiency of the human platform 150, while ensuring the accuracy of the answer.

[0059] In some embodiments, the target machine learning model 135 is further configured to analyze the dialogue record and output a dialogue summary. In some embodiments, if it is determined that the request demand indicates that a request for the human platform 150 to reply to the input information of the user 140 is to be initiated, the server device 130 can obtain the dialogue summary output by the target machine learning model 135 and request the human platform 150 to reply to the input information of the user 140 based on the dialogue summary by sending the dialogue summary to the human platform 150. Illustratively, before the server device 130 requests the target machine learning model 135 to make a dialogue summary, the dialogue record needs to be content-converted according to another predetermined dialogue template.

[0060] For example, the following converted content can be input to the target machine learning model 135: “You are now a customer service, you need to transfer the customer to another customer service, so you need to summarize what happened just now, please describe it in no more than 40 words, which can make another customer service quickly understand the problem”. The target machine learning model 135 can give a dialogue summary: “The return logistics caused the refund not to be credited, the white ticket repayment period will arrive, the customer asks whether to wait for the arrival of the logistics to automatically repay. Please help to solve the credit problem.” Then, the server device 130 can receive the dialogue summary and send it to the artificial platform 150, so that the artificial customer service replies to the input information of the user 140 directly based on the dialogue summary. In this way, by providing the dialogue summary to the artificial platform 150, the reading time of the artificial customer service for the dialogue record can be reduced, and the upper limit of the artificial customer service reception capacity can be effectively improved.

[0061] In this scheme, the powerful context understanding, question splitting and summarizing capabilities of the target machine learning model 135 are used to improve the traditional two-stage service mode into a mixed mode of dialogue machine 145 full-process response and artificial response type response. In this way, while ensuring accurate response, artificial intervention can be reduced, and artificial reply efficiency can be improved, the service potential of the dialogue machine 145 and the artificial customer service is maximized, so as to realize the improvement of the reception efficiency of the user.

[0062] FIG. 4 shows a flowchart of a process 400 for human-machine dialogue according to some embodiments of the present disclosure. The process 400 can be implemented in the environment 100 of FIG. 1. The process 400 can be shown as a specific embodiment of the steps of the process 200. In the process 400, first, the question of the user 140 (i.e., the input information) and the first reply of the dialogue machine 145 are input to the target machine learning model 135 (405). Then, the target machine learning model 135 identifies the question intention of the user 140 and identifies the intention of the reply of the dialogue machine 145 (410). Next, the target machine learning model 135 can analyze whether the first reply of the dialogue machine 145 solves the intention of the user 140 (420). After the analysis result is obtained by the server device 130, if it is determined that the first reply of the dialogue machine 145 solves the intention of the user 140, the server device 130 directly outputs the reply of the dialogue machine 145 to the IM console (such as an interactive window) for the user (430). Otherwise, the server device 130 makes the target machine learning model 135 read the dialogue record of the dialogue machine 145 and the user 140, gives the user intention, which intentions are answered by the dialogue machine 145, and which intentions are not answered (440), and then outputs to the IM console (such as the artificial platform 150) for the artificial customer service (450). Next, the artificial customer service answers the question through the artificial platform 150 (460). Then, the next input information of the user 140 is replied by the dialogue machine 145 (470).

[0063] FIG. 5 shows a schematic diagram of an architecture 500 for human-machine conversation, according to some embodiments of the present disclosure. The architecture 500 can be implemented in the environment 100 of FIG. 1. The architecture 500 can be used in the processes 200 and 400. In the architecture 500, after the dialogue machine 145 replies to the user question (510) and the human intervention reply (460) is made in combination with FIG. 4, the user question in the subsequent process is still replied by the dialogue machine 145 first (470). The target machine learning model 135 mainly performs dialogue record analysis (420) and dialogue summary (520), and the output result of the target machine learning model 135 is obtained by the server device 130 to determine whether the human intervention reply (460) is needed. Then, the target machine learning model 135 continuously supervises the dialogue record (530), and the cycle is repeated.

[0064] Through the above process, the traditional two-stage service mode can be improved into a hybrid mode of full-process reply of the dialogue machine 145 and human response reply, and the service function of the customer service is converted from the customer dimension to the problem dimension, which avoids the scenario of the dialogue machine 145 answering the wrong question and greatly improves the service efficiency.

[0065] In some other embodiments, the target machine learning model 135 can feed back the result that some answers of the dialogue machine 145 do not conform to the user's intention to the model algorithm of the dialogue machine 145 to improve the reply accuracy of the dialogue machine 145. In addition, the target machine learning model 135 can also fine-tune the training or optimize the prompt words in the business field to improve the analysis ability and the conversation summary accuracy of the target machine learning model 135.

[0066] FIG. 6 shows a schematic structural block diagram of an apparatus 600 for human-machine conversation, according to some embodiments of the present disclosure. The apparatus 600 can be implemented as or included in the server device 130. Various modules / components in the apparatus 600 can be implemented by hardware, software, firmware, or any combination thereof.

[0067] As shown, the apparatus 600 includes a dialogue record obtaining module 610 configured to obtain a dialogue record of a user and a dialogue machine, the dialogue record including input information of the user and a first reply of the dialogue machine to the input information.

[0068] The apparatus 600 further includes a dialogue record inputting module 620 configured to input the dialogue record to a trained target machine learning model to obtain an analysis result of the first reply output by the target machine learning model.

[0069] The apparatus 600 further comprises an effectiveness determination module 630 configured to determine, based on the analysis result, an effectiveness of the first reply of the dialog machine.

[0070] The apparatus 600 further comprises a demand determination module 640 configured to determine, based on the effectiveness of the first reply, a request demand for a human platform, the request demand indicating whether a request for the human platform to reply to the input information of the user is to be initiated.

[0071] In some embodiments, the effectiveness determination module 630 is further configured to determine, based on the analysis result, that the first reply of the dialog machine is either entirely valid or entirely invalid.

[0072] In some embodiments, the demand determination module 640 is further configured to determine, in response to determining that the first reply is entirely valid, that the human platform is not requested to reply to the input information of the user; and determine that a next input information of the user is replied by the dialog machine.

[0073] In some embodiments, the demand determination module 640 is further configured to determine, in response to determining that the first reply is entirely invalid, that the human platform is requested to reply to the input information of the user.

[0074] In some embodiments, the effectiveness determination module 630 is further configured to determine, based on the analysis result, whether the first reply of the dialog machine is partially valid; and wherein the demand determination module 640 is further configured to determine, in response to determining that the first reply is partially valid, that the human platform is requested to reply to the input information of the user.

[0075] In some embodiments, the target machine learning model is configured to split the input information into a plurality of sub-questions, and generate the analysis result of the first reply based on a result of the splitting, and the apparatus 600 further comprises a partially valid module configured to, in response to determining that the first reply is partially valid, send at least one sub-question corresponding to an invalid part of the first reply to the human platform; and request the human platform to reply to the at least one sub-question.

[0076] In some embodiments, the target machine learning model is further configured to analyze the conversation record and output a conversation summary, wherein the apparatus 600 further comprises a conversation summary module configured to, in response to determining that the request requirement indicates a request to initiate a reply to the input information of the user to the artificial platform, acquire the conversation summary output by the target machine learning model; and request the artificial platform to reply to the input information of the user based on the conversation summary by sending the conversation summary to the artificial platform.

[0077] In some embodiments, the conversation record input module 620 is further configured to perform content conversion on the conversation record according to a predetermined conversation template; and input the converted conversation record to the target machine learning model.

[0078] The units and / or modules included in the apparatus 600 can be implemented in various ways, including software, hardware, firmware, or any combination thereof. In some embodiments, one or more units and / or modules can be implemented using software and / or firmware, e.g., machine-executable instructions stored on a storage medium. In addition or as an alternative, part or all of the units and / or modules in the apparatus 600 can be implemented by one or more hardware logic components. As an example and not by way of limitation, example types of hardware logic components that can be used include field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SOCs), complex programmable logic devices (CPLDs), etc.

[0079] FIG. 7 shows a block diagram of an electronic device 700 in which one or more embodiments of the disclosure can be implemented. It should be understood that the electronic device 700 illustrated in FIG. 7 is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein. The electronic device 700 illustrated in FIG. 7 can be used to implement the server device 130 of FIG. 1 or the apparatus 600 of FIG. 6.

[0080] As shown in FIG. 7, the electronic device 700 is in the form of a general electronic device. The components of the electronic device 700 can include, but are not limited to, one or more processors or processing units 710, a memory 720, a storage device 730, one or more communication units 740, one or more input devices 750, and one or more output devices 760. The processing unit 710 can be a real or virtual processor and is capable of performing various processing according to programs stored in the memory 720. In a multi-processor system, multiple processing units perform computer-executable instructions in parallel to improve the parallel processing capability of the electronic device 700.

[0081] The electronic device 700 typically includes a plurality of computer storage media. Such media can be volatile and / or nonvolatile, removable and / or non-removable, and can be implemented in any method or technology for storage of information and / or data. For example, memory 720 can be implemented using a volatile and / or nonvolatile memory such as random access memory (RAM), cache memory, and / or flash memory. Storage 730 can be implemented using a non-transitory computer-readable medium, such as a solid state memory or a magnetic hard disk, to name a few. The computer-readable medium can be used to store data and / or instructions for use by and / or in connection with one or more computer programs and / or devices. For example, the computer-readable medium can be used for storing training data for training, and / or instructions for implementing the various embodiments of the present disclosure.

[0082] The electronic device 700 can further include additional removable / non-removable, volatile / non-volatile storage media. Although not shown in FIG. 7, a disk drive and / or a CD drive can be provided for reading from or writing to a removable, non-removable, volatile, and / or non-volatile media such as a CD, a DVD, and / or a floppy disk. In such instances, each drive can be connected to the bus by one or more data media interfaces. The memory 720 can include a computer program product 725 having one or more program modules configured to carry out the various methods or actions of the various embodiments of the present disclosure.

[0083] The communication unit 740 enables communications with other electronic devices over a communication medium. Additionally, the functionality of the components of the electronic device 700 can be implemented in a single computing cluster or a plurality of computer machines that are capable of communicating with one another over a communication connection. As such, the electronic device 700 can operate in a networked environment using logical connections to one or more other servers, network personal computers (PCs), or another network nodes.

[0084] The input device 750 can be one or more input devices such as a mouse, a keyboard, a trackball, etc. The output device 760 can be one or more output devices such as a display, a speaker, a printer, etc. The electronic device 700 can further communicate with one or more external devices (not shown) such as a storage device, a display device, etc. through the communication unit 740, with one or more devices that enable a user to interact with the electronic device 700, or with any devices (e.g., a network card, a modem, etc.) that enables the electronic device 700 to communicate in a network environment. Such communication can be enabled by an input / output (I / O) interface (not shown).

[0085] According to an example implementation of the present disclosure, a computer readable storage medium is provided having computer executable instructions stored thereon, where the computer executable instructions are executed by a processor to implement the method described above. According to an example implementation of the present disclosure, a computer program product is also provided that is tangibly stored on a non-transitory computer readable medium and includes computer executable instructions, where the computer executable instructions are executed by a processor to implement the method described above.

[0086] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0087] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0088] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0089] The computer program product of the present disclosure can have a signal including said computer program. This signal can be electronic, electromagnetic, optical, or any other suitable type of signal. Such a signal can be provided through a communication connection, such as electrical wiring, optical fiber, wireless interface, etc. Examples of computer program products include computer program implemented on a personal computer, server, or other networked device. A non-transitory computer readable medium, such as a floppy disk, CD-ROM, DVD-ROM, Blu-ray Disc, hard disk drive, or any other suitable non-transitory computer readable medium can store the computer program product. The computer program product can be provided in a computer program package, such as a floppy disk, CD-ROM, DVD-ROM, Blu-ray Disc, hard disk drive, or any other suitable computer program package. The computer program product can also be provided via a computer network, such as the Internet, a local area network, a wide area network, a wireless network, or any other suitable computer network.

[0090] Having described several implementations of the present disclosure, it is to be appreciated various alterations, modifications, and improvements will readily occur to those skilled in the art. Such alterations, modifications, and improvements are intended to be part of this disclosure. Accordingly, the foregoing description is by way of example only and is not intended to be limiting. The implementation described hereinabove is intended to be merely exemplary and those skilled in the art will recognize that equivalents, both to the implementation described and to the general concepts described herein, can be employed without departing from the scope of the disclosure.

Claims

1. A method for human-computer dialogue, comprising: Acquire the dialogue record between the user and the dialogue machine, the dialogue record including the user's input information and the dialogue machine's first response to the input information; The dialogue record is input into a trained target machine learning model to obtain the analysis results of the first response output by the target machine learning model; Based on the analysis results, the validity of the first response from the dialogue machine is determined; as well as Based on the validity of the first response, a request requirement for the human platform is determined, the request requirement indicating whether to initiate a request to the human platform to respond to the user's input information.

2. The method according to claim 1, wherein determining the validity of the first response of the dialogue machine based on the analysis results includes: Based on the analysis results, it is determined whether the first response from the dialogue machine is entirely valid or entirely invalid.

3. The method according to claim 2, wherein determining the request requirement for the human platform based on the validity of the first response includes: In response to determining that all the first responses are valid, it is determined that there is no need to request the human platform to respond to the user's input information; as well as It is determined that the dialogue machine will respond to the user's next input information.

4. The method according to claim 2, wherein determining the request requirement for the human platform based on the validity of the first response further includes: In response to determining that all of the first responses are invalid, the system requests the human platform to respond to the user's input information.

5. The method according to claim 1, wherein determining the validity of the first response of the dialog machine based on the analysis results further includes: Based on the analysis results, it is determined whether the first response of the dialogue machine is partially valid; Based on the validity of the first response, determining the request requirement for the human platform also includes: In response to determining that the first response portion is valid, the system requests the human platform to respond to the user's input information.

6. The method of claim 5, wherein the target machine learning model is configured to split the input information into a plurality of sub-questions, and to generate the analysis result of the first response based on the result of the splitting, and the method further comprises: In response to determining that the first response portion is valid, at least one sub-question corresponding to the invalid portion of the first response is sent to the artificial intelligence platform; as well as The human platform is requested to respond to at least one of the sub-questions.

7. The method of claim 1, wherein the target machine learning model is further configured to analyze the dialogue record and output a dialogue summary. The method further includes: In response to determining that the request requirement indicates that a request should be initiated to the human platform to respond to the user's input information, the dialogue summary output by the target machine learning model is obtained; as well as By sending the dialogue summary to the human platform, the human platform is requested to respond to the user's input information based on the dialogue summary.

8. The method of claim 1, wherein inputting the dialogue record into the target machine learning model comprises: The dialogue record is converted according to a predetermined dialogue template; as well as The converted dialogue record is then input into the target machine learning model.

9. A device for human-computer interaction, comprising: The dialogue record acquisition module is configured to acquire the dialogue record between the user and the dialogue machine, the dialogue record including the user's input information and the dialogue machine's first response to the input information; The dialogue record input module is configured to input the dialogue record into a trained target machine learning model to obtain the analysis results of the first response output by the target machine learning model; The validity determination module is configured to determine the validity of the first response of the dialog machine based on the analysis results; as well as The demand determination module is configured to determine a request demand to the human platform based on the validity of the first response, the request demand indicating whether to initiate a request to the human platform to respond to the user's input information.

10. An electronic device, comprising: At least one processor; as well as At least one memory coupled to the at least one processor and storing instructions for execution by the at least one processor, the instructions causing the electronic device to perform the method according to any one of claims 1 to 8 when executed by the at least one processor.

11. A computer-readable storage medium having a computer program stored thereon, the computer program being executable by a processor to implement the method according to any one of claims 1 to 8.

12. A computer program product tangibly stored in a computer storage medium and comprising computer-executable instructions that, when executed by a device, cause the device to perform the method according to any one of claims 1 to 8.

Images