Question and answer path determination method and device, equipment, storage medium and program product

Abstract

The invention relates to the technical field of computers, and discloses a question and answer path determination method and device, equipment, a storage medium and a program product.The method comprises the steps that a new session message initiated by a target object is received, a dialogue node tree corresponding to the new session message is determined, and the dialogue node tree comprises a plurality of dialogue levels, each dialogue hierarchy comprises at least one dialogue node; triggering acquisition of historical session information based on the new session message, and constructing dialogue prompt information by using the acquired historical session information; determining a target dialogue node of each dialogue level pointed by the new dialogue message based on the dialogue prompt information; and constructing a question and answer path of the new session message according to the target session node. By implementing the technical scheme of the invention, automatic selection of the target dialogue node is realized, the selection accuracy of the target dialogue node is improved, the construction accuracy of the question and answer path is ensured, and then the reply accuracy of the new dialogue message is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a method, device, equipment, storage medium and program product for determining a question-answer path. Background Art

[0002] With the development of computer technology, the application of intelligent conversational robots has received more and more attention, especially in the field of e-commerce. The application of intelligent customer service robots has greatly saved human resources and improved service efficiency. At present, conversational robots mainly use standard operation procedures (SOPs) and combine user status and questions to determine which branch or node to go to for each question. How to accurately and quickly determine the question-answering path related to the question has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] In view of this, the present disclosure provides a method, apparatus, device, storage medium and program product for determining a question-answer path to solve the problem of generating a question-answer path.

[0004] In a first aspect, the present disclosure provides a method for determining a question-and-answer path, comprising: receiving a new conversation message initiated by a target object, determining a dialog node tree corresponding to the new conversation message, the dialog node tree comprising multiple dialog levels, each dialog level comprising at least one dialog node; triggering acquisition of historical conversation information based on the new conversation message, and constructing conversation prompt information using the acquired historical conversation information; determining a target conversation node of each dialog level pointed to by the new conversation message based on the conversation prompt information; and constructing a question-and-answer path of the new conversation message according to the target conversation node.

[0005] In a second aspect, the present disclosure provides a device for determining a question-and-answer path, comprising: an information receiving module, used to receive a new conversation message initiated by a target object, and determine a dialog node tree corresponding to the new conversation message, wherein the dialog node tree includes multiple dialog levels, and each dialog level includes at least one dialog node; a prompt information construction module, used to trigger the acquisition of historical conversation information based on the new conversation message, and construct conversation prompt information using the acquired historical conversation information; a target determination module, used to determine a target conversation node of each dialog level pointed to by the new conversation message based on the conversation prompt information; and a question-and-answer path determination module, used to construct a question-and-answer path of the new conversation message according to the target conversation node.

[0006] In a third aspect, the present disclosure provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the method for determining the question and answer path of the above-mentioned first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0007] In a fourth aspect, the present disclosure provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for determining a question-and-answer path of the first aspect or any corresponding embodiment thereof.

[0008] In a fifth aspect, the present disclosure provides a computer program product, including computer instructions, which are used to enable a computer to execute the method for determining a question-and-answer path of the above-mentioned first aspect or any corresponding embodiment thereof.

[0009] The method, device, equipment, storage medium and program product for determining the question-answer path provided by the present disclosure trigger the acquisition of historical conversation information when receiving new conversation information of the target object, and use the historical conversation information to construct dialogue prompt information, thereby combining the dialogue prompt information to determine the target dialogue node pointed to by the new conversation message in each dialogue level of the dialogue node tree, thereby realizing the automatic selection of the target dialogue node. At the same time, by constructing the dialogue prompt information using the historical conversation information related to the new conversation message, the target dialogue node that the new conversation message is to pass through in the SOP process can be accurately and quickly located, thereby improving the accuracy of selecting the target dialogue node. Subsequently, the question-answer path is constructed using the target dialogue node, thereby ensuring the accuracy of the construction of the question-answer path, thereby improving the accuracy of the reply to the new conversation message. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the specific embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0011] Figure 1 is a flowchart of a method for determining a question-answer path according to an embodiment of the present disclosure;

[0012] Figure 2 is a schematic diagram of a dialog node tree according to an embodiment of the present disclosure;

[0013] Figure 3 is a flowchart of another method for determining a question-answer path according to an embodiment of the present disclosure;

[0014] Figure 4 is a specific schematic diagram of generating a question-answering path according to an embodiment of the present disclosure;

[0015] Figure 5 is a schematic diagram of the dialog node pruning effect according to an embodiment of the present disclosure;

[0016] Figure 6 is a structural block diagram of a device for determining a question-answer path according to an embodiment of the present disclosure;

[0017] Figure 7 It is a schematic diagram of the hardware structure of the computer device of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.

[0019] It is understandable that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, scope of use, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0020] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly prompt the user that the operation requested to be performed will require obtaining and using the user's personal information. Thus, the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, application, server, or storage medium that performs the operation of the technical solution of the present disclosure according to the prompt message.

[0021] As an optional but non-limiting implementation, in response to receiving an active request from the user, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0022] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that meet the relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0023] It is understandable that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and relevant provisions.

[0024] In the field of e-commerce, the application of conversational robots has become increasingly mainstream, thereby reducing labor costs and improving e-commerce service efficiency. However, conversational robots mainly use standard operating procedures (SOPs) and combine user status and questions to determine which branch or node to go to for each question. Wrong or inappropriate branch / node selection will lead to response deviation or even wrong response, which seriously affects the user's conversation experience.

[0025] Based on this, the disclosed technical solution uses a pre-trained conversation model to automatically select dialogue nodes, achieving the effect of pre-filling SOP in the conversation. When using the conversation model to decide the dialogue node of each SOP, in order to avoid the problem of taking too long, by combining the existing historical conversation data, the target dialogue node is determined, and the dialogue node pruning operation is performed using the probability of selecting each dialogue node, which realizes the rapid determination of the SOP question and answer path and the generation of correct reply information.

[0026] According to an embodiment of the present disclosure, an embodiment of a method for determining a question-and-answer path is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0027] In this embodiment, a method for determining a question-answer path is provided, which can be used in computer devices such as computers, mobile phones, tablet computers, etc. Figure 1 is a flow chart of a method for determining a question-answer path according to an embodiment of the present disclosure, such as Figure 1 As shown, the process includes the following steps:

[0028] Step S101: receiving a new conversation message initiated by a target object, and determining a dialog node tree corresponding to the new conversation message. The dialog node tree includes multiple dialog levels, and each dialog level includes at least one dialog node.

[0029] The target object is the user who initiates the conversation information through the interactive interface. The dialogue node tree is a tree structure composed of the dialogue nodes through which the dialogue reply information is generated. Figure 2 As shown, the dialog node tree includes multiple dialog levels, and each dialog level includes one or more dialog nodes.

[0030] Different dialog node trees can be set for different business scenarios. The dialog node trees are pre-stored in a database associated with the business scenarios. When a new conversation message initiated by a target object is received through an interactive interface, the computer device can determine the business scenario source of the new conversation message according to the information carried by the new conversation message, and then access the database corresponding to the business scenario to retrieve the dialog node tree associated with the business scenario from the database.

[0031] Step S102: triggering acquisition of historical conversation information based on the new conversation message, and constructing conversation prompt information using the acquired historical conversation information.

[0032] Historical session information is the session data initiated by the target object for the same business scenario within the historical time. For example, if the new session message is a product inquiry initiated based on the product purchase scenario, then the historical session information is the product inquiry data initiated by the target object for the same product in the product purchase scenario, such as product material, product details, product delivery time, etc.

[0033] All session data initiated by the target object for the business scenario is stored in the database associated with the business scenario. When a new session message is received, the access to the associated database can be triggered according to the business scenario of the new session message, and the historical session information associated with the new session message can be obtained from the database.

[0034] The dialogue prompt information is used to indicate the dialogue node to which the new conversation information is to flow, and the dialogue prompt information is composed of historical conversation information and conversation model prompt words. Specifically, the historical conversation information is converted into text information, and the text information and the conversation model prompt words are spliced ​​to obtain the corresponding dialogue prompt information. Among them, the splicing position of the text information can be determined based on the conversation model prompt words.

[0035] In a specific example, the text information corresponding to the historical conversation information is [text]. If the conversation model prompt is "Please analyze the probability of the target object in the three aspects of consultation A, consultation B and consultation C based on the above conversation, and output it in JSON format, where key is the dialogue node name and value is a decimal between 0 and 1", then the text information is spliced ​​before the conversation model prompt, and the conversation prompt information obtained is "[text] Please analyze the probability of pointing to the three aspects of A, B and C based on the above conversation, and output it in JSON format, where key is the dialogue node name and value is a decimal between 0 and 1".

[0036] If the conversation model prompt is "Please analyze the consultation probability of the target object in terms of A, B and C based on the following conversation, and output it in JSON format, where key is the conversation node name and value is a decimal between 0 and 1", the text information can be concatenated after the conversation model prompt to obtain the corresponding conversation prompt information "Please analyze the consultation probability of the target object in terms of A, B and C based on the following conversation, and output it in JSON format, where key is the conversation node name and value is a decimal between 0 and 1 [text]".

[0037] The conversation model is a pre-trained model for replying messages. Probability estimation algorithms, such as decision trees, random forests, and Bayesian classifiers, can be integrated into the conversation model to estimate the probability that new conversation information points to each dialogue node. The conversation model can be a large language model that integrates multiple model architectures, a deep learning model trained based on a deep learning model architecture, a machine learning model trained based on a machine learning model architecture, or a network model trained based on a neural network model architecture. The conversation model is not specifically limited here.

[0038] Step S103: determining the target dialog node of each dialog level to which the new conversation message points based on the dialog prompt information.

[0039] The target dialogue node is the dialogue node selected from each dialogue level according to the dialogue prompt information. Specifically, for each current dialogue level, the dialogue prompt information may include the prediction result of the previous dialogue level for the corresponding dialogue node, so as to accurately determine the possibility of pointing to each dialogue node in the current dialogue level. Combined with the possibility of pointing to each dialogue node in each dialogue level, the dialogue node with the greatest possibility in each dialogue level is determined, and the dialogue node with the greatest possibility is determined as the target dialogue node.

[0040] In a specific example, if the current dialogue level is N, the next dialogue level is N+1, and the dialogue level N contains M dialogue nodes, each dialogue node is associated with a number of leaf dialogue nodes in the dialogue level N+1. When it is determined that the probability that the current dialogue level N points to each dialogue node is Pm, where 1≤m≤M, the historical conversation information and the Pm corresponding to each dialogue node are spliced ​​into the conversation model prompt word to update the conversation prompt information. The updated conversation prompt information is input into the conversation model to use the updated conversation prompt information to determine the possibility that the new conversation message points to each dialogue node in the dialogue level N+1. By analogy, the possibility that the new conversation message points to each dialogue node in each dialogue level can be determined. According to the possibility of each dialogue node, the dialogue node with the greatest possibility is determined from each dialogue node in each dialogue level, and the dialogue node with the greatest possibility is determined as the target dialogue node.

[0041] Step S104: construct a question-answer path for a new conversation message according to the target dialogue node.

[0042] The question-answer path is the path formed by the target dialogue nodes that the new conversation message passes through. Specifically, according to the target dialogue nodes pointed to by the new conversation message at each dialogue level, the new conversation message is linked from the target dialogue node in the first dialogue level to the target dialogue node in the last dialogue level in sequence, and the question-answer path for the new conversation message can be constructed.

[0043] The method for determining the question-and-answer path provided in this embodiment triggers the acquisition of historical conversation information when receiving new conversation information of the target object, and uses the historical conversation information to construct dialogue prompt information, thereby combining the dialogue prompt information to determine the target dialogue node pointed to by the new conversation message in each dialogue level of the dialogue node tree, thereby realizing the automatic selection of the target dialogue node. At the same time, by constructing the dialogue prompt information using the historical conversation information related to the new conversation message, the target dialogue node that the new conversation message will pass through in the SOP process can be accurately and quickly located, thereby improving the accuracy of selecting the target dialogue node. Subsequently, the question-and-answer path is constructed using the target dialogue node, thereby ensuring the accuracy of the construction of the question-and-answer path, thereby improving the accuracy of the reply to the new conversation message.

[0044] In this embodiment, a method for determining a question-answer path is provided, which can be used in computer devices such as computers, mobile phones, tablet computers, etc. Figure 3 is a flow chart of a method for determining a question-answer path according to an embodiment of the present disclosure, such as Figure 3 As shown, the process includes the following steps:

[0045] Step S201, receiving a new conversation message initiated by a target object, determining a dialog node tree corresponding to the new conversation message, the dialog node tree including multiple dialog levels, each dialog level including at least one dialog node. For details, please refer to the relevant description of the corresponding steps in the above embodiment, which will not be repeated here.

[0046] Step S202: trigger the acquisition of historical conversation information based on the new conversation message, and construct the conversation prompt information using the acquired historical conversation information. For details, please refer to the relevant description of the corresponding steps in the above embodiment, which will not be repeated here.

[0047] Step S203: determining the target dialog node of each dialog level to which the new conversation message points based on the dialog prompt information.

[0048] Specifically, the above step S203 includes:

[0049] Step S2031: determining the probability that the new conversation message points to each conversation node at each conversation level based on the conversation prompt information.

[0050] The probability is estimated by the conversation model, which reflects the possibility of each dialogue node being selected. The conversation prompt information is input into the pre-trained conversation model, which is used to analyze the conversation prompt information and combine the context information in the historical conversation information to predict the possible reasons for the selection of each dialogue node in each dialogue level. Based on the possible reasons for the selection of the dialogue node, the probability of each dialogue node being selected is estimated.

[0051] In some optional implementations, the above step S2031 includes:

[0052] Step a1, obtaining a first probability that a new conversation message points to each first conversation node in the current conversation level.

[0053] Step a2, obtaining the next dialogue level corresponding to the current dialogue level.

[0054] Step a3: determining, based on the dialog prompt information and the first probability, a second probability that the new conversation message points to each second dialog node in the next dialog level.

[0055] The current dialogue level is the dialogue level currently traversed; the first dialogue node is the dialogue node included in the current dialogue level; the first probability represents the probability of selecting each first dialogue node. Specifically, the dialogue prompt information is input into the pre-trained conversation model, and the conversation model is used to analyze the historical conversation information and the first dialogue node in the current dialogue prompt information, extract key information from the dialogue prompt information, and combine the correlation between the new conversation message and the key information to predict the first probability of each first dialogue node being selected.

[0056] The dialog node tree is traversed according to the level of the dialog node tree to determine the next dialog level corresponding to the current dialog level. The second dialog node is the dialog node included in the next dialog level. The second probability represents the probability of selecting each second dialog node.

[0057] The dialogue node information of the current dialogue level and the first probability corresponding to each dialogue node are added to the dialogue prompt information to obtain the updated dialogue prompt information. The updated dialogue prompt information is input into the conversation model to obtain the probability of each second dialogue node in the next dialogue level, that is, the second probability. By analogy, the probability of each dialogue node being selected in each dialogue level can be calculated in turn.

[0058] By combining the dialogue prompt information and the first probabilities of the current dialogue level, the second probability of each dialogue node in the next dialogue level corresponding to the current dialogue level is determined. As a result, the second probability is predicted based on the first probability, which ensures the correlation between the dialogue nodes and improves the accurate prediction of the probability of each dialogue node.

[0059] In some optional implementations, the above step a1 includes:

[0060] Step a11: if there is a previous dialogue level for the current dialogue level, then the third probability corresponding to each third dialogue node in the previous dialogue level is obtained.

[0061] Step a12: determining the first probability corresponding to each first dialogue node in the current dialogue level based on the historical conversation information and the third probability.

[0062] Step a13: if there is no previous dialog level at the current dialog level, then based on the historical conversation information, determine the first probability corresponding to each first dialog node in the current dialog level.

[0063] The third dialogue node is a dialogue node included in the previous dialogue level corresponding to the current dialogue level; the third probability represents the probability of selecting each third dialogue node. According to the hierarchical structure of the dialogue node tree, it is determined whether the current dialogue level has a previous dialogue level. If it is determined that the current dialogue level has a previous dialogue level, it means that the current dialogue level is not the first dialogue level. At this time, the information of each third dialogue node in the previous dialogue level and the third probability corresponding to each third dialogue node are obtained.

[0064] The information of the third dialogue node of the previous dialogue level and the third probability corresponding to each third dialogue node are added to the dialogue prompt information to obtain updated dialogue prompt information. The updated dialogue prompt information is input into the pre-trained conversation model, and the historical conversation information and the third dialogue node in the updated dialogue prompt information are analyzed by the conversation model to extract key information from the dialogue prompt information, and the first probability of each first dialogue node being selected is predicted in combination with the correlation between the new conversation message and the key information.

[0065] If it is determined that the current dialogue level does not have a previous dialogue level, it means that the current dialogue level is the first dialogue level. At this time, the historical conversation information is directly spliced ​​with the conversation model prompt word to generate the conversation prompt information. The conversation prompt information is input into the pre-trained conversation model, and the conversation model is used to analyze the historical conversation information and the first dialogue node contained in the current dialogue level to extract key information from the conversation prompt information, and combine the correlation between the new conversation message and the key information to predict the first probability of each first dialogue node being selected.

[0066] By judging whether there is a previous dialogue level at the current dialogue level, if there is a previous dialogue level, the first probability of the current dialogue level is determined based on the dialogue prompt information and the third probability of the previous dialogue level; if there is no previous dialogue level, the first probabilities of the current dialogue level are determined based on the dialogue prompt information. In this way, the accuracy of the first probability of the current dialogue level determined is guaranteed, which is conducive to realizing the automatic selection of the target dialogue node according to the probability of each dialogue node.

[0067] In some optional implementations, the above step S2021 further includes:

[0068] Step b1, determining a first target node whose first probability is lower than a preset threshold from each first dialogue node.

[0069] Step b2: determining the second target node corresponding to the first target node from the second dialogue node.

[0070] Step b3, refusing to calculate the second probability corresponding to the second target node.

[0071] The preset threshold is a pre-set probability value for not continuing to perform the calculation. For different dialogue levels, the preset threshold can be set according to the situation, that is, different dialogue levels can correspond to different preset thresholds. The first probability corresponding to each first dialogue node is compared with the preset threshold to determine one or more first target nodes whose first probability is lower than the preset threshold.

[0072] The second target node is the leaf node corresponding to the first target node. Since the second dialog node is at the next dialog level of the current dialog level, the second dialog node corresponding to each first dialog node is the leaf node of the first dialog node. Combined with the association relationship between the first dialog node and the second dialog node, the second target node corresponding to the first target node can be determined from the second dialog node.

[0073] Since the first probability corresponding to the first target dialogue node is less than the preset threshold, the calculation is no longer performed for the first target dialogue node, that is, the second probability corresponding to the second target dialogue node is no longer calculated. Similarly, pruning can be performed for each dialogue node, which can reduce the dialogue nodes that need to calculate probabilities, form a thinking tree framework, and achieve the pruning effect of dialogue nodes.

[0074] The mind tree framework is a prompt engineering (PE) method that uses a sequential language sequence to represent the conversation model's thinking process and is used to express the intermediate steps of generating a question-answering path for a new conversation message. By using the mind tree, the conversation model can evaluate the intermediate thinking of the reasoning question-answering path, and combine the ability to generate and evaluate thinking with search algorithms (such as breadth-first search and depth-first search) so that the question-answering path can be forward verified and backtracked when systematically exploring thinking.

[0075] In a specific implementation, since the preset thresholds for different scenarios and different dialogue levels may also be different, artificially setting the preset thresholds will affect the computational efficiency. Therefore, when pruning dialogue nodes, a rejection model can be introduced. When the rejection model outputs No, it is considered that the rejection model determines that it is unnecessary to continue to execute. The rejection model is trained based on historical conversation data and the corresponding dialogue node selection probability. The rejection model is used to mark the dialogue nodes that are not necessarily calculated as abnormal. In the end, only a small number of dialogue nodes will be calculated to achieve the effect of pruning.

[0076] By comparing the probability of a dialogue node with a preset threshold, for dialogue nodes whose probability is lower than the preset threshold, probability calculation is no longer performed for the dialogue nodes at the next dialogue level, thus reducing the number of dialogue nodes that need to calculate probabilities and reducing calculation time, which is beneficial to improving the selection efficiency of the target dialogue node, thereby improving the efficiency of building the question-and-answer path.

[0077] Step S2032, based on the probabilities of the various dialogue nodes, determining the target dialogue node with the highest probability in each dialogue level from the multiple dialogue nodes.

[0078] After obtaining the probability of each dialogue node in each dialogue level, starting from the first dialogue level of the dialogue node tree, traverse the dialogue nodes in each dialogue level in turn, compare the probabilities of each dialogue node in the traversed dialogue levels, determine the dialogue node with the highest probability in each dialogue level, and determine the dialogue node with the highest probability as the target dialogue node.

[0079] Step S204: construct a question-answer path for a new conversation message according to the target dialogue node. Please refer to the relevant description of the steps in the above embodiment for details, which will not be repeated here.

[0080] The method for determining the question-answer path provided in this embodiment determines the probability of pointing to each dialogue node through dialogue indication information, and determines the target dialogue node from each dialogue node based on the maximum probability, so that the target dialogue node is the dialogue node that the new conversation message is expected to point to, thereby ensuring the prediction accuracy of the target dialogue node and facilitating improving the accuracy of constructing the question-answer path.

[0081] In some optional implementations, the method further includes: generating reply information for the new conversation message according to the question-answer path.

[0082] The reply information is a response generated for the new conversation message. The question-answer path can be used to determine the final target dialogue node to which the new conversation message points. The reply information corresponding to the new conversation message is automatically generated using the content represented by the final target dialogue node, and the target object is notified of the currently generated reply information through the interactive interface, so that the target object knows that the new conversation message has been processed accordingly.

[0083] The method for determining the question-answer path provided in this embodiment generates reply information according to the question-answer path, thereby ensuring that the generation of reply information meets the expectations of the target object and improving the accuracy of the question-answering.

[0084] As a specific application example of the embodiment of the present disclosure, the method for determining the above question-answer path is described here in combination with the commodity consultation application scenario. Figure 4 As shown in FIG. 1 , when the target object initiates a product question, the computer device can receive the product question information (I want to buy product A) initiated by the target object through the interactive interface, and enter the product consultation standard operation procedure (SOP) with three options: product use consultation, product delivery consultation, and product detail consultation. The computer device can obtain the historical conversation information between the target object and the customer service robot, and splice the historical conversation information as prompt text information into the prompt word of the conversation model (the conversation model is pre-deployed in the computer device), such as Figure 4The following figure shows the probability of product usage consultation, product delivery consultation, and product detail consultation according to the above conversation analysis, and output it in JSON format, where key is the name of the dialogue node and value is a decimal between 0 and 1. If there is useful information in the historical conversation information for executing SOP, such as the product delivery time has been mentioned in the historical conversation information, it can be inferred that the product delivery consultation will be selected with a high probability. Get the first dialogue level of SOP, input the dialogue prompt information prompt into the conversation model, and obtain the probability of each dialogue node in the first dialogue level. Get the dialogue node corresponding to each dialogue node in the next dialogue level, add the information of each dialogue node in the first dialogue level and the probability prediction result for each dialogue node output by the conversation model to the dialogue prompt information prompt, and input it into the model again to obtain the probability of each dialogue node in the next dialogue level. By analogy, finally get the probability of all the last-level dialogue nodes, and take the one with the largest probability as the final prediction result, that is, the SOP branch that the target object will eventually go to.

[0085] When calculating the dialogue nodes in each dialogue level, the rejection model is introduced. If the output is No, it is not necessary to continue to perform the probability calculation of the leaf nodes corresponding to the dialogue node, so as to form a thinking tree framework. Figure 5 Some of the dialogue nodes shown (filled with dotted lines) are rejected by the model and output No. Ultimately, only a small number of dialogue nodes (not filled) will have their probabilities calculated, thereby achieving the effect of dialogue node pruning.

[0086] Assuming that through probability calculation, it is determined that "Merchant Shipping Consultation" is the object option corresponding to the second dialogue level, the second round needs to obtain the delivery time options. For example, there are 3 delivery time options: delivery within 24 hours, delivery within 48 hours, and pre-sale within 5 days. Assuming that through probability calculation, it is determined that "delivery within 48 hours" is the object option corresponding to the third dialogue level, and enter the third round of selection. Similarly, the object options of each round are determined in turn through probability calculation. Finally, the target object will go to the reply node of product delivery consultation and generate the reply information that the target object wants. The reply node represents the most fine-grained question that the target object wants to consult.

[0087] In this embodiment, a device for determining a question-answer path is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0088] This embodiment provides a device for determining a question-answer path. Figure 6 As shown, including:

[0089] The information receiving module 301 is used to receive a new conversation message initiated by a target object, and determine a dialog node tree corresponding to the new conversation message. The dialog node tree includes multiple dialog levels, and each dialog level includes at least one dialog node.

[0090] The prompt information construction module 302 is used to trigger the acquisition of historical conversation information based on the new conversation message, and construct the dialogue prompt information using the acquired historical conversation information.

[0091] The target determination module 303 is used to determine the target dialog node of each dialog level to which the new conversation message points based on the dialog prompt information.

[0092] The question-answer path determining module 304 is used to construct a question-answer path of a new conversation message according to the target dialog node.

[0093] In some optional implementations, the target determination module 303 includes:

[0094] The probability determination unit is used to determine the probability that the new conversation message points to each conversation node at each conversation level based on the conversation prompt information.

[0095] The target dialogue node determination unit is used to determine the target dialogue node with the highest probability in each dialogue level from multiple dialogue nodes based on the probability of each dialogue node.

[0096] In some optional implementations, the probability determination unit includes:

[0097] The first probability determination subunit is used to obtain the first probability that the new conversation message points to each first conversation node in the current conversation level.

[0098] The level acquisition subunit is used to obtain the next dialogue level corresponding to the current dialogue level.

[0099] The second probability determination subunit is used to determine, based on the historical conversation information and the first probability, a second probability that the new conversation message points to each second conversation node in the next conversation level.

[0100] In some optional embodiments, the first probability determination subunit is specifically used to, if there is a previous dialogue level for the current dialogue level, obtain the third probability corresponding to each third dialogue node in the previous dialogue level; determine the first probability corresponding to each first dialogue node in the current dialogue level based on historical conversation information and the third probability; if there is no previous dialogue level for the current dialogue level, determine the first probability corresponding to each first dialogue node in the current dialogue level based on historical conversation information.

[0101] In some optional implementations, the probability determination unit further includes:

[0102] The first target node determination subunit is used to determine, from each first dialogue node, a first target node whose first probability is lower than a preset threshold.

[0103] The second target node determination subunit is used to determine the second target node corresponding to the first target node from the second dialogue node.

[0104] The rejection subunit is used to refuse to calculate the second probability corresponding to the second target node.

[0105] In some optional embodiments, the above device further includes:

[0106] The reply module is used to generate reply information for the new conversation message according to the question-answer path.

[0107] The question-answer path determination device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0108] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0109] The device for determining the question-and-answer path provided by the embodiment of the present disclosure triggers the acquisition of historical conversation information when receiving new conversation information of the target object, and uses the historical conversation information to construct dialogue prompt information, thereby combining the dialogue prompt information to determine the target dialogue node pointed to by the new conversation message in each dialogue level of the dialogue node tree, thereby realizing the automatic selection of the target dialogue node. At the same time, by constructing the dialogue prompt information using the historical conversation information related to the new conversation message, the target dialogue node that the new conversation message is to pass through in the SOP process can be accurately and quickly located, thereby improving the accuracy of selecting the target dialogue node. Subsequently, the question-and-answer path is constructed using the target dialogue node, thereby ensuring the accuracy of the construction of the question-and-answer path, thereby improving the accuracy of the reply to the new conversation message.

[0110] The present disclosure also provides a computer device having the above Figure 6 The device for determining the question and answer path shown.

[0111] See also Figure 7 , Figure 7 is a schematic diagram of a computer device provided by an optional embodiment of the present disclosure, such as Figure 7As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 7 A processor 10 is taken as an example.

[0112] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0113] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0114] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the use of a computer device based on the presentation of a small program landing page, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0115] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0116] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 7 The example of connecting through bus is taken in the following.

[0117] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0118] The computer device also includes a communication interface, which is used for the computer device to communicate with other devices or a communication network.

[0119] The embodiments of the present disclosure also provide a computer-readable storage medium. The above-mentioned method according to the embodiments of the present disclosure can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium and downloaded through a network, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0120] A part of the present disclosure may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present disclosure through the operation of the computer. Those skilled in the art should understand that the existence of computer program instructions in computer-readable media includes, but is not limited to, source files, executable files, installation package files, etc., and accordingly, the way in which computer program instructions are executed by a computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to the computer.

[0121] Although the embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for determining a question-answer path, characterized in that: The method comprises: Receive a new conversation message initiated by a target object, and determine a dialog node tree corresponding to the new conversation message, wherein the dialog node tree includes a plurality of dialog levels, and each of the dialog levels includes at least one dialog node; triggering acquisition of historical conversation information based on the new conversation message, and constructing conversation prompt information using the acquired historical conversation information; Determine, based on the conversation prompt information, a target conversation node at each of the conversation levels to which the new conversation message points; A question-answer path of the new conversation message is constructed according to the target dialog node.

2. The method according to claim 1, characterized in that The determining, based on the dialog prompt information, a target dialog node of each dialog level to which the new dialog message points includes: Determining, based on the dialogue prompt information, the probability that the new conversation message points to each dialogue node at each dialogue level; Based on the probabilities of the respective dialog nodes, a target dialog node with the highest probability in each dialog level is determined from the plurality of dialog nodes.

3. The method according to claim 2, characterized in that The determining, based on the dialog prompt information, the probability that the new conversation message points to each dialog node at each dialog level includes: Obtaining a first probability that the new conversation message points to each first conversation node in the current conversation level; Obtaining a next dialog level corresponding to the current dialog level; Based on the dialog prompt information and the first probability, a second probability that the new conversation message points to each second dialog node in the next dialog level is determined.

4. The method according to claim 3, characterized in that Also includes: Determine, from each of the first dialogue nodes, a first target node whose first probability is lower than a preset threshold; Determine a second target node corresponding to the first target node from the second dialogue node; Refuse to calculate the second probability corresponding to the second target node.

5. The method according to claim 3, characterized in that: The obtaining of the first probability that the new conversation message points to each first conversation node in the current conversation level includes: If there is a previous dialogue level for the current dialogue level, obtaining a third probability corresponding to each third dialogue node in the previous dialogue level; Based on the historical conversation information and the third probability, a first probability corresponding to each of the first dialog nodes in the current dialog level is determined.

6. The method according to claim 5, characterized in that The obtaining the first probability that the new conversation message points to each first conversation node in the current conversation level includes: If the current dialog level does not have a previous dialog level, a first probability corresponding to each of the first dialog nodes in the current dialog level is determined based on the historical conversation information.

7. The method according to claim 1, characterized in that Also includes: Generate reply information for the new conversation message according to the question-answer path.

8. A device for determining a question-answer path, characterized in that: The device comprises: An information receiving module, configured to receive a new conversation message initiated by a target object, and determine a dialog node tree corresponding to the new conversation message, wherein the dialog node tree includes a plurality of dialog levels, and each of the dialog levels includes at least one dialog node; A prompt information construction module, used to trigger the acquisition of historical conversation information based on the new conversation message, and construct dialogue prompt information using the acquired historical conversation information; A target determination module, configured to determine a target dialog node of each dialog level to which the new conversation message points based on the dialog prompt information; The question-answer path determining module is used to construct the question-answer path of the new conversation message according to the target dialogue node.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for determining a question-and-answer path according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for determining a question-and-answer path according to any one of claims 1 to 7.

11. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the method for determining a question-answer path according to any one of claims 1 to 7.

Images