A response method, device, equipment and computer-readable storage medium

Abstract

The present application discloses an answering method, comprising: obtaining a response corpus for query information; wherein the query information represents query information in multiple rounds of answers; determining, from the response corpus, answering nodes to be screened that match the query information; obtaining a response weight for each answering node to be screened, and screening the answering nodes to be screened based on the response weight to determine a target answering node; and answering the query information based on the answer information of the target answering node. The present application also discloses an answering device, apparatus, and computer-readable storage medium.

Description

Technical Field

[0001] The present invention relates to the field of human-computer interaction technology, and in particular to a response method, device, equipment and computer-readable storage medium. Background Art

[0002] In commercial intelligent response systems, multi-round responses are generally supported to ensure user experience. In contrast, in technology, the sequence of response nodes in multi-round responses is usually fixed, and no relevant mechanism for jump responses between unrelated nodes is given. This makes it easy for response nodes in multi-round responses to have poor response effects, which makes the response intelligence low. Summary of the Invention

[0003] In order to solve the above technical problems, the embodiments of the present application hope to provide a response method, device, equipment and computer-readable storage medium, which improve the response effect when entering a multi-round response process, thereby improving the response intelligence.

[0004] To achieve the above objectives, the technical solution of this application is implemented as follows:

[0005] A response method, comprising:

[0006] Acquire a response corpus for query information; wherein the query information represents consulting information in multiple rounds of responses;

[0007] Determining, from the response corpus, a response node to be screened that matches the query information;

[0008] Obtaining a response weight of each of the to-be-screened response nodes, and screening the to-be-screened response nodes based on the response weights to determine a target response node;

[0009] The query information is responded to based on the response information of the target response node.

[0010] In some embodiments, determining, from the response corpus, a response node to be screened that matches the query information includes:

[0011] Obtaining the round number of the query information in the multiple rounds of responses;

[0012] Obtaining a pre-matched response node for the round number in the multiple rounds of responses;

[0013] If it is determined that there is no to-be-matched answer node matching the query information in the pre-matched answer node, the to-be-screened answer node matching the query information is determined from the answer corpus.

[0014] In some implementations, obtaining the response weight of each of the to-be-screened response nodes includes:

[0015] Obtaining a distance value between each of the to-be-screened answer nodes and the pre-matched answer node;

[0016] Obtaining the number of responses of each of the response nodes to be screened within a preset period;

[0017] Based on the distance value and the number of responses, a response weight of each of the to-be-screened response nodes is determined.

[0018] In some embodiments, the method further comprises:

[0019] When it is determined that there is no to-be-screened answer node matching the query information from the answer corpus, obtaining the initial answer node; wherein the initial answer node represents the first answer node in the multiple rounds of answers;

[0020] Processing the query information and the initial response node within a preset period to generate a feedback mapping table; wherein the feedback mapping table is used to represent the correspondence between the query information and the initial response node;

[0021] Updating the response corpus based on the feedback mapping table to obtain a target corpus;

[0022] The next round of inquiry information is answered based on the target corpus.

[0023] In some implementations, processing the query information and the initial response node within a preset period to generate a feedback mapping table includes:

[0024] Processing each of the initial response nodes and each of the query information within a preset period to generate information to be processed;

[0025] Processing the information to be processed using a text clustering algorithm to obtain a group of information to be fed back that matches a target number; wherein the target number represents the number of the initial response nodes within a preset period;

[0026] Perform intent recognition and classification on the information group to be fed back, and generate a feedback mapping table.

[0027] In some embodiments, the processing of each initial response node and each query message within a preset period to generate information to be processed includes:

[0028] Perform word segmentation processing on each query information within a preset period and remove stop words to generate a corpus to be processed;

[0029] Performing word segmentation processing and stop word removal on the response information of each of the initial response nodes within a preset period to obtain pre-processed information of each of the initial response nodes;

[0030] Using regular expressions to filter the pre-processed information for to-be-processed vocabulary, and obtaining a business word for each of the initial response nodes; wherein the to-be-processed vocabulary represents information related to item transactions;

[0031] The business words are matched with the corpus to be processed to generate the information to be processed corresponding to the initial response node.

[0032] In some embodiments, the performing of intent recognition and classification on the information group to be fed back and generating a feedback mapping table includes:

[0033] Obtaining sample information, and training the sample information to generate a language classification model;

[0034] Using the language classification model, classify each piece of information to be fed back in the group of information to be fed back according to business, and perform intent recognition on the classification results to obtain a target corpus for each piece of information to be fed back;

[0035] Based on the corresponding relationship between the target corpus and the information to be fed back, and the corresponding relationship between the information to be fed back and the initial response node, a feedback mapping table between the initial response node and the target corpus is generated.

[0036] In some implementations, after processing the query information and the initial response node within a preset period to generate a feedback mapping table, the method further includes:

[0037] Create a preprocessing feedback branch;

[0038] Based on the corpus similarity strategy and the preprocessing feedback branch, the feedback mapping table is updated to a target feedback mapping table.

[0039] A response device, comprising: an acquisition module, a determination module and a response module;

[0040] The acquisition module is used to acquire a response corpus for the inquiry information; wherein the inquiry information represents the consultation information in multiple rounds of responses;

[0041] The determining module is configured to determine, from the response corpus, a response node to be screened that matches the query information;

[0042] The determining module is further configured to obtain a response weight of each of the to-be-screened response nodes, and screen the to-be-screened response nodes based on the response weight to determine a target response node;

[0043] The response module is configured to respond to the inquiry information based on the response information of the target response node.

[0044] A response device comprising: a processor, a memory, and a communication bus;

[0045] The communication bus is used to realize the communication connection between the processor and the memory;

[0046] The processor is configured to execute the program in the memory to implement any of the above-mentioned response methods.

[0047] A computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the above-mentioned response method.

[0048] The response method, apparatus, device and computer-readable storage medium provided in the embodiments of the present application obtain a response corpus for query information; wherein the query information represents the query information in multiple rounds of responses; determine the response nodes to be screened that match the query information from the response corpus; obtain the response weight of each of the response nodes to be screened, and screen the response nodes to be screened based on the response weight to determine the target response node; respond to the query information based on the response information of the target response node; in this way, in the response corpus, determine the response nodes to be screened that match the query information, and then screen the target response nodes that match the query information from the response nodes to be screened based on the response weight, and respond to the query information based on the response information of the target response node, thereby improving the response effect in the multiple rounds of response process and thereby improving the intelligence of the response. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a diagram of the configuration of the order modification process in the relative technology of this application;

[0050] Figure 2 A flowchart of a response method provided in an embodiment of the present application;

[0051] Figure 3 A flowchart of another response method provided in an embodiment of the present application;

[0052] Figure 4 A flowchart of another response method provided in an embodiment of the present application;

[0053] Figure 5 A flowchart of another response method provided in an embodiment of the present application;

[0054] Figure 6A schematic diagram of the configuration of the order modification process provided in an embodiment of the present application;

[0055] Figure 7 A schematic diagram of a system structure for performing multiple rounds of responses provided in an embodiment of the present application;

[0056] Figure 8 A schematic diagram of a system flow for performing multiple rounds of responses provided in an embodiment of the present application;

[0057] Figure 9 A schematic diagram of the structure of a response device provided in an embodiment of the present application;

[0058] Figure 10 A schematic diagram of the structure of a response device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application.

[0060] It should be understood that the “embodiments of the present application” or “the aforementioned embodiments” mentioned throughout the specification mean that the specific features, structures or characteristics related to the embodiments are included in at least one embodiment of the present application. Therefore, “in the embodiments of the present application” or “in the aforementioned embodiments” appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. In the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The above-mentioned serial numbers of the embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments.

[0061] In order to enable a more detailed understanding of the features and technical contents of the embodiments of the present application, the implementation of the embodiments of the present application is described in detail below with reference to the accompanying drawings. The attached drawings are for reference only and are not used to limit the present application.

[0062] In terms of technology, intelligent response systems are generally composed of natural language understanding (NLU) and dialogue management (DM). Among them, NLU is mainly based on the convolutional neural network (CNN) deep learning algorithm model to identify user intent, and then DM decides and outputs the response based on the NLU recognition results.

[0063] At the same time, in order to ensure user experience, commercial intelligent response systems all support multi-round responses. Among them, multi-round response is a function and strategy based on multi-round interactive decision-making and response. It is a process in which the response system continuously decides the optimal response based on the current and historical status during the response process. At the same time, the core of multi-round response is the multi-round response decision-making ability of DM. The current mainstream commercial multi-round response system mainly implements multi-round response based on pre-configured response processes. The multi-round response process configuration is pre-configured by business operations personnel and takes effect after activation.

[0064] The main basic framework for executing multi-round responses is the process engine, and its summary execution process is as follows:

[0065] Step 1: First, determine the first execution response node: if it is a new session, start execution from the process head; otherwise, find the last response node as the starting response node;

[0066] Step 2: Starting from the first execution of the answer node, match the subsequent branch rules. If the match is successful, enter the branch answer. If all matches fail, repeat pushing the current node answer;

[0067] Step 3: Record the response node in the process response status. When the next response comes in, match and respond from this node.

[0068] like Figure 1 A schematic diagram of the configuration of the order modification process in the relative technology is given; among them, the first consulting question entered into the multi-round response is: "Modify order", and the answer node corresponding to "Modify order" is entered. The system automatically replies: "Please enter the order number you want to modify", and enters the "Order number filling slot"; among them, after the system obtains the relevant information and displays that the order number is empty, it continues to display "Order number filling slot", if the order number is not empty, it enters the initial answer node corresponding to the consulting information, that is, the "Modify answer" node, and the system replies according to the node: "Your order XXX is a third-party delivery, has been shipped, and may not support modification; you can go to the "My Order" page to select the order and try to modify it"; the user enters similar language For example: Why can't I see the Modify button? If I can't see the Modify button, the system enters the "Cancel Response" node, and the system replies based on the current node: "Well, if you can't see the Modify button, it may have been canceled; if you really want to modify it, you can choose to cancel and place another order; you can go to the "My Orders" page and click the Cancel button for the corresponding order"; if the user inputs or the system receives another inquiry question: "Why can't I see the Cancel button?", the system enters the "Reject Response" node again, and the system replies based on the current node: "Well, if the Cancel button is not seen, it should be that the order status does not support cancellation; if you still insist on modifying it, you can choose to contact the delivery to refuse or wait for the delivery to contact you to refuse." At the same time Figure 1Taking the modification of order process configuration as an example, the response effect is shown in Table 1 below:

[0069]

[0070] Table 1

[0071] Among them, the user input matching in the rule is based on the semantic similarity matching model, which matches the semantic similarity of the user input with a similar corpus set. If the similarity is greater than the similarity threshold, the match is considered successful; the similarity matching model adopts the Deep Structured Semantic Models (DSSM).

[0072] In terms of technology, pre-configured feedback responses in multi-round responses are mainly achieved through two methods: 1. Operations personnel infer user feedback from the answers and configure multi-round responses (for example, modifying the order process configuration); 2. Operations personnel invest manpower every week to annotate and sort out the multi-round user feedback in the online response system log and configure the response process. Judging from the operation and sorting situation, user input is open and unrestricted. The configuration strategies of methods 1 and 2 have a very small coverage range and are highly dependent on the input of operations personnel, which is very costly.

[0073] Furthermore, the sequence of response nodes in multiple rounds of responses is fixed, making the flow inflexible. The process lacks a fallback mechanism, and responses cannot be redirected between unrelated nodes. This leads to the following problems: Feedback relationships exist between multiple responses, requiring configuration to form a complex mesh structure that is difficult to maintain. If a node's subsequent rules fail to match, or if there are no subsequent rules, the same response is repeated, resulting in poor response quality. As shown in Table 2, in the order modification process, any subsequent user input of inquiry information will result in the same answer being pushed.

[0074]

[0075] Table 2

[0076] Based on the above problems, the embodiment of the present application provides a response method, which is applied to a response device, referring to Figure 2 As shown, the method includes the following steps:

[0077] Step 101: Obtain a response corpus for query information.

[0078] The inquiry information represents the consultation information in multiple rounds of responses.

[0079] In the embodiment of the present application, the inquiry information is generally composed of a question sentence; wherein the inquiry information can be generated by the user inputting a relevant phrase or sentence, and the answering device needs to respond to the information.

[0080] Among them, multi-round responses can be automatic questions and answers during multiple rounds of interactions; among them, the number of responses in a multi-round response can be two or more, and this application does not impose any restrictions on this; at the same time, multi-round responses can be conversational questions and answers between devices, or conversational questions and answers between devices and people.

[0081] It should be noted that the response corpus is a corpus of inquiry information obtained by the reply response device provided by the system, wherein the response corpus includes response nodes and the response information of the response nodes; the number of response nodes and the number of response information corresponding to the response nodes are not limited in this application.

[0082] Step 102: Determine the response nodes to be screened that match the query information from the response corpus.

[0083] In an embodiment of the present application, the response device can select a response node to be screened that matches the query information from the response corpus based on a matching condition; wherein the matching condition can represent the presence of keywords similar to the query information in the response node, or can represent a certain correlation relationship between the response node and the query information.

[0084] The number of response nodes to be screened is not limited in this application.

[0085] Step 103: Obtain the response weight of each to-be-screened response node, and screen the to-be-screened response nodes based on the response weight to determine a target response node.

[0086] In the embodiment of the present application, the response weight of each response node to be screened may be obtained by calculating relevant parameters, or may be automatically generated by the system.

[0087] In a feasible implementation, the response device may obtain the response weight of the response node by performing correlation calculation on the distance between the response node and the current response node and the number of responses of the response node within a preset period.

[0088] It should be noted that the response device selects the response node with the highest response weight from the response nodes to be screened as the target response node according to the calculated response weight.

[0089] Step 104: Respond to the inquiry information based on the response information of the target response node.

[0090] In an embodiment of the present application, the response device responds to the inquiry information based on the response information of the target response node; wherein the correlation between the response information of the target response node and the inquiry information is the information with the highest correlation with the inquiry information among the response nodes to be screened.

[0091] The response method provided in the embodiment of the present application obtains a response corpus for query information; wherein the query information represents the query information in multiple rounds of responses; determines the to-be-screened response nodes that match the query information from the response corpus; obtains the response weight of each to-be-screened response node, and screens the to-be-screened response nodes based on the response weight to determine the target response node; responds to the query information based on the response information of the target response node; in this way, the to-be-screened response nodes that match the query information are determined in the response corpus, and then the target response nodes that match the query information are screened from the to-be-screened response nodes based on the response weight, and the query information is responded to based on the response information of the target response node, thereby improving the response effect in the multiple-round response process and thereby improving the response intelligence.

[0092] Based on the above embodiments, the embodiments of the present application provide a response method, referring to Figure 3 As shown, the method includes the following steps:

[0093] Step 201: The answering device obtains a response corpus for query information.

[0094] The inquiry information represents the consultation information in multiple rounds of responses.

[0095] Step 202: The answering device obtains the round number of the inquiry information in the multi-round answering.

[0096] In an embodiment of the present application, the number of rounds of multi-round responses is N; wherein N is greater than or equal to 1; illustratively, when N is 5, the number of rounds of inquiry information in the multi-round responses can be the first round, the fifth round, or even the third round, and the present application does not impose any limitation on this.

[0097] Step 203: The response device obtains a pre-matched response node with a round number in the multi-round response.

[0098] In the embodiment of the present application, the response node corresponding to each round number in the multiple rounds of response includes at least one response node and its corresponding response information.

[0099] In one feasible implementation, when the multi-round response is a multi-round response for modifying an order, when the multi-round response is the first round, the corresponding pre-match response node is "Modify Order"; when the multi-round response is the second round, the corresponding pre-match response node may be "Modify Response," "Change Modification Method," etc. This application does not impose any limitations on this.

[0100] Step 204: If the answering device determines that there is no answering node to be matched that matches the query information in the pre-matched answering nodes, it determines an answering node to be screened that matches the query information from the answer corpus.

[0101] In an embodiment of the present application, if the answering device does not match the answering node to be matched related to the query information from the acquired pre-matching answering nodes, it needs to determine the answering node to be screened that matches the query information from the answer corpus.

[0102] Step 205: The answering device obtains the answering weight of each answering node to be screened, and screens the answering nodes to be screened based on the answering weight to determine a target answering node.

[0103] In a feasible implementation, when the answering device performs step 205 to obtain the answering weight of each answering node to be screened, it can also be implemented by performing steps 205a to 205c:

[0104] Step 205a: The answering device obtains the distance value between each answering node to be screened and the pre-matched answering node.

[0105] In the embodiment of the present application, the answering device calculates the distance value between each answering node to be screened and the pre-matched answering node by using Dijkstra algorithm to calculate relevant parameters.

[0106] It should be noted that Dijkstra mainly solves the algorithm for the shortest path problem in a weighted graph, that is, the shortest path algorithm from one vertex to the remaining vertices.

[0107] Each answer node to be screened is represented by an X node, and the distance values of the remaining pre-matched answer nodes can be represented by distance(X).

[0108] Step 205b: The answering device obtains the number of responses of each answering node to be screened within a preset period.

[0109] Based on the above, the number of responses of each to-be-screened response node within a preset period can be represented by responseTime(X); wherein, the preset period is not limited in this application and can be one month, one week, or even one day.

[0110] Step 205c: The answering device determines the answering weight of each answering node to be screened based on the distance value and the number of answers.

[0111] In the embodiment of the present application, the response weight of each to-be-screened response node can be represented by W(X); wherein W(X) can be calculated by the following formula (1):

[0112] W(X)=lg(responseTime(X))*1 / distance(X) (1)

[0113] Step 206: The response device responds to the inquiry information based on the response information of the target response node.

[0114] It should be noted that, for the description of the same steps and contents in this embodiment as those in other embodiments, reference can be made to the description in other embodiments and will not be repeated here.

[0115] The response method provided in the embodiment of the present application determines the response nodes to be screened that match the query information in the response corpus, and then screens the target response nodes that match the query information from the response nodes to be screened based on the response weight, and responds to the query information based on the response information of the target response node, thereby improving the response effect in the multi-round response process and thereby improving the intelligence of the response.

[0116] Based on the above embodiments, the embodiments of the present application provide a response method, referring to Figure 4 As shown, the method includes the following steps:

[0117] Step 301: The answering device obtains a response corpus for query information.

[0118] The inquiry information represents the consultation information in multiple rounds of responses.

[0119] Step 302: The answering device obtains the round number of the inquiry information in the multi-round answering.

[0120] Step 303: The response device obtains a pre-matched response node with a round number in the multi-round response.

[0121] Step 304: If the answering device determines that there is no answering node to be matched that matches the query information in the pre-matched answering nodes, it determines an answering node to be screened that matches the query information from the answer corpus.

[0122] As shown in Table 3, in the multiple rounds of responses for modifying an order, after obtaining the initial response node "modify response", the query information input by the user into the response device does not receive any feedback, that is, there is no corresponding relationship in the response corpus with the response to be matched with the query information.

[0123]

[0124] Table 3

[0125] Step 305: When the response device determines from the response corpus that there is no response node to be screened that matches the query information, it obtains an initial response node.

[0126] In an embodiment of the present application, when the response device cannot match the to-be-screened response node related to the query information in the response corpus, it needs to obtain the first response node in the multiple rounds of responses in which the query information is located.

[0127] Step 306: The response device processes the query information and the initial response node within a preset period to generate a feedback mapping table.

[0128] The feedback mapping table is used to represent the correspondence between the query information and the initial response node.

[0129] In a feasible implementation, when the answering device executes step 306, it may also execute steps 306a to 306c:

[0130] Step 306a: The response device processes each initial response node and each inquiry message within a preset period to generate information to be processed.

[0131] In an embodiment of the present application, the response device obtains at least one initial response node and at least one inquiry message within a preset period, processes the at least one initial response node and the at least one inquiry message respectively, and obtains information to be processed based on the processing results.

[0132] The initial response node is the first response node when the query information in the multi-round response obtained by the system does not match the corresponding response node in the answer corpus.

[0133] Step 306b: The answering device processes the information to be processed using a text clustering algorithm to obtain a group of information to be fed back that matches the target number.

[0134] The target number represents the number of initial response nodes within a preset period;

[0135] In an embodiment of the present application, the response device clusters the information to be processed using a text clustering algorithm based on the number of initial response nodes to obtain a group of information to be fed back that matches the target number; wherein the group of information to be fed back includes at least one piece of information to be fed back.

[0136] It should be noted that the text clustering algorithm is mainly based on the idea that documents of the same type have a greater similarity, while documents of different types have a smaller similarity to perform clustering processing on text information.

[0137] Step 306c: The response device performs intent recognition and classification on the group of information to be fed back, and generates a feedback mapping table.

[0138] In an embodiment of the present application, intent recognition is to classify text information based on the intent represented by the text information.

[0139] The answering device performs intent recognition and classification on each piece of information to be fed back in the group of information to be fed back, and obtains a classification result.

[0140] Step 307: The response device updates the response corpus based on the feedback mapping table to obtain a target corpus.

[0141] In an embodiment of the present application, the response device updates the response corpus based on the generated feedback mapping table; in a feasible implementation, the feedback mapping table may be integrated into the response corpus according to preset rules to generate a target corpus.

[0142] Step 308: The answering device answers the next round of inquiry information based on the target corpus.

[0143] In an embodiment of the present application, the answering device will answer the inquiry information in a new round of multi-round answering based on the newly generated target corpus.

[0144] It should be noted that, for the description of the same steps and contents in this embodiment as those in other embodiments, reference can be made to the description in other embodiments and will not be repeated here.

[0145] The response method provided in the embodiment of the present application determines the response nodes to be screened that match the query information in the response corpus, and then screens the target response nodes that match the query information from the response nodes to be screened based on the response weight, and responds to the query information based on the response information of the target response node, thereby improving the response effect in the multi-round response process and thereby improving the intelligence of the response.

[0146] Based on the above embodiments, the embodiments of the present application provide a response method, referring to Figure 5 As shown, the method includes the following steps:

[0147] Step 401: The answering device obtains a response corpus for query information.

[0148] The inquiry information represents the consultation information in multiple rounds of responses.

[0149] Step 402: The answering device obtains the round number of the inquiry information in the multi-round answering.

[0150] Step 403: The response device obtains a pre-matched response node with a round number in the multi-round response.

[0151] Step 404: If the answering device determines that there is no answering node to be matched that matches the query information in the pre-matched answering nodes, it determines an answering node to be screened that matches the query information from the answer corpus.

[0152] Step 405: When the response device determines from the response corpus that there is no response node to be screened that matches the query information, it obtains an initial response node.

[0153] Step 406: The response device processes each initial response node and each inquiry message within a preset period to generate information to be processed.

[0154] In a feasible implementation, when executing step 406, the answering device may also implement the following by executing steps 406a to 406d:

[0155] Step 406a: The answering device performs word segmentation processing on each inquiry message within a preset period and removes stop words to generate a corpus to be processed.

[0156] In the embodiment of the present application, the answering device performs word segmentation processing on each query message within a preset period and removes stop words within each query message to obtain a corpus to be processed.

[0157] The answering device may perform word segmentation processing on each query information based on NLU.

[0158] It should be noted that stop words can generally be divided into two categories. One category is function words contained in human language. These function words are extremely common and have no actual meaning compared to other words, such as these, which, this or that. Another category of words cannot guarantee that they can provide truly relevant search results when used in search engines. They are difficult to help narrow the search scope and will also reduce the efficiency of the search. Table 4 shows the results of word segmentation processing for different query information and the processing results after stop words are removed.

[0159] Request information Word segmentation results Processing results after stop words are removed Why third-party delivery? "Third Party", "Delivery" reserve What to do if modification is not supported "Revise" reserve Don't quite understand "Understand" delete

[0160] Table 4

[0161] Step 406b: The response device performs word segmentation processing on the response information of each initial response node within a preset period and removes stop words to obtain pre-processed information of each initial response node.

[0162] In an embodiment of the present application, the response device performs word segmentation processing on the response information corresponding to each initial response node within a preset period based on NLU, and removes stop words therein to obtain preprocessing information of each initial response node; Table 5 shows the results of word segmentation processing and stop word removal of the response information in the "Modify Response" node.

[0163]

[0164] Table 5

[0165] Step 406c: The answering device uses regular expressions to filter the pre-processed information for to-be-processed vocabulary to obtain the business words of each initial answering node.

[0166] The vocabulary to be processed represents information related to item transactions.

[0167] In the embodiment of the present application, the response device uses regular expressions to filter the information related to the item transaction in the pre-processed information of each initial response node to obtain the business words of each initial response node.

[0168] The vocabulary to be processed, i.e., information related to the item transaction, may be: order number, product number, telephone number, address, etc.; this application does not impose any limitation on this.

[0169] Step 406d: The answering device matches the business words with the corpus to be processed, and generates information to be processed corresponding to the initial answering node.

[0170] In the embodiment of the present application, the response device matches the business words of each initial response node with the corpus to be processed to generate information to be processed corresponding to each initial response node.

[0171] Step 407: The answering device processes the information to be processed using a text clustering algorithm to obtain a group of information to be fed back that matches the target number.

[0172] Step 408: The response device performs intent recognition and classification on the group of information to be fed back, and generates a feedback mapping table.

[0173] In a feasible implementation, when the answering device executes step 408, it may also execute steps 408a to 408c:

[0174] Step 408a: The answering device obtains sample information and trains the sample information to generate a language classification model.

[0175] In an embodiment of the present application, the answering device may obtain sample information including a variety of corpora for training to generate a language classification model; wherein the language classification model may be a language classification model obtained based on NLU.

[0176] Step 408b: The answering device uses a language classification model to classify each piece of information to be fed back in the feedback information group according to the business, and performs intent recognition on the classification results to obtain the target corpus of each piece of information to be fed back.

[0177] In the embodiment of the present application, the answering device classifies each piece of information to be fed back in each group of information to be fed back according to the business, and performs intent recognition on the classification results to obtain the target corpus of each piece of information to be fed back.

[0178] Among them, the answering device uses a language classification model to classify and filter the information to be fed back in each information group to be fed back according to business, and obtains a classification result; the answering device performs intent recognition on each result in the classification result, such as if the proportion of intent represented by a certain classification result meets the preset conditions; in a feasible implementation method, a preset threshold of 80% can be given, that is, if the intention proportion exceeds 80%, it means that the intention proportion meets the preset conditions; the result is retained as the target corpus of the information to be fed back; as shown in Table 6, the classification results corresponding to the two types of information to be fed back and the relationship between the target corpus are given.

[0179] Pending feedback information Classification results target corpus 1 Delivery type and method Why third-party delivery? 2 Modify order What to do if modification is not supported

[0180] Table 6

[0181] Step 408c: The response device generates a feedback mapping table between the initial response node and the target corpus based on the correspondence between the target corpus and the information to be fed back, and the correspondence between the information to be fed back and the initial response node.

[0182] The feedback mapping table is used to represent the correspondence between the query information and the initial response node.

[0183] As shown in Table 7, the correspondence between the initial response node, the target corpus, and the information to be fed back is given.

[0184]

[0185] Table 7

[0186] Step 409: The response device updates the response corpus based on the feedback mapping table to obtain a target corpus.

[0187] Step 410: The answering device answers the next round of inquiry information based on the target corpus.

[0188] In the embodiment of the present application, since the target corpus has a feedback mapping table, in the next round of "modify order" multi-round response process, the following can be used: Figure 6Reply in the given way, among which; enter the first consultation question entered in multiple rounds of responses: "Modify Order", enter the answer node corresponding to "Modify Order", the system automatically replies: "Please enter the order number you want to modify", enter the "Order Number Filling Slot"; among which, after the system obtains the relevant information and displays that the order number is empty, it will continue to display "Order Number Filling Slot", if the order number is not empty, then enter the initial answer node corresponding to the consultation information, that is, the "Modify Answer" node, the system replies according to the node: "Your order XXX is a third-party delivery, has been shipped, and may not support modification; you can go to the "My Orders" page to select the order and Try to modify it"; if the user enters similar corpus such as: "Why can't I see the modify button", that is, the modify button cannot be seen, the system enters the "cancel response" node. If the user enters similar corpus "Why is it a third-party delivery" or "Which third-party delivery company is it", that is, delivery type and method, the system enters feedback response 1; if the user enters similar corpus such as "What should I do if modification is not supported", "I tried, but cannot modify" or "I just can't modify it", that is, "delivery type and method", the system enters feedback response 2; the user enters the next query information in turn, and the system will match the keyword of the target corpus to find the corresponding feedback response n.

[0189] It should be noted that, for the description of the same steps and contents in this embodiment as those in other embodiments, reference can be made to the description in other embodiments and will not be repeated here.

[0190] The response method provided in the embodiment of the present application determines the response nodes to be screened that match the query information in the response corpus, and then screens the target response nodes that match the query information from the response nodes to be screened based on the response weight, and responds to the query information based on the response information of the target response node, thereby improving the response effect in the multi-round response process and thereby improving the intelligence of the response.

[0191] In other embodiments of the present application, Figure 7As shown, a system structure diagram for performing multi-round responses is provided; the response method provided by the embodiment of the present application can be applied to this architecture, and the process of executing multi-round responses is based on a process engine. The process engine provides a corresponding response node for the first inquiry information input by the user based on the system's response process configuration and weighted matching response module. At the same time, the first inquiry information can be the first round of questions in the multi-round response, or it can be the Nth round of questions; N is a positive integer greater than or equal to 2; if the weighted matching response module matches the to-be-screened response node associated with the inquiry information in the response node corresponding to its round number in the multi-round response link, the to-be-screened response node can be matched and the response weight of the relevant response node can be calculated, and then the target response node can be determined based on the response weight. And based on the response information of the target response node, the second query information input by the user is responded to; the response process is configured to match the first query information input by the user with the response scheme in the relative technology; the feedback branch clustering module is based on the response process configuration or the weighted matching response module. When the matching node corresponding to the query information cannot be obtained, the initial response node in the multi-round response where the query information is located is obtained at the same time, and the corresponding feedback mapping table is obtained based on the query information and the initial response node, and the response corpus is updated based on the feedback mapping table to obtain the target corpus; at the same time, in the next multi-round response process or when the next round of query information in this multi-round response comes, the weighted matching response module matches and responds to the query information based on the target corpus; as shown in Table 8, the application Figure 7 In the system shown in FIG1 , in multiple rounds of responses, query information of the first round, query information of the second round, and query information of the third round respectively correspond to response information of the response node.

[0192] like Figure 8 As shown in the figure, a schematic diagram of the system flow of multiple rounds of responses is given; it is mainly aimed at Figure 7The specific implementation process in the weighted matching response module and the feedback branch clustering module is described. In the feedback branch clustering module, first execute steps 801 and 802, obtain and analyze the query information and the corresponding initial response node within the preset period; and execute step 803 to judge the initial response node of each query information. If a query information does not have a corresponding initial response node, that is, the query information enters the first round of multiple rounds of responses, and the system does not match the corresponding response node, then the process ends directly; if the query information has a corresponding initial response node, execute step 804 to match and filter the query information and the initial response node; obtain the matching result step 805; if there is no matching result, repeat the process. The query information and the initial response node within the preset period are newly obtained and analyzed; if there is a matching result, step 806 is executed, and the information to be processed is processed using a text clustering algorithm. After processing, the information group to be fed back is obtained, and step 807 is executed to perform intent recognition and classification on the information group to be fed back, wherein a language classification model is used to classify it; after executing step 807, step 808 is executed to determine whether there is a clustering branch after the classification processing is completed, and then step 809 is executed to generate a mapping table. At the same time, step 810 can also be executed to create a preprocessing feedback branch to update the feedback mapping table in step 809.

[0193] At the same time, the user inputs query information in the weighted matching response module, and step 811 is executed based on the process engine to determine whether the response node in the relative technology matches it; if so, step 812 is executed to obtain all response nodes in the system's existing response corpus, and then step 813 is executed to match from all response nodes, and step 814 is executed to determine whether there are response nodes to be screened. If so, step 815 is executed to calculate the response weights of the response nodes to be screened, and the response node with the highest response weight is selected to obtain the matching result, i.e. step 816; if no corresponding response node is found in step 817, step 818 is executed to process the current response node and its query information to obtain the target corpus.

[0194]

[0195] Table 8

[0196] In other embodiments of the present application, after the answering device performs the step of processing the query information and the initial answering node within a preset period to generate a feedback mapping table, that is, after the answering device performs step 306 or step 408, it may further perform the following steps A1 to A2:

[0197] Step A1: The answering device creates a pre-processing feedback branch.

[0198] In other embodiments of the present application, the response device may create a new preprocessing feedback branch based on preset rules; in one feasible implementation, the response device creates a new preprocessing feedback branch based on the obtained query information; in another feasible implementation, the response device creates a new preprocessing feedback branch based on the response information of the initial response node; in the embodiments of the present application, there is no limitation on the information based on which the preprocessing feedback branch is created.

[0199] Step A2: The answering device updates the feedback mapping table to a target feedback mapping table based on the corpus similarity strategy and the pre-processing feedback branch.

[0200] In an embodiment of the present application, after creating a preprocessing branch based on the query information or the response information of the initial response node, the response device adds the preprocessing branch to the feedback mapping table based on the corpus similarity strategy to generate a target feedback mapping table.

[0201] Based on the above embodiments, the embodiment of the present application further provides a response device 9, which can be applied to Figures 2 to 5 In a response method provided in the corresponding embodiment, refer to Figure 9 As shown, the response device 9 may include: an acquisition module 901, a determination module 902 and a response module 903, wherein:

[0202] Acquisition module 901 is used to acquire a response corpus for query information; wherein the query information represents consultation information in multiple rounds of responses;

[0203] A determination module 902 is configured to determine, from the response corpus, a response node to be screened that matches the query information;

[0204] The determination module 902 is configured to obtain a response weight of each to-be-screened response node, and screen the to-be-screened response nodes based on the response weight to determine a target response node;

[0205] The response module 903 is configured to respond to the query information based on the response information of the target response node.

[0206] The acquisition module 901 is further used to obtain the round number of the query information in the multi-round response;

[0207] Obtain the pre-matched response node in the multi-round response;

[0208] If it is determined that there is no to-be-matched answer node matching the query information in the pre-matched answer node, a to-be-screened answer node matching the query information is determined from the answer corpus.

[0209] The determination module 902 is further configured to obtain a distance value between each to-be-screened answer node and the pre-matched answer node;

[0210] Obtain the number of responses of each to-be-screened response node within a preset period;

[0211] Based on the distance value and the number of responses, the response weight of each response node to be screened is determined.

[0212] The determination module 902 is further configured to obtain an initial response node when it is determined from the response corpus that there is no response node to be screened that matches the query information; wherein the initial response node represents the first response node in the multiple rounds of responses;

[0213] Processing the query information and the initial response node within a preset period to generate a feedback mapping table; wherein the feedback mapping table is used to represent the correspondence between the query information and the initial response node;

[0214] Update the response corpus based on the feedback mapping table to obtain the target corpus;

[0215] Respond to the next round of inquiry information based on the target corpus.

[0216] The determination module 902 is further configured to process each initial response node and each query message within a preset period to generate information to be processed;

[0217] Using a text clustering algorithm to process the information to be processed, a group of information to be fed back matching the target number is obtained; wherein the target number represents the number of initial response nodes within a preset period;

[0218] Perform intent recognition and classification on the feedback information group and generate a feedback mapping table.

[0219] The determination module 902 is further configured to perform word segmentation processing on each query information within a preset period and remove stop words to generate a corpus to be processed;

[0220] Perform word segmentation processing on the response information of each initial response node within a preset period and remove stop words to obtain pre-processed information of each initial response node;

[0221] Regular expressions are used to filter the pre-processed information for to-be-processed vocabulary to obtain the business words of each initial response node; wherein the to-be-processed vocabulary represents information related to the item transaction;

[0222] Match the business words with the corpus to be processed to generate the information to be processed corresponding to the initial response node.

[0223] The determination module 902 is further configured to obtain sample information and train the sample information to generate a language classification model;

[0224] Using a language classification model, each piece of information to be fed back in the feedback information group is classified according to business, and the intent of the classification results is identified to obtain the target corpus of each piece of information to be fed back;

[0225] Based on the correspondence between the target corpus and the information to be fed back, and the correspondence between the information to be fed back and the initial response node, a feedback mapping table between the initial response node and the target corpus is generated.

[0226] The determination module 902 is further configured to create a pre-processing feedback branch;

[0227] Based on the corpus similarity strategy and the preprocessing feedback branch, the feedback mapping table is updated to the target feedback mapping table.

[0228] It should be noted that the specific implementation process of the steps performed by each module in this embodiment can be referred to Figures 2 to 5 The implementation process of the response method provided in the corresponding embodiment will not be repeated here.

[0229] The response device provided in the embodiment of the present application determines the response nodes to be screened that match the query information in the response corpus, and then screens the target response nodes that match the query information from the response nodes to be screened based on the response weight, and responds to the query information based on the response information of the target response node, thereby improving the response effect in the multi-round response process and thereby improving the response intelligence.

[0230] Based on the above embodiments, the embodiment of the present application further provides a response device 10, which can be applied to Figures 2 to 5 In a response method provided in the corresponding embodiment, refer to Figure 10 As shown, the answering device 10 may include: a processor 1001, a memory 1002 and a communication bus 1003, wherein:

[0231] The communication bus 1003 is used to implement communication connection between the processor 1001 and the memory 1002.

[0232] The processor 1001 is configured to execute the program of the response method stored in the memory 1002 to implement the following steps:

[0233] Obtaining a response corpus for query information; wherein the query information represents consulting information in multiple rounds of responses;

[0234] Determining a response node to be screened that matches the query information from the response corpus;

[0235] Obtaining the response weight of each to-be-screened response node, and screening the to-be-screened response nodes based on the response weight to determine a target response node;

[0236] The query message is responded to based on the response message of the target response node.

[0237] In other embodiments of the present application, the processor 1001 is configured to execute the steps stored in the memory 1002 to determine, from the response corpus, response nodes to be screened that match the query information, and further implement the following steps:

[0238] Get the round number of the query information in the multi-round response;

[0239] Obtain the pre-matched response node in the multi-round response;

[0240] If it is determined that there is no to-be-matched answer node matching the query information in the pre-matched answer node, a to-be-screened answer node matching the query information is determined from the answer corpus.

[0241] In other embodiments of the present application, the processor 1001 is configured to execute the steps stored in the memory 1002 to obtain the response weight of each to-be-screened response node, and further implement the following steps:

[0242] Obtaining the distance between each to-be-screened answer node and the pre-matched answer node;

[0243] Obtain the number of responses of each to-be-screened response node within a preset period;

[0244] Based on the distance value and the number of responses, the response weight of each response node to be screened is determined.

[0245] In other embodiments of the present application, the processor 1001 is configured to execute the response method stored in the memory 1002, and may further implement the following steps:

[0246] When it is determined from the response corpus that there is no response node to be screened that matches the query information, an initial response node is obtained; wherein the initial response node represents the first response node in multiple rounds of responses;

[0247] Processing the query information and the initial response node within a preset period to generate a feedback mapping table; wherein the feedback mapping table is used to represent the correspondence between the query information and the initial response node;

[0248] Update the response corpus based on the feedback mapping table to obtain the target corpus;

[0249] Respond to the next round of inquiry information based on the target corpus.

[0250] In other embodiments of the present application, the processor 1001 is configured to execute the instructions stored in the memory 1002 to process the query information and the initial response node within a preset period to generate a feedback mapping table, and may further implement the following steps:

[0251] Process each initial response node and each query message within a preset period to generate information to be processed;

[0252] Using a text clustering algorithm to process the information to be processed, a group of information to be fed back matching the target number is obtained; wherein the target number represents the number of initial response nodes within a preset period;

[0253] Perform intent recognition and classification on the feedback information group and generate a feedback mapping table.

[0254] In other embodiments of the present application, the processor 1001 is configured to execute the instructions stored in the memory 1002 to process each initial response node and each query message within a preset period, generate information to be processed, and further implement the following steps:

[0255] Perform word segmentation and stop word removal on each query information within a preset period to generate corpus to be processed;

[0256] Perform word segmentation processing on the response information of each initial response node within a preset period and remove stop words to obtain pre-processed information of each initial response node;

[0257] Regular expressions are used to filter the pre-processed information for to-be-processed vocabulary to obtain the business words of each initial response node; wherein the to-be-processed vocabulary represents information related to the item transaction;

[0258] Match the business words with the corpus to be processed to generate the information to be processed corresponding to the initial response node.

[0259] In other embodiments of the present application, the processor 1001 is configured to execute the intent recognition and classification of the feedback information group stored in the memory 1002 and generate a feedback mapping table, and may further implement the following steps:

[0260] Obtain sample information and train the sample information to generate a language classification model;

[0261] Using a language classification model, each piece of information to be fed back in the feedback information group is classified according to business, and the intent of the classification results is identified to obtain the target corpus of each piece of information to be fed back;

[0262] Based on the correspondence between the target corpus and the information to be fed back, and the correspondence between the information to be fed back and the initial response node, a feedback mapping table between the initial response node and the target corpus is generated.

[0263] In other embodiments of the present application, the processor 1001 is configured to execute the query information and the initial response node stored in the memory 1002 within a preset period to generate a feedback mapping table, and further implement the following steps:

[0264] Create a preprocessing feedback branch;

[0265] Based on the corpus similarity strategy and the preprocessing feedback branch, the feedback mapping table is updated to the target feedback mapping table.

[0266] It should be noted that the specific implementation process of the steps executed by the processor in this embodiment can be referred to Figures 2 to 5 The implementation process of the response method provided in the corresponding embodiment will not be repeated here.

[0267] The response device provided in the embodiment of the present application determines the response nodes to be screened that match the query information in the response corpus, and then screens the target response nodes that match the query information from the response nodes to be screened based on the response weight, and responds to the query information based on the response information of the target response node, thereby improving the response effect in the multi-round response process and thereby improving the response intelligence.

[0268] Based on the above embodiments, the embodiments of the present application provide a computer-readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the following. Figures 2 to 5 The corresponding embodiment provides steps of the response method.

[0269] It should be noted that the above-mentioned computer-readable storage medium can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory (Flash Memory), a magnetic surface storage, an optical disc, or a compact disc read-only memory (CD-ROM); it can also be various answering devices including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0270] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0271] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0272] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0273] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0274] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0275] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0276] The above are only preferred embodiments of the present application and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A response method, characterized in that: The method comprises: Acquire a response corpus for query information; wherein the query information represents consulting information in multiple rounds of responses; Obtaining the round number of the query information in the multiple rounds of responses, and obtaining a pre-matching response node for the round number in the multiple rounds of responses; If it is determined that there is no to-be-matched answer node matching the query information in the pre-matched answer node, determining a to-be-screened answer node matching the query information from the answer corpus; Obtaining a response weight of each of the to-be-screened response nodes, and screening the to-be-screened response nodes based on the response weight to determine a target response node; wherein the response weight is determined based on a distance between each of the to-be-screened response nodes and the pre-matched response node and the number of responses of each of the to-be-screened response nodes within a preset period; The query information is responded to based on the response information of the target response node.

2. The method according to claim 1, characterized in that The obtaining of the response weight of each of the to-be-screened response nodes includes: Obtaining a distance value between each of the to-be-screened answer nodes and the pre-matched answer node; The number of responses of each of the to-be-screened response nodes within a preset period is obtained.

3. The method according to claim 1, characterized in that The method further comprises: When it is determined that there is no to-be-screened answer node matching the query information from the answer corpus, obtaining an initial answer node; wherein the initial answer node represents the first answer node in the multiple rounds of answers; Processing the query information and the initial response node within a preset period to generate a feedback mapping table; wherein the feedback mapping table is used to represent the correspondence between the query information and the initial response node; Updating the response corpus based on the feedback mapping table to obtain a target corpus; The next round of inquiry information is answered based on the target corpus.

4. The method according to claim 3, characterized in that The processing of the query information and the initial response node within a preset period to generate a feedback mapping table includes: Processing each of the initial response nodes and each of the query information within a preset period to generate information to be processed; Processing the information to be processed using a text clustering algorithm to obtain a group of information to be fed back that matches a target number; wherein the target number represents the number of the initial response nodes within a preset period; Perform intent recognition and classification on the information group to be fed back, and generate a feedback mapping table.

5. The method according to claim 4, characterized in that The processing of each initial response node and each query message within a preset period to generate information to be processed includes: Perform word segmentation processing on each query information within a preset period and remove stop words to generate a corpus to be processed; Performing word segmentation processing and stop word removal on the response information of each of the initial response nodes within a preset period to obtain pre-processed information of each of the initial response nodes; Using regular expressions to filter the pre-processed information for to-be-processed vocabulary, and obtaining a business word for each of the initial response nodes; wherein the to-be-processed vocabulary represents information related to item transactions; The business words are matched with the corpus to be processed to generate the information to be processed corresponding to the initial response node.

6. The method according to claim 4, characterized in that The performing intention recognition and classification on the information group to be fed back and generating a feedback mapping table includes: Obtaining sample information, and training the sample information to generate a language classification model; Using the language classification model, classify each piece of information to be fed back in the group of information to be fed back according to business, and perform intent recognition on the classification results to obtain a target corpus for each piece of information to be fed back; Based on the corresponding relationship between the target corpus and the information to be fed back, and the corresponding relationship between the information to be fed back and the initial response node, a feedback mapping table between the initial response node and the target corpus is generated.

7. The method according to claim 3, characterized in that After processing the query information and the initial response node within a preset period to generate a feedback mapping table, the method further includes: Create a preprocessing feedback branch; Based on the corpus similarity strategy and the preprocessing feedback branch, the feedback mapping table is updated to a target feedback mapping table.

8. A response device, characterized in that: The response device includes: an acquisition module, a determination module and a response module; The acquisition module is used to acquire a response corpus for the inquiry information; wherein the inquiry information represents the consultation information in multiple rounds of responses; The determining module is configured to obtain the round number of the query information in the multiple rounds of responses, and obtain pre-matching response nodes with the round number in the multiple rounds of responses; if it is determined that there is no to-be-matched response node matching the query information among the pre-matching response nodes, determine a to-be-screened response node matching the query information from the response corpus; The determination module is further configured to obtain a response weight of each of the to-be-screened response nodes, and screen the to-be-screened response nodes based on the response weight to determine a target response node; wherein the response weight is determined based on a distance between each of the to-be-screened response nodes and the pre-matched response node and a number of responses by each of the to-be-screened response nodes within a preset period; The response module is configured to respond to the inquiry information based on the response information of the target response node.

9. An answering device, characterized in that: The answering device includes: a processor, a memory and a communication bus; The communication bus is used to realize the communication connection between the processor and the memory; The processor is configured to execute the program in the memory to implement the response method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the response method according to any one of claims 1 to 7.

Images