Customer service standard operation process mining method, intelligent dialogue method and electronic equipment

Abstract

The embodiment of the invention provides a customer service standard operation process mining method, an intelligent dialogue method, electronic equipment, a storage medium and a computer program product. The customer service standard operation process mining method comprises the steps of performing information extraction and aggregation processing on service trajectory data of historical customer service dialogues to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenes; according to the time sequence of customer service operation in the service track data, frequent customer service operation path sequence mining is carried out on the service track data in the set, and target service paths associated with various consultation questions in different customer service scenes are obtained; and based on the target service path, generating customer service standard operation process trees corresponding to various consultation questions in different customer service scenes. According to the method, the acquisition efficiency of the customer service standard operation process is improved.

Description

(19) State Intellectual Property Office (12) Invention Patent Application (10) Application Publication Number (43) Application Publication Date (21) Application Number 202511300460.0 (22) Application Date 2025.09.11 (71) Applicant Hangzhou Alibaba Overseas Internet Industry Co., Ltd. Address 303, 3rd Floor, Building 5, No. 699, Wangshang Road, Changhe Street, Binjiang District, Hangzhou City, Zhejiang Province, 310051 (72) Inventor Wei Zhen (74) Patent Agency Beijing Junyixin Intellectual Property Agency Co., Ltd. 11789 Patent Attorney Qian Xiuru (51) Int.Cl. G06F 16 / 3329 (2025.01) G06F 16 / 334 (2025.01) G06F 16 / 338 (2019.01) G06Q 30 / 015 (2023.01) (54) Invention Title: Customer Service Standard Operating Procedure Mining Method, Intelligent Dialogue Method, Electronic Device (57) Abstract: This application provides a customer service standard operating procedure mining method, an intelligent dialogue method, an electronic device, a storage medium, and a computer program product. The customer service standard operating procedure mining method includes: extracting and aggregating information from historical customer service dialogue service trajectory data to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios; mining frequent customer service operation path sequences in the service trajectory data according to the time order of customer service operations in the service trajectory data to obtain target service paths associated with various consultation questions in different customer service scenarios; and generating customer service standard operating procedure trees corresponding to various consultation questions in different customer service scenarios based on the target service paths. This method improves the efficiency of obtaining customer service standard operating procedures. Claims 3 pages, Description 16 pages, Drawings 4 pages, CN 121434336 A 2026.01.30 CN 1 21 43 43 36 A 1. A method for mining standard customer service operation processes, characterized in that the method includes: extracting and aggregating information from service trajectory data of historical customer service dialogues to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios; mining frequent customer service operation path sequences from the service trajectory data in the set according to the time order of customer service operations in the service trajectory data to obtain target service paths associated with various consultation questions in different customer service scenarios; generating a standard customer service operation process tree corresponding to various consultation questions in different customer service scenarios based on the target service paths. 2. The method according to claim 1, characterized in that the extraction and aggregation of information from service trajectory data of historical customer service dialogues to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios includes:The method involves summarizing and extracting information from historical customer service dialogue service trajectory data to obtain clustering information for the service trajectory data. This clustering information includes the following: scenario information, dialogue identifiers, and consultation questions. Multi-stage aggregation is performed on the clustering information to obtain a set of dialogue identifiers corresponding to consultation questions in different customer service scenarios. Based on the set of dialogue identifiers corresponding to consultation questions in different customer service scenarios, and the service trajectory data, a set of service trajectory data corresponding to consultation questions in different customer service scenarios is obtained. 3. The method according to claim 2, wherein the multi-stage aggregation of the clustering information to obtain a set of dialogue identifiers corresponding to consultation questions in different customer service scenarios includes: Based on the scenario information, aggregation processing is performed on the consultation questions and dialogue identifiers in the clustering information to obtain a first set of consultation questions and dialogue identifiers in different customer service scenarios; Aggregation processing is performed on the consultation questions in the first set to obtain a set of dialogue identifiers corresponding to consultation questions in different customer service scenarios. 4. The method according to claim 3, characterized in that, the step of aggregating the consultation questions in the first set to obtain a set of dialogue identifiers corresponding to the consultation questions in different customer service scenarios includes: performing text vectorization processing on the consultation questions in the first set to obtain text vectors for each consultation question; calculating the similarity between consultation questions in the same first set based on the text vectors; aggregating consultation questions whose similarity satisfies a preset similarity condition into the same type of consultation questions, and constructing a set of dialogue identifiers corresponding to the corresponding type of consultation questions in the customer service scenario corresponding to the first set based on the dialogue identifiers corresponding to the consultation questions aggregating into the same type of consultation questions. 5. The method according to claim 1, characterized in that, according to the time sequence of customer service operations in the service trajectory data, the step of performing frequent customer service operation path sequence mining on the service trajectory data in the set to obtain target service paths associated with various types of consultation questions under different customer service scenarios includes: extracting information from the service trajectory data in the set according to the time sequence of customer service operations in the service trajectory data to obtain service flow lines corresponding to various types of consultation questions under different customer service scenarios, wherein the service nodes in the service flow lines correspond to the customer service operations, and the paths in the service flow lines represent the upstream and downstream relationships of customer service operations; using the service nodes as graph nodes, constructing directed edges connecting the corresponding graph nodes according to the upstream and downstream relationships represented by the paths, setting the weights of the corresponding directed edges according to the frequency of occurrence of the paths in the service flow lines, and obtaining a directed acyclic graph as the service flow line graph corresponding to various types of consultation questions under different customer service scenarios; Claims 1 / 3 page 2 CN 121434336 ABased on the service flow graph, frequent customer service operation path sequence mining is performed to obtain target service paths associated with various types of consultation questions under different customer service scenarios. 6. The method according to claim 5, characterized in that, the step of mining frequent customer service operation path sequences based on the service flow graph to obtain target service paths associated with various types of consultation questions under different customer service scenarios includes: performing relaxation operations on the outgoing edges of the nodes in the service flow graph in topological order to search for target service paths associated with various types of consultation questions under different customer service scenarios, wherein the sum of the weights of each directed edge in the target service path is greater than the sum of the weights of each directed edge in the known service paths. 7. An intelligent dialogue method, characterized in that the method includes: after receiving dialogue text input by a user during an inbound customer service dialogue, extracting information from the dialogue text to obtain scenario information and consultation questions matching the dialogue text; obtaining a pre-stored standard customer service operation process tree corresponding to the scenario information and the consultation questions; inputting the dialogue text, the standard customer service operation process tree, tool information of preset operation tools, and / or preset query capability information into a pre-fine-tuned generative large model to guide the generative large model to generate operation results for the current customer service operation steps in the inbound customer service dialogue; and generating the current intelligent reply text for the inbound customer service dialogue based on the operation results. 8. The method according to claim 7, wherein the customer service standard operation flow tree is used to describe one or more service paths for a specified consultation question, the service path includes service nodes corresponding to customer service operation steps, and the step of inputting the dialogue text, the customer service standard operation flow tree, the tool information of the preset operation tool and / or the preset query capability information into the pre-fine-tuned generative large model, and guiding the generative large model to generate the operation result of the current customer service operation step for the inbound customer service dialogue, includes: inputting the dialogue text, the customer service standard operation flow tree, the tool information of the preset operation tool and / or the preset query capability information into the pre-fine-tuned generative large model, guiding the generative large model to generate the current customer service operation step to be executed using the customer service standard operation flow tree as a constraint, and executing the current customer service operation step to be executed based on the tool information of the preset operation tool and / or the preset query capability information to obtain the operation result of the current customer service operation step for the inbound customer service dialogue. 9. The method according to claim 7, characterized in that, before extracting information from the dialogue text input by the user during the incoming customer service dialogue and obtaining the scenario information and consultation questions matched by the dialogue text, it further includes: performing knowledge accumulation on the service trajectory data of historical customer service dialogues to obtain customer service standard operation process trees corresponding to various consultation questions under different customer service scenarios;The standard customer service operation process tree is associated and stored with the scenario information and the consultation question. 10. The method according to claim 9, wherein the step of knowledge accumulation on the service trajectory data of historical customer service dialogues to obtain the standard customer service operation process tree corresponding to various consultation questions under different customer service scenarios includes: extracting and aggregating information from the service trajectory data of historical customer service dialogues to obtain a set of service trajectory data corresponding to consultation questions under different customer service scenarios; mining frequent customer service operation path sequences on the service trajectory data in the set according to the time order of customer service operations in the service trajectory data to obtain the target service paths associated with various consultation questions under different customer service scenarios; generating the standard customer service operation process tree corresponding to various consultation questions under different customer service scenarios based on the target service paths. 11. The method according to claim 7, wherein after generating the current intelligent reply text of the incoming customer service dialogue based on the operation result, it further includes: obtaining evaluation data of the intelligent reply text of the incoming customer service dialogue; optimizing the standard customer service operation process tree based on the evaluation data. 12. An electronic device, characterized in that it comprises: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of claims 1-11. 13. A computer-readable storage medium, characterized in that the computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-11. 14. A computer program product, comprising a computer program / computer-executable instructions, characterized in that, when the computer program / computer-executable instructions are executed by a processor in an electronic device, they implement the method as described in any one of claims 1-11. Claims 3 / 3 Page 4 CN 121434336 A Customer Service Standard Operation Process Mining Method, Intelligent Dialogue Method, Electronic Device Technical Field

[0001] This application relates to the field of intelligent customer service technology, and in particular to a customer service standard operation process mining method, an intelligent dialogue method, an electronic device, a storage medium, and a computer program product. Background Art

[0002] In customer service standard operating procedure mining applications, the service SOP (Standard Operating Procedure) refers to the standard service handling process for different customer inquiries. In the prior art, SOPs are generally...Most of the configuration is done manually by customer service personnel, resulting in low overall efficiency and unreliable accuracy and timeliness. Furthermore, the overall SOP cannot be optimized and adjusted in a timely manner after being arranged, leading to situations where the customer service standard operating procedures cannot keep up with product updates and iterations in terms of problem-solving effectiveness.

[0003] It is evident that the existing methods for mining customer service standard operating procedures still need improvement. Summary of the Invention

[0004] This application provides a method for mining customer service standard operating procedures, which can effectively improve the efficiency of obtaining customer service standard operating procedures and is more timely than manually configuring operating procedures.

[0005] Correspondingly, this application also provides a method for mining customer service standard operating procedures, an intelligent dialogue method, an electronic device, a storage medium, and a computer program product to ensure the implementation and application of the above-mentioned method for mining customer service standard operating procedures.

[0006] To address the above problems, this application discloses a method for mining standard customer service operation processes. The method includes:

[0007] Extracting and aggregating information from historical customer service dialogue service trajectory data to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios;

[0008] Mining frequent customer service operation path sequences from the service trajectory data in the set according to the time order of customer service operations in the service trajectory data to obtain target service paths associated with various consultation questions in different customer service scenarios;

[0009] Generating standard customer service operation process trees corresponding to various consultation questions in different customer service scenarios based on the target service paths.

[0010] To solve the above problems, this application discloses an intelligent dialogue method, including:

[0011] After receiving dialogue text input by a user during an inbound customer service dialogue, extracting information from the dialogue text to obtain scenario information and consultation questions matching the dialogue text;

[0012] Obtaining a pre-stored customer service standard operation process tree corresponding to the scenario information and the consultation questions;

[0013] Inputting the dialogue text, the customer service standard operation process tree, tool information of preset operation tools, and / or preset query capability information into a pre-fine-tuned generative large model to guide the generative large model to generate operation results for the current customer service operation steps of the inbound customer service dialogue;

[0014] Generating the current intelligent reply text of the inbound customer service dialogue based on the operation results.

[0015] This application also discloses an electronic device, including: a processor, and a memory connected to the processor (see page 1 / 16 of CN 121434336 A);

[0016] the memory stores computer execution instructions;

[0017] the processor executes the computer execution instructions stored in the memory to implement the method described in this application embodiment.

[0018] This application also discloses a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method described in this application.

[0019] This application also discloses a computer program product, including a computer program / computer-executable instructions, which, when executed by a processor in an electronic device, implement the method described in this application.

[0020] Compared with the prior art, this application has the following advantages:

[0021] By extracting and aggregating information from historical customer service dialogue service trajectory data, a set of service trajectory data corresponding to consultation questions under different customer service scenarios is obtained; then, according to the time sequence of customer service operations in the service trajectory data, frequent customer service operation path sequence mining is performed on the service trajectory data in the set to obtain target service paths associated with various consultation questions under different customer service scenarios; and based on the target service paths, a customer service standard operation process tree corresponding to various consultation questions under different customer service scenarios is generated, which effectively improves the efficiency of obtaining customer service standard operation processes; by transforming real-time service operations into knowledge accumulation, the dynamic evolution of application rules is supported, and the generated customer service standard operation process is more reliable and more practical.

[0022] Figure 1 is a flowchart of the steps of the customer service standard operation process mining method disclosed in the embodiments of this application;

[0023] Figure 2 is a schematic diagram of the service flow generated by the customer service standard operation process mining method disclosed in the embodiments of this application;

[0024] Figure 3 is a schematic diagram of the service flow generated by the customer service standard operation process mining method disclosed in the embodiments of this application;

[0025] Figure 4 is a schematic diagram of the structure of the customer service standard operation process tree generated by the customer service standard operation process mining method disclosed in the embodiments of this application;

[0026] Figure 5 is one of the flowcharts of the intelligent dialogue method disclosed in the embodiments of this application;

[0027] Figure 6 is another flowchart of the intelligent dialogue method disclosed in the embodiments of this application;

[0028] Figure 7 is a schematic diagram of the structure of an exemplary device provided in an embodiment of this application. Detailed Description

[0029] To make the above-mentioned objects, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] The customer service standard operating procedure mining method disclosed in this application, through mining customer service assistant service flow data (including service operation path, specific operation content, service response content, service evaluation, etc.), trains the data, classifies it according to problem type, and precipitates the customer service standard operating procedure (Standard Operating Procedure).The Standard Operating Procedure (SOP) is used to guide intelligent customer service. The standard operating procedure includes, but is not limited to, a series of operations such as replying with scripts, selecting and calling tools, querying the knowledge base, creating work orders, and work order transfer. Through a large model's understanding of the standard operating procedure, and through a thought chain under constraints, intelligent customer service is achieved, further improving the intelligent service solution effect.

[0031] As shown in Figure 1, the customer service standard operating procedure mining method disclosed in this application includes steps 102 to 106.

[0032] Step 102: Information extraction and aggregation processing is performed on the service trajectory data of historical customer service dialogues to obtain a set of service trajectory data corresponding to the consultation questions under different customer service scenarios.

[0033] Wherein, the historical customer service dialogues are positive dialogues selected based on user evaluations.

[0034] In some optional embodiments, service trajectory data of historical customer service dialogues is collected by marking the service trajectory data of inbound customer service dialogues on the consultation reception platform. In some alternative embodiments, a brain-computer interface can be added to the consultation reception platform to collect stress response data of customer service personnel during inbound customer service dialogues and convert it into service trajectory data. In the embodiments of this application, the specific implementation of obtaining service trajectory data from historical customer service dialogues is not limited.

[0035] Wherein, the service trajectory data is used to describe the customer service personnel's information during the service process, including but not limited to: tool call parameters, tool execution results, service content used, knowledge information used, operation type performed (such as transfer, escalation), and reply content, etc.

[0036] Optionally, the service trajectory data includes but is not limited to one or more of the following: operation time, dialogue identifier, customer consultation context, service operation type, operation input parameters, operation output parameters, reply content, etc. Wherein, the operation time refers to the timestamp of the customer service representative's page operation; the dialogue identifier is used to uniquely identify an incoming customer service dialogue; the customer consultation context is used to describe the content of the customer's incoming consultation; the service operation type is used to describe the type of operation action for the incoming customer service dialogue, including but not limited to one or more of the following: viewing historical incoming records, asking questions in return, querying the knowledge base, selecting and calling tools, creating work orders, transferring work orders, replying to consultations / soothing emotions, etc.; the operation input parameters are used to describe the relevant parameters for performing the service operation, including tool query parameters, knowledge base query parameters, etc.; the operation output parameters are used to describe the results obtained after performing the service action, including knowledge details, tool results, etc.; the reply content is used to describe the reply to the customer's consultation questions after performing a series of related service actions.

[0037] Optionally, the service trajectory data of historical customer service dialogues is processed by information extraction and aggregation to obtain notThe set of service trajectory data corresponding to the consultation question in the same customer service scenario includes: summarizing and extracting information from the service trajectory data of historical customer service dialogues to obtain clustering information of the service trajectory data. The clustering information includes the following information: scenario information, dialogue identifier, and consultation question; performing multi-stage aggregation on the clustering information to obtain a set of dialogue identifiers corresponding to the consultation question in different customer service scenarios; and obtaining a set of service trajectory data corresponding to the consultation question in different customer service scenarios based on the set of dialogue identifiers corresponding to the consultation question in different customer service scenarios and the service trajectory data.

[0038] Optionally, the large model in the prior art can be used to summarize and extract information from the service trajectory data of historical customer service dialogues to obtain the scenario information, dialogue identifier, and consultation question associated with the service trajectory data of each customer service call. The scenario information, dialogue identifier, and consultation question associated with the service trajectory data are recorded as the clustering information of the service trajectory data. The scenario information and consultation question are determined in combination with the scenario of the specific customer service dialogue application. The scenario information includes, but is not limited to: question type and / or consultation scenario. For example, candidate question types, candidate consultation scenarios, and candidate consultation questions can be given. Then, the large model is prompted to select the question type, consultation scenario, and consultation question that matches each historical customer service dialogue from the given candidate question types, candidate consultation scenarios, and candidate consultation questions.

[0039] The service trajectory data of historical customer service dialogues is summarized and information is extracted to obtain the clustering information of the service trajectory data. See the prior art for specific implementation methods, which will not be repeated in the embodiments of this application.

[0040] Using the above method, a clustering information associated with the service trajectory data of each historical customer service dialogue can be extracted. Next, based on the clustering information, the service trajectory data is aggregated to obtain a set of dialogue identifiers of the incoming customer service dialogues associated with each type of consultation question under the specified question type and consultation scenario.

[0041] Optionally, the step of performing multi-stage aggregation on the clustering information to obtain a set of dialogue identifiers corresponding to consultation questions in different customer service scenarios includes: based on the scenario information, performing aggregation processing on the consultation questions and dialogue identifiers in the clustering information to obtain a first set of consultation questions and dialogue identifiers in different customer service scenarios; and performing aggregation processing on the consultation questions in the first set to obtain a set of dialogue identifiers corresponding to consultation questions in different customer service scenarios. The scenario information can be selected from high-evaluation customer service scenarios.

[0042] In the first stage, using question type and consultation scenario as two dimensions as aggregation conditions, the clustering information...The consultation questions and the dialogue identifiers are aggregated to group the consultation questions and dialogue identifiers in each cluster of information with the same question type and consultation scenario into the same set, which is denoted as "first set" in the embodiments of this application. The information in each first set can be represented as: [question type i, consultation scenario j, [{dialogue identifier l, consultation question l}, {dialogue identifier 2, consultation question 2}, ...]], where i and j are the numbers of question type and consultation scenario, respectively.

[0043] After performing the first stage of aggregation, the set of consultation questions and dialogue identifiers under each combination of question type and consultation scenario can be obtained.

[0044] In the second stage, the consultation questions in each first set are further aggregated to obtain the set of dialogue identifiers corresponding to each consultation question under different customer service scenarios (such as different combinations of question type and consultation scenario). Usually, each customer service scenario includes multiple consultation questions. In the second stage, the consultation questions in each first set are aggregated to obtain the set of dialogue identifiers corresponding to each consultation question under different customer service scenarios.

[0045] Optionally, the step of aggregating the consultation questions in the first set to obtain a set of dialogue identifiers corresponding to the consultation questions in different customer service scenarios includes: performing text vectorization processing on the consultation questions in the first set to obtain text vectors for each consultation question; calculating the similarity between consultation questions in the same first set based on the text vectors; aggregating consultation questions whose similarity satisfies a preset similarity condition into the same type of consultation questions, and constructing a set of dialogue identifiers corresponding to the corresponding type of consultation questions in the customer service scenario corresponding to the first set based on the dialogue identifiers corresponding to the consultation questions aggregating into the same type of consultation questions.

[0046] For example, after completing the customer service scenario aggregation, the Word2Vec method is used to perform text vectorization processing on the consultation questions in each first set to obtain text vectors for each consultation question; then, for each first set, the cosine similarity between the consultation questions in the first set is calculated, and the consultation questions in the first set are clustered based on the calculated cosine similarity, aggregating similar consultation questions into one type of consultation questions, thereby dividing the consultation questions in each first set into multiple types of consultation questions. Taking a cosine similarity value range of [0,1] as an example, for the first set A, consultation questions with a cosine similarity greater than 0.9 in the first set A can be classified into one category of consultation questions, that is, similar consultation questions in the same first set are regarded as one category of consultation questions, thus obtaining the multiple categories of consultation questions included in the first set A.

[0047] Next, each category of consultation questions in the first set is recorded as a target consultation question, and each category of consultation questions is further classified into target consultation questions.The dialogue identifiers corresponding to the questions are aggregated into a set, which is the set of dialogue identifiers corresponding to this type of consultation question. This type of consultation question is the set of dialogue identifiers corresponding to the next type of consultation question in the customer service scenario corresponding to the first set. (See specification 4 / 16, page 8, CN 121434336 A). For example, after multi-stage clustering, the clustering information of a certain customer service scenario can obtain a set of dialogue identifiers similar to the following: Set 1: [Question type 1, Consultation scenario 1, Consultation question 1, [Dialogue identifier 1, Dialogue identifier 2, ...]]; Set 2: [Question type 1, Consultation scenario 1, Consultation question 2, [Dialogue identifier 3, Dialogue identifier 4, ...]]; Set 3: [Question type 1, Consultation scenario 2, Consultation question 3, [Dialogue identifier 5, Dialogue identifier 6, ...]]. Wherein, Question type 1 and Consultation scenario 1 correspond to one customer service scenario, and Question type 1 and Consultation scenario 2 correspond to another consultation scenario.

[0048] After obtaining the set of dialogue identifiers corresponding to the consultation questions in different customer service scenarios, service trajectory data of corresponding historical customer service dialogues is obtained from the service trajectory data according to the dialogue identifiers in the set of dialogue identifiers corresponding to the consultation questions in different customer service scenarios (such as the above set 1, set 2 and set 3), and a set of service trajectory data corresponding to the consultation questions in different customer service scenarios is constructed. For example, the service trajectory data of historical customer service dialogues identified by dialogue identifier 1 and dialogue identifier 2 are obtained and added to the set of service trajectory data of consultation question 1 under question type 1 and consultation scenario 1.

[0049] Step 104: According to the time order of customer service operations in the service trajectory data, frequent customer service operation path sequence mining is performed on the service trajectory data in the set to obtain the target service paths associated with various consultation questions in different customer service scenarios.

[0050] Wherein, the target service path is the high-frequency service path extracted from the service trajectory data.

[0051] As mentioned above, each set of service trajectory data corresponds to a type of consultation question in a specified scenario. Next, for each set, each service trajectory data in the set is analyzed to obtain a service path, thereby obtaining multiple service paths for a type of consultation question in a specified scenario. Then, based on the multiple service paths for each type of consultation question, the target service path for that type of question is mined.

[0052] Optionally, the step of mining frequent customer service operation path sequences in the service trajectory data of the set according to the time order of customer service operations in the service trajectory data to obtain the target service paths associated with various types of consultation questions in different customer service scenarios includes: extracting information from the service trajectory data of the set according to the time order of customer service operations in the service trajectory data to obtain the service paths corresponding to various types of consultation questions in different customer service scenarios.The service nodes in the service flow correspond to the customer service operations, and the paths in the service flow represent the upstream and downstream relationships of the customer service operations. Using the service nodes as graph nodes, directed edges connecting the corresponding graph nodes are constructed according to the upstream and downstream relationships represented by the paths. The weights of the corresponding directed edges are set according to the frequency of the paths in the service flow to obtain a directed acyclic graph, which serves as the service flow graph corresponding to various types of consultation questions in different customer service scenarios. Based on the service flow graph, frequent customer service operation path sequence mining is performed to obtain the target service paths associated with various types of consultation questions in different customer service scenarios.

[0053] For each service trajectory data, customer service operations of a specified service operation type are extracted as service nodes, and the upstream and downstream relationships of customer service operations are determined according to the time sequence of customer service operations, thereby obtaining the service flow corresponding to each service trajectory data. For a specific example, for the service trajectory data corresponding to dialogue identifier 1001, the operation time of the customer service operation of "consultation and counter-question" is 09:00, the operation time of the customer service operation of "knowledge base query" is 09:01, the operation time of the customer service operation of "tool call" is 09:03, and the operation time of the customer service operation of "consultation reply" is 09:07, thus obtaining the service flow shown in Figure 2. In the service flow obtained based on the service trajectory data corresponding to dialogue identifier 1001, the service nodes include: service node 1 corresponding to the customer service operation "consultation and counter-question", service node 2 corresponding to the customer service operation "knowledge base query", service node 3 corresponding to the customer service operation "tool call", and service node 4 corresponding to the customer service operation "consultation reply"; the paths include: path 1 from service node 1 to service node 2, path 2 from service node 2 to service node 3, and path 3 from service node 3 to service node 4.

[0054] After extracting information from the set of specified service trajectory data using the above method, several service paths can be obtained. Next, for a specified set of service trajectory data, the service nodes in the service paths obtained from the service trajectory data in the set are used as the graph nodes of the service path graph to be generated. The directed edges of the service path graph to be generated are constructed based on the paths (i.e., the upstream and downstream relationships of the service nodes) in the service paths obtained from the service trajectory data in the set, resulting in a directed acyclic graph. On the other hand, for each directed edge, the weight of the directed edge is set according to the number of times the path corresponding to the directed edge appears in the service paths obtained from the set. Finally, the directed acyclic graph with weighted directed edges is used as the service path graph corresponding to the set. The service path graph obtained in this embodiment is shown in Figure 3.

[0055] In the service flow diagram shown in Figure 3, the arrows point to downstream service nodes. As can be seen from Figure 3, after a user initiates a consultation, they can perform various types of customer service operations. During the service process, they go through different operation procedures and finally execute a consultation reply. The service flow diagram shown in Figure 3 is obtained based on the analysis of historical service trajectory data. Some operation paths are short, some are long, and some paths (i.e., the upstream and downstream relationships of customer service operations) appear frequently in historical service trajectory data. In order to improve customer service efficiency and quality, in the embodiments of this application, frequent customer service operation path sequence mining is further performed based on the service flow diagram to obtain target service paths associated with various types of consultation questions, which are used to generate a standard customer service operation process tree.

[0056] Optionally, the frequent customer service operation path sequence mining based on the service flow diagram to obtain target service paths associated with various types of consultation questions under different customer service scenarios includes: performing relaxation operations on the outgoing edges of the nodes in the service flow diagram in the topological order to search for target service paths associated with various types of consultation questions under different customer service scenarios, wherein the sum of the weights of each directed edge in the target service path is greater than the sum of the weights of each directed edge in the known service path.

[0057] For the current service flow diagram, firstly, obtain the set of graph nodes and the set of directed edges in the service flow diagram. Taking the set of graph nodes as V and the set of directed edges as E as an example, the set V contains the entry node S and the exit node T of the service flow diagram, and each directed edge in the set E is represented as (v, u), that is, the directed edge from graph node v to graph node u, (v, u) ∈ E, and the weight of edge (v, u) is represented as W(v, u).

[0058] The following describes the process of performing relaxation operations on the outgoing edges of the nodes in the service flow diagram in the topological order to search for the target service paths associated with various consultation questions in different customer service scenarios.

[0059] First, use dynamic programming to initialize the weight table of each path in the service flow diagram, and initialize the predecessor node record table. Among them, the weight of the path is the sum of the maximum weights of each directed edge in the complete service flow from the entry graph node to the exit graph node. In specific implementation, an array dist[v] can be defined to represent the maximum sum of the directed edges between the entry node S and the exit node T. The initial weight dist[S] of the entry node is initialized to 0, and the values ​​of dist[v] (v≠S) of other nodes are initialized to negative infinity, indicating that the nodes are unreachable. On the other hand, an array prev[v] is defined to record the predecessor nodes of node v, which are queried when backtracking the path.

[0060] Next, the outgoing edges of the nodes in the service path graph are relaxed according to the topological order, and the weight table and the predecessor node record table are updated according to the relaxation results. By comparing the weight of the currently known path with the weight of the newly discovered path, the weight of the path is found.For the path with the larger weight, update the preferred path of the node. For example, process each pair of nodes v and u in the service flow graph in the topological order, where u is the node pointed to by the outgoing edge of node v. For each outgoing edge (u, v) ∈ E of u, perform a relaxation operation: if dist[v] < dist[u] + w(u, v), then dist[v] ← dist[u] + w(u, v), prev[u] ← u. That is, if the maximum sum of the weights of the directed edges from the entry node to node u to node v is less than the sum of the weights of the directed edges from the entry node to node u and the sum of the weights of the directed edges between nodes u and v, update the maximum sum of the weights of the directed edges from the entry node to node u to the sum of the weights of the directed edges from the entry node to node u and the sum of the weights of the directed edges between nodes u and v, and record node u as the predecessor node of node v. Finally, dist[T] of the exit node T is the maximum weight sum from the entry node to the exit node, which can be used as the weight of the newly discovered path.

[0061] Finally, according to the predecessor node record table, the target service path is obtained, and the weight corresponding to the target service path is obtained through the weight table.

[0062] For example, by relaxing the outgoing edges of the nodes in the service flow graph, the predecessor node record table is obtained. According to the predecessor node record table, the complete path can be obtained by backtracking back to the entry node through prev[T]. Correspondingly, the weight table records the maximum sum of the weights of the directed edges of the preferred path after each updated node (corresponding to the target service path obtained by backtracking), which can be used as the weight of the corresponding target service path.

[0063] According to the above method, frequent customer service operation path sequence mining is performed based on the service flow graph of each type of consultation question in each customer service scenario to obtain the target service paths associated with each type of consultation question in different customer service scenarios.

[0064] In general, each customer service scenario includes multiple types of consultation questions. Next, for each customer service scenario, the target service path associated with each type of consultation question in that customer service scenario is processed separately, thereby generating a customer service standard operation process tree corresponding to each type of consultation question in that customer service scenario.

[0065] Step 106: Based on the target service path, generate customer service standard operation process trees corresponding to each type of consultation question in different customer service scenarios.

[0066] Optionally, generating customer service standard operation process trees corresponding to each type of consultation question in different customer service scenarios based on the target service path includes: merging the starting node of the target service path to obtain the root node of the customer service standard operation process tree; creating each level node of the root node according to the target service path; merging child nodes with the same parent node in the same level node to obtain the customer service standard operation process tree.

[0067] According to the method of mining target service paths, each target service path starts from the entry node and ends at the exit node. By merging the entry nodes of each target service path into one entry node, which serves as the root node of the customer service standard operation process tree, multiple target service paths are constructed into an operation process tree, with each target service path serving as a subtree of the operation process tree. Next, starting from the root node, the node hierarchy of the operation process tree can be operated. Starting from the root node, the nodes on each subtree are traversed layer by layer. For nodes at the same level that correspond to the same customer service operation and have the same parent node, a node merging operation is performed to standardize the operation process tree and obtain the customer service standard operation process tree.

[0068] Taking the customer service standard operation process tree shown in Figure 4 as an example, the customer service standard operation process tree describes the customer service operation steps to be executed sequentially to solve a type of consultation problem in a certain customer service scenario.

[0069] Afterwards, the mapping relationship between the relevant consultation problems of different customer service scenarios and the customer service standard operation process tree is stored for subsequent intelligent customer service. For example, the mapping relationship can be stored in the following forms: (Question Type 1, Consultation Scenario 1, Consultation Question 1, Customer Service Standard Operation Flow Tree 1), (Question Type 1, Consultation Scenario 2, Consultation Question 2, Customer Service Standard Operation Flow Tree 2), (Question Type 2, Consultation Scenario 1, Consultation Question 6, Customer Service Standard Operation Flow Tree 3), etc.

[0070] In some optional embodiments, after generating the customer service standard operation flow tree for consultation questions, the customer service standard operation flow tree can be manually verified to further improve the accuracy of the customer service standard operation flow.

[0071] In summary, the customer service standard operation process mining method disclosed in this application extracts and aggregates information from historical customer service dialogue service trajectory data to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios; then, according to the time order of customer service operations in the service trajectory data, frequent customer service operation path sequence mining is performed on the service trajectory data in the set to obtain target service paths associated with various consultation questions in different customer service scenarios; and based on the target service paths, a customer service standard operation process tree corresponding to various consultation questions in different customer service scenarios is generated, which effectively improves the efficiency of obtaining customer service standard operation processes. Furthermore, this method transforms real-time service operations into knowledge accumulation, supports the dynamic evolution of application rules, and generates customer service standard operation processes with higher reliability and stronger practicality.

[0072] Based on the foregoing embodiments, this application also discloses an intelligent dialogue method, as shown in Figure 5, the method includes: steps 502 to 508.

[0073] Step 502: After receiving the dialogue text input by the user during the incoming customer service conversation, process the dialogue text...Information extraction is performed to obtain the scenario information and consultation questions matching the dialogue text.

[0074] Optionally, the intelligent conversation is a dialogue between the user of the consultation reception platform and the customer service robot. The consultation reception platform engages in dialogue with the user by creating an inbound customer service dialogue to answer the user's questions. After the user inputs the dialogue text through the inbound customer service dialogue window, the client of the consultation reception platform obtains the dialogue text and sends it to the server. The server extracts information from the dialogue text to obtain the scenario information and consultation questions matching the dialogue text. The scenario information and consultation questions are determined in combination with the specific scenario of the customer service dialogue application. The scenario information is used to describe the customer service scenario. The scenario information includes, but is not limited to: question type and / or consultation scenario. For example, the generative large model in the prior art can be used to extract information from the dialogue text to obtain the scenario information and consultation questions matching the dialogue text.

[0075] For the specific implementation of extracting information from the dialogue text to obtain the customer service scenario and consultation questions matching the dialogue text, please refer to the relevant description in the previous embodiments, which will not be repeated here.

[0076] Step 504: Obtain the pre-stored customer service standard operation process tree corresponding to the scenario information and the consultation question.

[0077] Optionally, after receiving the dialogue text input by the user during the inbound customer service dialogue, before extracting information from the dialogue text and obtaining the scenario information and consultation question matching the dialogue text, as shown in Figure 6, the method further includes: steps 500 and 501.

[0078] Step 500: Perform knowledge accumulation on the service trajectory data of historical customer service dialogues to obtain the customer service standard operation process tree corresponding to various consultation questions under different customer service scenarios.

[0079] Optionally, knowledge accumulation is performed on the service trajectory data of historical customer service dialogues to obtain customer service standard operation process trees corresponding to various consultation questions in different customer service scenarios. This includes: extracting and aggregating information from the service trajectory data of historical customer service dialogues to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios; according to the time order of customer service operations in the service trajectory data, frequent customer service operation path sequence mining is performed on the service trajectory data in the set to obtain target service paths associated with various consultation questions in different customer service scenarios; based on the target service paths, customer service standard operation process trees corresponding to various consultation questions in different customer service scenarios are generated. The customer service scenario is described by scenario information, which includes: question type and / or consultation scenario.

[0080] For the specific implementation methods of each step of knowledge accumulation on the service trajectory data of historical customer service dialogues to obtain customer service standard operation process trees corresponding to various consultation questions in different customer service scenarios, please refer to the relevant descriptions in the previous embodiments. These will not be repeated in the embodiments of this application.

[0081] Step 501: Associate and store the customer service standard operation process tree with the scenario information and the consultation question.

[0082] Optionally, the customer service standard operation process tree can be stored in the following form: (question type, consultation scenario, consultation question, customer service standard operation process tree), wherein the question type and consultation scenario are used as scenario information to describe the scenario in the customer service manual 8 / 16 page 12 CN 121434336 A. In specific applications, the pre-stored customer service standard operation process tree can be matched based on the scenario information and the consultation question.

[0083] Step 506: Input the dialogue text, the customer service standard operation process tree, the tool information of the preset operation tool and / or the preset query capability information into the pre-fine-tuned generative large model, and guide the generative large model to generate the operation result of the current customer service operation steps for the incoming customer service dialogue.

[0084] As can be seen from the method for generating the customer service standard operation flow tree described above: the customer service standard operation flow tree is used to describe one or more service paths for a specified consultation question, and the service path includes service nodes corresponding to customer service operation steps. Optionally, the step of inputting the dialogue text, the customer service standard operation flow tree, the tool information of the preset operation tool and / or the preset query capability information into the pre-fine-tuned generative large model, and guiding the generative large model to generate the operation result of the current customer service operation step for the inbound customer service dialogue, includes: inputting the dialogue text, the customer service standard operation flow tree, the tool information of the preset operation tool and / or the preset query capability information into the pre-fine-tuned generative large model, using a thinking chain approach to guide the generative large model to generate the customer service operation step to be executed currently using the customer service standard operation flow tree as a constraint, and executing the customer service operation step to be executed currently based on the tool information of the preset operation tool and / or the preset query capability information to obtain the operation result of the current customer service operation step for the inbound customer service dialogue. Optionally, the customer service standard operation process tree is first textualized, and the textualized customer service standard operation process tree is input into the generative large model.

[0085] Optionally, the customer service operation steps correspond to any of the following service operation types: viewing historical call records, asking follow-up questions, querying the knowledge base, calling tools, creating work orders, transferring work orders, etc. In the customer service standard operation process tree, not only are the service paths and customer service operation steps for specific problem types, specific consultation scenarios, and specific consultation questions recorded, but also the task content and execution method of the corresponding customer service operation steps are recorded. For example, the tool information, service interface, knowledge base to be queried, service operation input parameters, service operation output parameters, and one or more other information such as the tool information, service interface, knowledge base to be queried, service operation input parameters, and service operation output parameters of the tools to be called to execute the corresponding customer service operation steps are recorded.

[0086] After determining the customer service operation steps to be executed based on the user-input dialogue text and the customer service standard operation flow tree, the generative big model further obtains the specific execution task content and task execution method of the customer service operation steps to be executed from the customer service standard operation flow tree. Then, it executes the task content according to the task execution method to obtain the operation result of the current customer service operation step. For example, after the generative big model determines that the customer service operation step to be executed is the historical call record viewing step based on the dialogue text and the customer service standard operation flow tree, it further obtains the service interface for viewing historical call records from the customer service standard operation flow tree, calls the service interface, and obtains the historical call records as the operation result of the current customer service operation step.

[0087] In specific implementation, the generative big model is obtained by fine-tuning the generative big model commonly used in the prior art based on the service trajectory data of customer service dialogue, the pre-generated customer service standard operation flow tree, and the tools, services, knowledge bases, etc. integrated in the intelligent customer service system. The fine-tuned generative big model has the ability to infer and generate the customer service operation steps to be executed based on the currently input incoming customer dialogue context and the customer service standard operation flow tree. On the other hand, the fine-tuned generative large model also has the ability to execute the customer service operation steps that need to be executed according to the task content and execution method required by the specified customer service operation steps described in the customer service standard operation process tree.

[0088] The fine-tuning method of the generative large model is described in the prior art and will not be repeated in the embodiments of this application.

[0089] Step 508: Based on the operation result, generate the current intelligent reply text of the incoming customer service dialogue.

[0090] The operation result may include, but is not limited to, any of the following: tool call result, knowledge base query result, large model output text, etc. Optionally, the operation result can be directly used as the current intelligent reply text of the incoming customer service dialogue and output to the user through the dialogue window of the incoming customer service dialogue.

[0091] In some optional embodiments, as shown in FIG6, after generating the current intelligent reply text of the incoming customer service dialogue based on the operation result, the method further includes: steps 510 and 512.

[0092] Step 510: Obtain evaluation data of the intelligent reply text for the incoming customer service dialogue.

[0093] Evaluation data of the intelligent reply text for the incoming customer service dialogue can be obtained using methods in the art. For example, user evaluations of each incoming customer service dialogue can be collected through a consultation reception platform.

[0094] Step 512: Optimize the customer service standard operation process tree based on the evaluation data.

[0095] Optionally, the customer service standard operation process tree with low service evaluation can be optimized by manual verification based on the service evaluation. Alternatively, the service trajectory data of the incoming customer service dialogue with high service evaluation can be used to optimize and iterate knowledge accumulation, and update the customer service standard operation process tree.

[0096] By optimizing the customer service standard operation process tree based on service evaluation, feedback-driven optimization of the customer service standard operation process tree is achieved, so that the customer service standard operation process tree not only supports the dynamic evolution of application rules, but also makes the operation process closer to user needs and improves the user experience of intelligent dialogue products.

[0097] In summary, the intelligent dialogue method disclosed in this application extracts information from the dialogue text input by the user during the incoming customer service dialogue to obtain the scenario information and consultation question matching the dialogue text; obtains the pre-stored customer service standard operation process tree corresponding to the scenario information and consultation question; inputs the dialogue text, the customer service standard operation process tree, the tool information of the preset operation tool and / or the preset query capability information into the pre-fine-tuned generative large model, and uses the customer service standard operation process tree as a constraint to guide the generative large model to generate the operation result of the current customer service operation step for the incoming customer service dialogue; finally, based on the operation result, the current intelligent reply text of the incoming customer service dialogue is generated. Compared with directly calling the generative large model to generate intelligent reply text, this method effectively suppresses the illusion problem of the large model and improves the reliability of the reply text.

[0098] On the other hand, this method transforms customer service experience into an iterative digital process and applies it to intelligent dialogue products, realizing a positive cycle of "human service wisdom - standardized process - intelligent service capability", which is a major breakthrough compared with traditional solutions. Furthermore, this method can transform customer service experience into a dynamic digital process in real time and allow it to evolve dynamically. It not only supports the dynamic evolution of application rules but also mines standard customer service operation processes for complex customer service scenarios.

[0099] Based on the above embodiments, this embodiment also provides a customer service standard operation process mining device, which includes:

[0100] a service trajectory data set acquisition module, used to extract and aggregate information from historical customer service dialogue service trajectory data to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios;

[0101] a target service path search module, used to mine frequent customer service operation path sequences in the service trajectory data according to the time order of customer service operations in the service trajectory data, to obtain target service paths associated with various consultation questions in different customer service scenarios;

[0102] a customer service standard operation process tree generation module, used to generate customer service standard operation process trees corresponding to various consultation questions in different customer service scenarios based on the target service paths.

[0103] Optionally, the service trajectory data set acquisition module is further configured to:

[0104] summarize and extract information from the service trajectory data of historical customer service dialogues to obtain clustering information of the service trajectory data, wherein the clustering information includes the following information: scenario information, dialogue identifier, and consultation question;

[0105] perform multi-stage aggregation on the clustering information to obtain a set of dialogue identifiers corresponding to consultation questions in different customer service scenarios;

[0106] obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios based on the set of dialogue identifiers corresponding to consultation questions in different customer service scenarios and the service trajectory data.

[0107] Optionally, the step of performing multi-stage aggregation on the clustering information to obtain a set of dialogue identifiers corresponding to the consultation questions in different customer service scenarios includes:

[0108] Based on the scenario information, performing aggregation processing on the consultation questions and dialogue identifiers in the clustering information to obtain a first set of consultation questions and dialogue identifiers in different customer service scenarios;

[0109] Performing aggregation processing on the consultation questions in the first set respectively to obtain a set of dialogue identifiers corresponding to the consultation questions in different customer service scenarios.

[0110] Optionally, the step of aggregating the consultation questions in the first set to obtain a set of dialogue identifiers corresponding to the consultation questions in different customer service scenarios includes:

[0111] performing text vectorization processing on the consultation questions in the first set to obtain text vectors for each consultation question;

[0112] calculating the similarity between consultation questions in the same first set based on the text vectors;

[0113] aggregating consultation questions whose similarity satisfies a preset similarity condition into the same type of consultation questions, and constructing a set of dialogue identifiers corresponding to the corresponding type of consultation questions in the customer service scenario corresponding to the first set based on the dialogue identifiers corresponding to the consultation questions aggregating into the same type of consultation questions.

[0114] Optionally, the target service path search module is further configured to:

[0115] extract information from the service trajectory data in the set according to the time sequence of customer service operations in the service trajectory data, and obtain service flow lines corresponding to various types of consultation questions under different customer service scenarios, wherein the service nodes in the service flow lines correspond to the customer service operations, and the paths in the service flow lines represent the upstream and downstream relationships of customer service operations;

[0116] using the service nodes as graph nodes, constructing directed edges connecting the corresponding graph nodes according to the upstream and downstream relationships represented by the paths, setting the weights of the corresponding directed edges according to the number of times the paths appear in the service flow lines, and obtaining a directed acyclic graph as the service flow line graph corresponding to various types of consultation questions under different customer service scenarios;

[0117] Based on the service flow diagram, frequent customer service operation path sequence mining is performed to obtain target service paths associated with various types of consultation questions under different customer service scenarios.

[0118] Optionally, the step of mining frequent customer service operation path sequences based on the service flow diagram to obtain target service paths associated with various types of consultation questions under different customer service scenarios includes:

[0119] Performing relaxation operations on the outgoing edges of the nodes in the service flow diagram in topological order to search for target service paths associated with various types of consultation questions under different customer service scenarios, wherein the sum of the weights of each directed edge in the target service path is greater than the sum of the weights of each directed edge in the known service path.

[0120] The customer service standard operation process mining device disclosed in this application embodiment is used to implement the above-mentioned customer service standard operation process mining method. For the specific implementation of each module of the device, please refer to the specific implementation of the corresponding steps in the foregoing method embodiment, which will not be repeated here.

[0121] In summary, the customer service standard operation process mining device disclosed in this application extracts and aggregates information from historical customer service dialogue service trajectory data to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios. Then, according to the time sequence of customer service operations in the service trajectory data, frequent customer service operation path sequence mining is performed on the service trajectory data in the set to obtain target service paths associated with various consultation questions in different customer service scenarios. Based on the target service paths, a customer service standard operation process tree corresponding to various consultation questions in different customer service scenarios as described on pages 11 / 16 of the specification, CN 121434336 A, is generated, effectively improving the efficiency of obtaining customer service standard operation processes. Furthermore, this method transforms real-time service operations into knowledge accumulation, supports the dynamic evolution of application rules, and generates customer service standard operation processes with higher reliability and stronger practicality.

[0122] Based on the above embodiments, this embodiment also provides an intelligent dialogue device, the device comprising:

[0123] a scenario information and consultation question acquisition module, used to extract information from the dialogue text input by the user during an incoming customer service dialogue after receiving the dialogue text, and acquire scenario information and consultation questions matching the dialogue text;

[0124] a customer service standard operation process tree acquisition module, used to acquire a pre-stored customer service standard operation process tree corresponding to the scenario information and the consultation question;

[0125] an operation result acquisition module, used to input the dialogue text, the customer service standard operation process tree, tool information of preset operation tools and / or preset query capability information into a pre-fine-tuned generative large model, and guide the generative large model to generate operation results for the current customer service operation steps of the incoming customer service dialogue;

[0126] The response module is used to generate the current intelligent response text of the incoming customer service dialogue based on the operation result.

[0127] Optionally, the customer service standard operation flow tree is used to describe one or more service paths for a specified consultation question. The service path includes service nodes corresponding to customer service operation steps. The operation result acquisition module is further used to:

[0128] input the dialogue text, the customer service standard operation flow tree, the tool information of the preset operation tool and / or the preset query capability information into the pre-fine-tuned generative large model, guide the generative large model to generate the customer service operation steps to be executed currently using the customer service standard operation flow tree as a constraint, and execute the customer service operation steps to be executed currently based on the tool information of the preset operation tool and / or the preset query capability information to obtain the operation result of the current customer service operation steps of the incoming customer service dialogue.

[0129] Optionally, before extracting information from the dialogue text input by the user during the incoming customer service dialogue and obtaining the scenario information and consultation question matched by the dialogue text after receiving the dialogue text, the device further includes:

[0130] a customer service standard operation process tree generation module, used to perform knowledge accumulation on the service trajectory data of historical customer service dialogues to obtain customer service standard operation process trees corresponding to various consultation questions under different customer service scenarios;

[0131] a customer service standard operation process tree storage module, used to associate and store the customer service standard operation process tree with the scenario information and the consultation question.

[0132] Optionally, the step of knowledge accumulation of service trajectory data of historical customer service dialogues to obtain customer service standard operation process trees corresponding to various consultation questions in different customer service scenarios includes:

[0133] extracting and aggregating information from the service trajectory data of historical customer service dialogues to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenarios;

[0134] mining frequent customer service operation path sequences of the service trajectory data in the set according to the time order of customer service operations in the service trajectory data to obtain target service paths associated with various consultation questions in different customer service scenarios;

[0135] generating customer service standard operation process trees corresponding to various consultation questions in different customer service scenarios based on the target service paths.

[0136] Optionally, after generating the current intelligent reply text of the incoming customer service dialogue based on the operation results, the device further includes:

[0137] a customer service standard operation process tree optimization module, used to obtain evaluation data of the intelligent reply text of the incoming customer service dialogue; and to optimize the customer service standard operation process tree based on the evaluation data. Instruction manual, pages 12 / 16, CN 121434336 A

[0138] The intelligent dialogue device disclosed in this application is used to implement the above-mentioned intelligent dialogue method. For the specific implementation of each module of the device, please refer to the specific implementation of the corresponding steps in the foregoing method embodiments, which will not be repeated here.

[0139] In summary, the intelligent dialogue device disclosed in this application extracts information from the dialogue text input by the user during the incoming customer service dialogue to obtain the scenario information and consultation question matching the dialogue text; obtains the pre-stored customer service standard operation process tree corresponding to the scenario information and consultation question; inputs the dialogue text, the customer service standard operation process tree, the tool information of the preset operation tool and / or the preset query capability information into the pre-fine-tuned generative large model, and uses the customer service standard operation process tree as a constraint condition to guide the generative large model to generate the operation result of the current customer service operation step for the incoming customer service dialogue; finally, based on the operation result, the current intelligent reply text of the incoming customer service dialogue is generated. Compared with directly calling the generative large model to generate intelligent reply text, it effectively suppresses the illusion problem of the large model and improves the reliability of the reply text.

[0140] On the other hand, this device transforms customer service experience into an iterable digital process and applies it to intelligent dialogue products, realizing a positive cycle of "human service wisdom - standardized process - intelligent service capability", which is a major breakthrough compared with traditional solutions. Furthermore, this device can transform customer service experience into dynamic digital processes in real time and evolve them dynamically. It not only supports the dynamic evolution of application rules but also uncovers standard customer service operation processes for complex customer service scenarios.

[0141] This application embodiment also provides a non-volatile readable storage medium, which stores one or more modules (programs). When these one or more modules are applied to a device, they can enable the device to execute the instructions of each method step in this application embodiment.

[0142] This application embodiment also provides a computer-readable storage medium, which stores computer-executable instructions. When executed by a processor, the computer-executable instructions are used to implement the method described in this application embodiment.

[0143] This application embodiment also provides an electronic device, including: a processor and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method described in this application embodiment. In this application embodiment, the electronic device includes devices such as servers and terminal devices.

[0144] This application also discloses a computer program product, including a computer program / computer executable instructions, which, when executed by a processor in an electronic device, implement the method described in this application.

[0145] Embodiments of this disclosure can be implemented as an apparatus configured as desired using any suitable hardware, firmware, software, or any combination thereof, which may include electronic devices such as servers (clusters), terminals, etc. FIG7 schematically illustrates an exemplary apparatus 700 that may be used to implement the various embodiments described in this application.

[0146] For one embodiment, FIG7 illustrates an exemplary apparatus 700 having one or more processors 702, a control module (chipset) 704 coupled to at least one of the processors(one or more) 702, a memory 706 coupled to the control module 704, a non-volatile memory (NVM) / storage device 708 coupled to the control module 704, one or more input / output devices 710 coupled to the control module 704, and a network interface 712 coupled to the control module 704.

[0147] The processor 702 may include one or more single-core or multi-core processors, and the processor 702 may include any combination of general-purpose processors or special-purpose processors (e.g., graphics processors, application processors, baseband processors, etc.). In some embodiments, device 700 can function as a server, terminal, or other device as described in the embodiments of this application. Specification 13 / 16 pages 17 CN 121434336 A

[0148] In some embodiments, device 700 may include one or more computer-readable media (e.g., memory 706 or NVM / storage device 708) having instructions 714 and one or more processors 702 configured to execute instructions 714 in conjunction with the one or more computer-readable media to implement a module thereby performing the actions described in this disclosure.

[0149] In one embodiment, control module 704 may include any suitable interface controller to provide any suitable interface to at least one of the processors 702 and / or any suitable device or component communicating with control module 704.

[0150] Control module 704 may include a memory controller module to provide an interface to memory 706. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0151] Memory 706 may be used, for example, to load and store data and / or instructions 714 for device 700. In one embodiment, memory 706 may include any suitable volatile memory, such as suitable DRAM. In some embodiments, memory 706 may include double data rate type quad synchronous dynamic random access memory (DDR4 SDRAM).

[0152] In one embodiment, control module 704 may include one or more input / output controllers to provide an interface to NVM / storage device 708 and (one or more) input / output devices 710.

[0153] For example, NVM / storage device 708 may be used to store data and / or instructions 714. NVM / storage device 708 may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable (one or more) non-volatile storage devices (e.g., one or more hard disk drive (HDD), one or more optical disc (CD) drives, and / or one or more digital universal optical disc (DVD) drives).

[0154] NVM / storage device 708 may include storage resources that are part of a device on which device 700 is mounted, or that are accessible by the device but do not necessarily have to be part of the device. For example, NVM / storage device 708 may be accessed via a network via (one or more) input / output devices 710.

[0155] (One or more) input / output devices 710 may provide an interface for device 700 to communicate with any other suitable device, and input / output devices 710 may include communication components, audio components, sensor components, etc. Network interface 712 may provide an interface for device 700 to communicate via one or more networks. Device 700 may wirelessly communicate with one or more components of a wireless network according to any standard and / or protocol in one or more wireless network standards and / or protocols, such as accessing wireless networks based on communication standards, such as Bluetooth, WiFi, 2G, 3G, 4G, 5G, etc., or combinations thereof.

[0156] In one embodiment, at least one of the processors 702 may be logically packaged with one or more controllers (e.g., memory controller modules) of control module 704. In one embodiment, at least one of the processors 702 may be logically packaged with one or more controllers of control module 704 to form a system-in-package (SiP). In one embodiment, at least one of the processors 702 may be logically integrated with one or more controllers of control module 704 on the same die. In one embodiment, at least one of the processors 702 may be logically integrated with one or more controllers of control module 704 on the same die to form a system-on-a-chip (SoC).

[0157] In various embodiments, device 700 may be, but is not limited to, a terminal device such as a server, desktop computing device, or mobile computing device (e.g., laptop computing device, handheld computing device, tablet computer, netbook, etc.). In various embodiments, device 700 may have more or fewer components and / or different architectures. For example, in some embodiments, device 700 includes one or more cameras, a keyboard, a liquid crystal display (LCD) screen (including a touchscreen display), a non-volatile memory port, multiple antennas, a graphics chip, an application-specific integrated circuit (ASIC), and a speaker.

[0158] The detection device may use a main control chip as a processor or control module. Sensor data, location information, etc., are stored in a memory or NVM / storage device. The sensor group may be used as an input / output device. The communication interface may include the interface described in the network specification page 14 / 16, 18 CN 121434336 A.

[0159] This application embodiment also provides an electronic device, including: a processor; and a memory, on which executable code is stored. When the executable code is executed, the processor performs one or more of the methods described in this application embodiment. In this application embodiment, the memory may store various data, such as target files, file and application association data, etc., and may also include user behavior data, thereby providing a data basis for various processing.

[0160] This application embodiment also provides one or more machine-readable media, on which executable code is stored. When the executable code is executed, the processor performs one or more of the methods described in this application embodiment.

[0161] For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For relevant parts, please refer to the description of the method embodiment.

[0162] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably.

[0163] The embodiments of this application are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0164] These computer program instructions may also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing terminal device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0165] These computer program instructions may also be loaded onto a computer or other programmable data processing terminal device to cause a series of operational steps to be performed on the computer or other programmable terminal device to produce a computer-implemented process, thereby enabling...Instructions executed on a computer or other programmable terminal device provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0166] Although preferred embodiments of the embodiments of this application have been described, those skilled in the art, once they learn the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of this application.

[0167] Finally, it should also be noted that, in this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further restrictions, an element defined by the phrase "including one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0168] The above provides a detailed description of a customer service standard operation process mining method, an intelligent dialogue method, an electronic device, a storage medium, and a computer program product provided by this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and its core ideas in the specification of this application (pages 15 / 16, CN 121434336 A). At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application. Instruction manual 16 / 16 pages 20 CN 121434336 A Figure 1 Figure 2 Figure 3 Instruction manual Figure 1 / 4 pages 21 CN 121434336 A Figure 4 Figure 5 Instruction manual Figure 2 / 4 pages 22 CN 121434336 A Figure 6 Instruction manual Figure 3 / 4 pages 23 CN 121434336 A Figure 7 Instruction manual Figure 4 / 4 pages 24 CN 121434336 A Abstract The embodiment of the invention provides a customer service standard operation process miningmethod, an intelligent dialogue method, electronic equipment, a storage medium and a computer program product. The customer service standard operation process mining method comprises the steps of performing information extraction and aggregation processing on service trajectory data of historical customer service dialogues to obtain a set of service trajectory data corresponding to consultation questions in different customer service scenes; according to the time sequence of customer service operation in the service track data, frequent customer service operation path sequence mining is carried out on the service track data in the set, and target service paths associated with various consultation questions in different customer service scenes are obtained; and based on the target service path, generating customer service standard operation process trees corresponding to various consultation questions in different customer service scenes. According to the method, the acquisitionefficiency of the customer service standard operation process is improved.

Claims

1. A method for mining standard customer service operating procedures, characterized in that, The method comprises: information extraction and aggregation processing of service track data of historical customer service dialogues to obtain a set of service track data corresponding to consultation problems under different customer service scenarios; According to the time sequence of the customer service operations in the service track data, the service track data in the set is respectively subjected to frequent customer service operation path sequence mining to obtain target service paths respectively associated with each type of consultation problem under different customer service scenarios; Based on the target service path, a customer service standard operation process tree corresponding to each type of consultation problem under different customer service scenarios is generated.

2. The method of claim 1, wherein, The information extraction and aggregation processing of the service track data of the historical customer service dialogues to obtain the set of service track data corresponding to the consultation problems under different customer service scenarios comprises: Summarizing and information extraction of the service track data of the historical customer service dialogues to obtain clustering information of the service track data, wherein the clustering information comprises the following information: scenario information, dialogue identifier, consultation problem; The clustering information is subjected to multi-stage aggregation to obtain a set of dialogue identifiers corresponding to the consultation problems under different customer service scenarios; According to the set of dialogue identifiers corresponding to the consultation problems under different customer service scenarios and the service track data, a set of service track data corresponding to the consultation problems under different customer service scenarios is obtained.

3. The method of claim 2, wherein, The multi-stage aggregation of the clustering information to obtain the set of dialogue identifiers corresponding to the consultation problems under different customer service scenarios comprises: Based on the scenario information, the consultation problems and the dialogue identifiers in the clustering information are subjected to aggregation processing to obtain a first set of consultation problems and dialogue identifiers under different customer service scenarios; The consultation problems in the first set are respectively subjected to aggregation processing to obtain a set of dialogue identifiers corresponding to the consultation problems under different customer service scenarios.

4. The method of claim 3, wherein, The aggregation processing of the consultation problems in the first set to obtain the set of dialogue identifiers corresponding to the consultation problems under different customer service scenarios comprises: Text vectorization processing of the consultation problems in the first set is performed to obtain text vectors of each consultation problem; Similarities between the consultation problems in the same first set are calculated based on the text vectors; The consultation problems satisfying the preset similarity condition are aggregated into the same type of consultation problems, and a set of dialogue identifiers corresponding to the corresponding type of consultation problems under the customer service scenario corresponding to the first set is constructed based on the dialogue identifiers corresponding to the consultation problems aggregated into the same type of consultation problems.

5. The method of claim 1, wherein, The frequent customer service operation path sequence mining of the service track data in the set according to the time sequence of the customer service operations in the service track data to obtain the target service paths respectively associated with each type of consultation problem under different customer service scenarios comprises: According to the time sequence of the customer service operations in the service trajectory data, information extraction is performed on the service trajectory data in the set to obtain service action lines corresponding to various types of consultation questions in different customer service scenarios, service nodes in the service action lines correspond to the customer service operations, and paths in the service action lines represent upstream and downstream relationships of the customer service operations; The service nodes are taken as graph nodes, directed edges connecting the corresponding graph nodes are constructed according to the upstream and downstream relationships represented by the paths, the weights of the corresponding directed edges are set according to the number of occurrences of the paths in the service action lines, and a directed acyclic graph is obtained as a service action line graph corresponding to various types of consultation questions in different customer service scenarios; Based on the service action line graph, frequent customer service operation path sequence mining is performed to obtain target service paths associated with various types of consultation questions in different customer service scenarios.

6. The method of claim 5, wherein, The method comprises: After receiving the dialogue text input by the user in the incoming customer service dialogue process, information extraction is performed on the dialogue text to obtain scene information and consultation questions matched by the dialogue text; 7. An intelligent dialog method, characterized by, A pre-stored customer service standard operation process tree corresponding to the scene information and the consultation questions is obtained; The dialogue text, the customer service standard operation process tree, tool information of a preset operation tool and / or preset query capability information are input into a pre-tuned generative large model to guide the generative large model to generate an operation result of a current customer service operation step for the incoming customer service dialogue; Based on the operation result, a current intelligent reply text of the incoming customer service dialogue is generated. The customer service standard operation process tree is used to describe one or more service paths of a specified consultation question, the service paths include service nodes corresponding to customer service operation steps, and the dialogue text, the customer service standard operation process tree, tool information of a preset operation tool and / or preset query capability information are input into a pre-tuned generative large model to guide the generative large model to generate an operation result of a current customer service operation step for the incoming customer service dialogue, which comprises: The dialogue text, the customer service standard operation process tree, tool information of a preset operation tool and / or preset query capability information are input into a pre-tuned generative large model, and the generative large model is guided to generate the current customer service operation step to be executed by taking the customer service standard operation process tree as a constraint in a thinking chain manner, and the current customer service operation step to be executed is executed based on the tool information of the preset operation tool and / or the preset query capability information to obtain the operation result of the current customer service operation step for the incoming customer service dialogue.

8. The method of claim 7, wherein, ​ ​ 9. The method of claim 7, wherein, Before the information extraction on the dialogue text input by the user in the incoming line customer service dialogue process and the acquisition of the scene information and the consultation question matched by the dialogue text, the method further comprises: Knowledge is deposited on service track data of historical customer service dialogues to obtain customer standard operation process trees corresponding to various consultation questions in different customer service scenarios; The customer standard operation process trees are stored in association with the scene information and the consultation question.

10. The method of claim 9, wherein, The knowledge deposition on the service track data of the historical customer service dialogues to obtain the customer standard operation process trees corresponding to various consultation questions in different customer service scenarios comprises: Information extraction and aggregation processing are performed on service track data of historical customer service dialogues to obtain a set of service track data corresponding to consultation questions in different customer service scenarios; According to a time sequence of customer service operations in the service track data, frequent customer service operation path sequence mining is performed on the service track data in the set to obtain target service paths associated with various consultation questions in different customer service scenarios; Based on the target service paths, customer standard operation process trees corresponding to various consultation questions in different customer service scenarios are generated.

11. The method of claim 7, wherein, After the generation of the current intelligent reply text of the incoming line customer service dialogue based on the operation result, the method further comprises: Evaluation data of the intelligent reply text of the incoming line customer service dialogue is acquired; Based on the evaluation data, the customer standard operation process trees are optimized.

12. An electronic device, comprising: Comprise: A processor and a memory connected in communication with the processor; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory to realize the method in any one of claims 1-11.

13. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to realize the method in any one of claims 1-11.

14. A computer program product comprising computer program / computer executable instructions, characterized in that, The computer program / computer executable instructions are executed by the processor in the electronic device to realize the method in any one of claims 1-11.

Images