Flow mining system and mining method based on conversation scenarios

Abstract

This invention discloses a process mining system and method based on conversation scenarios, belonging to the field of process mining technology. This invention can automatically obtain the corresponding user intent and the corresponding conversation flow after analyzing historical conversation data. Compared with traditional methods that rely on experience to classify user intent types and design processes, this method is more efficient and considers user intents more comprehensively, avoiding the problem of missing user intent statistics due to limited personal energy and ability, thus preventing the loss of corresponding conversation flows. This invention automatically analyzes a complete dialogue text and groups the dialogue text logs according to their purpose, reducing the differences within the same group of dialogue text logs. This helps improve the rationality and accuracy of the finalized mining process, significantly reducing the workload of manual analysis and minimizing the negative impact of personal experience on the accuracy of the classification results.

Description

Technical Field

[0001] This invention belongs to the field of process mining technology, specifically, it relates to a process mining system and mining method based on conversation scenarios. Background Technology

[0002] Workflow mining is a relatively new application of data mining in the field of workflow management. The initial goal of workflow mining was to analyze the logs generated during workflow execution to reconstruct the actual business process, and then use this knowledge to analyze and optimize the workflow.

[0003] With the development of internet technology, intelligent customer service has become widely used. However, in the current technology, when formulating intelligent customer service processes in the field of customer service consultation, the main approach is to rely on the experience of staff to classify user intent types and design processes. However, this method is affected by factors such as staff experience, and the designed processes may not be simple. Moreover, as the number of user intents increases, the design workload and completion rate will be affected. In addition, the incomplete statistics of user intents result in some user intents not being included in the designed customer process, requiring manual transfer for answering corresponding customer service inquiries in actual use. To solve the above problems, this invention provides a process mining system and method that can automatically mine processes and significantly improve the quality of process design. The following technical solution is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide a process mining system and method based on conversation scenarios, which solves the problem that the design of customer service processes in existing customer service consultation services relies on the experience of staff, resulting in low efficiency, incomplete user intent statistics, and affecting the comprehensiveness of customer service processes.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] The process mining method based on conversation scenarios includes the following steps:

[0007] S1. Obtain the dialogue text and divide it into several dialogue text combinations according to the purpose of the dialogue text;

[0008] S2. For dialogue texts in the same dialogue text combination, obtain the customer service intents within them, and mark the customer service intents that appear in more than γ% of the dialogue texts in the dialogue text combination as necessary customer service intents, where γ is a preset value.

[0009] The customer service intents are assigned values ​​according to their time sequence. Within the scope of the dialogue text in which all necessary customer service intents exist, the necessary customer service intents are assigned values ​​of 1, 2, ..., v in the order in which they appear in a dialogue text, where v is the number of necessary customer service intents.

[0010] Calculate the sum Fh of the values ​​assigned to a necessary customer service intent in all dialogue texts, and sort the necessary customer service intents in ascending order of the sum Fh.

[0011] S3. Obtain the corresponding dialogue statements based on the necessary customer service intent;

[0012] The session flow is generated based on the order of necessary customer service intentions and the corresponding dialogue statements for each necessary customer service intention.

[0013] As a further embodiment of the present invention, the value of γ is 70.

[0014] As a further aspect of the present invention, the method for obtaining the target of the dialogue text is as follows:

[0015] S11. Based on a complete dialogue text, obtain the user's intent set and the customer service's intent set in the dialogue text through intent recognition;

[0016] S12. Process the dialogue text in the conversation data storage unit to obtain the user intent set and customer service intent set in each dialogue text. Represent the user intent A and customer service intent B as A{B1k1, B2k2, ..., Bnkn}, where Biki represents the number of dialogue texts containing the i-th customer service intent Bi among the n dialogue texts containing user intent A.

[0017] Where 1≤i≤n;

[0018] S13. When ki / n≥β holds true, the corresponding customer service intent Bi is considered to be a sub-intent of the corresponding user intent A, where β is a preset percentage coefficient.

[0019] S14. The number of dialogue texts containing user intent A, including each sub-intent corresponding to user intent A, will be sequentially marked as z1, z2, ..., zs.

[0020] The intensity value Y of user intent A is calculated according to the formula Y = (z1 / n + z2 / n + ... + zs / n) * s. When Y is greater than the preset value Y1, the user intent A corresponding to Y is considered to be the higher-level intent, which is the purpose of the corresponding dialogue text.

[0021] As a further embodiment of the present invention, the value of β is 0.65.

[0022] As a further aspect of the present invention, when two or more higher-level intentions exist simultaneously in a dialogue text, the corresponding dialogue text is deleted from the conversation data storage unit and is not used as a sample for subsequent dialogue text combination and division.

[0023] As a further aspect of the present invention, during customer service conversations, the user's intent is analyzed through intent recognition, and the corresponding conversation flow is matched according to different conversation scenarios.

[0024] As a further aspect of the present invention, a process mining system based on a conversation scenario is also disclosed, the process mining system comprising:

[0025] A session data storage unit is used to store session data;

[0026] The text log generation unit is used to analyze and process conversation data to generate dialogue text;

[0027] The intent recognition unit analyzes the dialogue text using an intent recognition algorithm to obtain the user intent and the customer service intent, and then transmits the obtained user intent and customer service intent to the control unit.

[0028] The control unit is used to analyze the user intent and customer service intent in the dialogue text and obtain the order in which the customer service intent appears in dialogue texts with different purposes.

[0029] The conversation flow generation unit generates a conversation flow based on the customer service intent and the order in which the customer service intent appears.

[0030] As a further aspect of the present invention, the dialogue text is a text log generated from text chat records.

[0031] As a further aspect of the present invention, the dialogue text is a text log generated by converting voice logs.

[0032] The beneficial effects of this invention are:

[0033] (1) The present invention can automatically obtain the corresponding user intent and the corresponding conversation flow after analyzing historical conversation data. Compared with the traditional method of classifying user intent types and designing processes based on experience, it is more efficient and considers user intent more comprehensively, avoiding the problem of missing user intent statistics due to limited personal energy and ability, which in turn leads to the missing corresponding conversation flow.

[0034] (2) This invention utilizes the characteristic that only user intents that are highly related to the corresponding customer service intents are the main purpose around which the dialogue text content revolves. This invention automatically analyzes a complete dialogue text and groups the dialogue text logs according to their purpose orientation, reducing the differences in the same group of dialogue text logs, which is beneficial to the subsequent analysis work and improves the rationality and accuracy of the formed mining process. In addition, automatically grouping according to the text content can also significantly reduce the workload of manual analysis and reduce the negative impact of personal experience on the accuracy of the division results.

[0035] (3) This invention deletes a portion of customer service intentions by setting a certain threshold. The threshold can be reasonably adjusted according to the actual application scenario. It can delete customer service intentions that are not necessary to complete the main purpose in the corresponding dialogue text combination, reduce or eliminate interference caused by personal habits. In addition, by assigning values ​​in sequence and then calculating the sum of the assigned values, the overall order of appearance of necessary customer service intentions can be intuitively expressed, thereby automatically completing the sorting. Attached Figure Description

[0036] The invention will now be further described with reference to the accompanying drawings.

[0037] Figure 1 This is a schematic diagram of the framework structure of the process mining system based on the conversation scenario of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] A process mining system based on conversation scenarios, such as Figure 1 As shown, it includes:

[0040] A session data storage unit is used to store session data within a certain range;

[0041] The certain range includes a certain range in time and a certain range in content. The certain range in time includes selecting dialogue texts within a preset time t1 in the past when selecting dialogue texts as samples to ensure the updates of the samples. The certain range in content includes dialogue texts generated when solving user problems in a certain area of ​​the same industry, characterized by the limited intent of the users.

[0042] The text log generation unit is used to analyze and process conversation data to generate dialogue text;

[0043] The dialogue text can be a text log generated from text chat records or a text log generated by converting voice logs.

[0044] The intent recognition unit analyzes the dialogue text using an intent recognition algorithm to obtain the user intent and the customer service intent, and then transmits the obtained user intent and customer service intent to the control unit.

[0045] The user intent mentioned above refers to the intent obtained by analyzing and identifying the user's dialogue text, while the customer service intent refers to the intent obtained by analyzing and identifying the customer service staff's dialogue text.

[0046] The control unit is used to analyze the user intent and customer service intent in the dialogue text and obtain the order in which the customer service intent appears in dialogue texts with different purposes.

[0047] The conversation flow generation unit generates a conversation flow based on the customer service intent and the order in which the customer service intent appears;

[0048] This invention also discloses a process mining method based on conversation scenarios. This method is performed using the aforementioned mining system and includes the following steps:

[0049] S1. Obtain the dialogue text and divide it into several dialogue text combinations according to the purpose of the dialogue text;

[0050] In one embodiment of the present invention, the method for obtaining the destination of the dialogue text is as follows:

[0051] S11. Based on a complete dialogue text, obtain the user's intent set and the customer service's intent set in the dialogue text through intent recognition;

[0052] The complete dialogue text refers to the text of a single conversation completed in the process of resolving the same user's problem;

[0053] S12. Process the dialogue text in the conversation data storage unit to obtain the user intent set and customer service intent set in each dialogue text. Represent the user intent A and customer service intent B as A{B1k1, B2k2, ..., Bnkn}, where Biki represents the number of dialogue texts containing the i-th customer service intent Bi among the n dialogue texts containing user intent A.

[0054] Where 1≤i≤n;

[0055] S13. When ki / n≥β holds true, the corresponding customer service intent Bi is considered to be a sub-intent of the corresponding user intent A.

[0056] Wherein β is a preset percentage coefficient, and in one embodiment of the present invention, the value of β is 0.65;

[0057] S14. The number of dialogue texts containing user intent A, including each sub-intent corresponding to user intent A, will be sequentially marked as z1, z2, ..., zs.

[0058] The intensity value Y of user intent A is calculated according to the formula Y = (z1 / n + z2 / n + ... + zs / n) * s. When Y is greater than the preset value Y1, the user intent A corresponding to Y is considered to be the higher-level intent, which is the purpose of the corresponding dialogue text.

[0059] When dividing a dialogue text into several groups, dialogue texts with the same overarching intent are grouped into the same dialogue text group.

[0060] In one embodiment of the present invention, when two or more higher-level intentions exist simultaneously in a dialogue text, the corresponding dialogue text is deleted from the conversation data storage unit and is not used as a sample for subsequent dialogue text combination and division.

[0061] In actual customer service conversations, users typically have one or more primary objectives. Around these primary objectives, customer service representatives and users usually engage in 3-5 question-and-answer sessions to obtain the necessary auxiliary information and resolve the primary objective. While different human customers and users may differ in the order and process of their questions and answers when resolving the same primary objective, the auxiliary information they need is largely the same. Therefore, this invention utilizes this characteristic: only user intentions highly related to the corresponding customer service intentions represent the primary objective around which the dialogue text content revolves. This invention automatically analyzes a complete dialogue text and groups the dialogue text logs according to their objective orientation, reducing the discrepancies within the same group of dialogue text logs. This facilitates subsequent analysis and improves the rationality and accuracy of the established data mining process. Furthermore, automatically grouping based on text content significantly reduces the workload of manual analysis and minimizes the negative impact of personal experience on the accuracy of the grouping results.

[0062] S2. For dialogue texts in the same dialogue text combination, obtain the customer service intents within them, and mark the customer service intents that appear in more than γ% of the dialogue texts in the dialogue text combination as necessary customer service intents.

[0063] Wherein γ is a preset value, and in one embodiment of the present invention, the value of γ is 70;

[0064] The customer service intents are assigned values ​​according to the timing of each necessary customer service intent. Specifically:

[0065] Within the scope of the dialogue text in which all necessary customer service intentions exist, assign values ​​1, 2, ..., v in the order in which the necessary customer service intentions appear in a dialogue text, where v is the number of necessary customer service intentions;

[0066] Calculate the sum Fh of the values ​​assigned to a necessary customer service intent in all dialogue texts, and sort the necessary customer service intents in ascending order of the sum Fh, with the necessary customer service intent with the smallest sum Fh listed first.

[0067] This step deletes some customer service intentions by setting a certain threshold. The threshold can be reasonably adjusted according to the actual application scenario. It can delete customer service intentions in the corresponding dialogue text combination that are not necessary to complete the main purpose, reduce or eliminate interference caused by personal habits. In addition, by assigning values ​​in sequence and then calculating the sum of the values, it can intuitively express the overall order of appearance of necessary customer service intentions, thereby automatically completing the sorting.

[0068] S3. Input or generate corresponding dialogue statements according to the necessary customer service intent;

[0069] The conversation flow is generated based on the order of necessary customer service intentions and the corresponding dialogue statements for each necessary customer service intention.

[0070] During customer service conversations, intent recognition is used to analyze user intent and match the corresponding conversation flow according to different conversation scenarios.

[0071] This invention can automatically obtain the corresponding user intent and the corresponding conversation flow after analyzing historical conversation data. Compared with the traditional method of classifying user intent types and designing processes based on experience, it is more efficient and considers user intents more comprehensively. It avoids the problem of missing user intent statistics due to limited personal energy and ability, which leads to the lack of corresponding conversation flows.

[0072] In the description of this specification, the references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0073] The above description is merely an example and illustration of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. A process mining method based on conversation scenarios, characterized in that, Includes the following steps: S1. Obtain the dialogue text and divide it into several dialogue text combinations according to the purpose of the dialogue text; S2. For dialogue texts in the same dialogue text combination, obtain the customer service intents within them, and mark the customer service intents that appear in more than γ% of the dialogue texts in the dialogue text combination as necessary customer service intents, where γ is a preset value. The customer service intents are assigned values ​​according to their time sequence. Within the scope of the dialogue text in which all necessary customer service intents exist, the necessary customer service intents are assigned values ​​of 1, 2, ..., v in the order in which they appear in a dialogue text, where v is the number of necessary customer service intents. Calculate the sum Fh of the values ​​assigned to a necessary customer service intent in all dialogue texts, and sort the necessary customer service intents in ascending order of the sum Fh. S3. Obtain the corresponding dialogue statements based on the necessary customer service intent; The conversation flow is generated based on the order of necessary customer service intentions and the corresponding dialogue statements for each necessary customer service intention. The method for obtaining the target of the dialogue text is as follows: S11. Based on a complete dialogue text, obtain the user's intent set and the customer service's intent set in the dialogue text through intent recognition; S12. Process the dialogue text in the conversation data storage unit to obtain the user intent set and customer service intent set in each dialogue text. Represent the user intent A and customer service intent B as A{B1k1, B2k2, ..., Bnkn}, where Biki represents the number of dialogue texts containing the i-th customer service intent Bi among the n dialogue texts containing user intent A. Where 1≤i≤n; S13. When ki / n≥β holds true, the corresponding customer service intent Bi is considered to be a sub-intent of the corresponding user intent A, where β is a preset percentage coefficient. S14. The number of dialogue texts containing user intent A, including each sub-intent corresponding to user intent A, will be sequentially marked as z1, z2, ..., zs. The intensity value Y of user intent A is calculated according to the formula Y=(z1 / n+z2 / n+、…、+zs / n)*s. When Y is greater than the preset value Y1, the user intent A corresponding to Y is considered to be the higher-level intent, which is the purpose of the corresponding dialogue text.

2. The process mining method based on conversation scenarios according to claim 1, characterized in that, The value of γ is 70.

3. The process mining method based on conversation scenarios according to claim 1, characterized in that, The value of β is 0.

65.

4. The process mining method based on conversation scenarios according to claim 1, characterized in that, When a dialogue text contains two or more overarching intents, the corresponding dialogue text is deleted from the conversation data storage unit and is not used as a sample for subsequent dialogue text combination and segmentation.

5. The process mining method based on conversation scenarios according to claim 1, characterized in that, During customer service conversations, intent recognition is used to analyze user intent and match the corresponding conversation flow according to different conversation scenarios.

6. A process mining system based on conversation scenarios, characterized in that, include: A session data storage unit is used to store session data; The text log generation unit is used to analyze and process conversation data to generate dialogue text; The intent recognition unit analyzes the dialogue text using an intent recognition algorithm to obtain the user intent and the customer service intent, and then transmits the obtained user intent and customer service intent to the control unit. The control unit is used to analyze the user intent and customer service intent in the dialogue text and obtain the order in which the customer service intent appears in dialogue texts with different purposes. The conversation flow generation unit generates a conversation flow based on the customer service intent and the order in which the customer service intent appears; The method for obtaining the target of the dialogue text is as follows: S11. Based on a complete dialogue text, obtain the user's intent set and the customer service's intent set in the dialogue text through intent recognition; S12. Process the dialogue text in the conversation data storage unit to obtain the user intent set and customer service intent set in each dialogue text. Represent the user intent A and customer service intent B as A{B1k1, B2k2, ..., Bnkn}, where Biki represents the number of dialogue texts containing the i-th customer service intent Bi among the n dialogue texts containing user intent A. Where 1≤i≤n; S13. When ki / n≥β holds true, the corresponding customer service intent Bi is considered to be a sub-intent of the corresponding user intent A, where β is a preset percentage coefficient. S14. The number of dialogue texts containing user intent A, including each sub-intent corresponding to user intent A, will be sequentially marked as z1, z2, ..., zs. The intensity value Y of user intent A is calculated according to the formula Y=(z1 / n+z2 / n+、…、+zs / n)*s. When Y is greater than the preset value Y1, the user intent A corresponding to Y is considered to be the higher-level intent, which is the purpose of the corresponding dialogue text.

7. The process mining system based on conversation scenarios according to claim 6, characterized in that, The dialogue text is a text log generated from text chat records.

8. The process mining system based on conversation scenarios according to claim 7, characterized in that, The dialogue text is a text log generated from voice logs.

Images