Customer intention analysis method, device, equipment and storage medium

Abstract

The present invention relates to the field of artificial intelligence, and discloses a customer intention analysis method, device, equipment and storage medium for improving the efficiency of customer intention analysis. The customer intention analysis method includes: preprocessing customer historical processing data, classifying the preprocessed historical data based on a preset decision tree algorithm, and obtaining a willingness classification standard; configuring intention rules based on a process speech template to obtain customer intention rules, receiving voice interaction data returned by an artificial intelligence voice robot, and comparing the voice interaction data with the customer intention rules to obtain a voice comparison result; extracting offline interaction data from the voice comparison result, matching the offline interaction data with a preset customer intention degree label to obtain a customer intention matching result; classifying the voice comparison result and the customer intention degree matching result to obtain a customer intention analysis result. In addition, the present invention also relates to blockchain technology, and the customer intention analysis result can be stored in the blockchain.

Description

Technical Field

[0001] The present invention relates to the field of similarity matching, and in particular to a customer intention analysis method, device, equipment and storage medium. Background Art

[0002] Customer intention analysis is mainly carried out by learning all voice data after the current artificial intelligence AI outbound calls, as well as the customer's corresponding outbound call situation, basic information, credit card usage, etc., to train the customer intention analysis model in order to discover customers with processing intentions and improve the processing rate.

[0003] There are two traditional methods for analyzing customer intentions. One is to listen back to the AI ​​outbound call recordings, manually analyze the interaction texts between AI and customers, extract the keywords and intentions in the customer interactions, and then follow up with the intended customers through offline data collection. The other is to set the intentions and nodes of the intended customers in advance when designing the AI ​​dialogue process. If the customer triggers the preset intention node in the actual interaction, these customers are defined as intended customers. However, the analysis process of the above two methods is complex and time-consuming, and requires analysis of the entire amount of interaction data, resulting in low efficiency in customer intention analysis. Summary of the invention

[0004] The present invention provides a customer intention analysis method, device, equipment and storage medium, which are used to classify pre-processed historical data based on a preset decision tree algorithm to obtain a willingness classification standard, compare voice interaction data with customer intention rules to obtain voice comparison results, extract offline interaction data in the voice comparison results, match the offline interaction data with preset customer intention labels to obtain customer intention matching results, and classify the voice comparison results and customer intention matching results based on the willingness classification standard to obtain customer intention analysis results, thereby improving the efficiency of customer intention analysis.

[0005] The first aspect of the present invention provides a customer intention analysis method, including: obtaining customer historical processing data, preprocessing the customer historical processing data to obtain preprocessed historical data, classifying the preprocessed historical data based on a preset decision tree algorithm to obtain a willingness classification standard; configuring intention rules based on a preset process speech template to obtain customer intention rules, receiving voice interaction data returned by an artificial intelligence voice robot, comparing the voice interaction data with the customer intention rules to obtain a voice comparison result; extracting offline interaction data in the voice comparison result, matching the offline interaction data with a preset customer intention degree label to obtain a customer intention matching result; based on the willingness classification standard, classifying the voice comparison result and the customer intention degree matching result respectively to obtain a customer intention analysis result.

[0006] Optionally, in a first implementation method of the first aspect of the present invention, the acquiring of customer historical processing data, preprocessing of the customer historical processing data to obtain preprocessed historical data, and classification of the preprocessed historical data based on a preset decision tree algorithm to obtain a willingness grading standard includes: acquiring customer historical processing data, filling in missing values, filtering outliers and filtering duplicate values ​​for the customer historical processing data to obtain preprocessed historical data; calling a preset decision tree algorithm to traverse the preprocessed historical data to obtain a target decision tree, wherein the target decision tree comprises multiple leaf nodes; acquiring the business processing rate corresponding to each leaf node in the target decision tree, and sorting each leaf node in order of the business processing rate from large to small to obtain a leaf node sorting result; and classifying the leaf node sorting results according to a preset customer volume grading standard to obtain a willingness grading standard.

[0007] Optionally, in a second implementation method of the first aspect of the present invention, the preset decision tree algorithm is called to traverse the preprocessed historical data to obtain a target decision tree, and the target decision tree contains multiple leaf nodes, including: traversing the preprocessed historical data to obtain traversal results, and classifying the traversal results according to preset customer group characteristics to obtain an initial decision tree; pruning the initial decision tree to obtain a target decision tree, and the target decision tree contains multiple leaf nodes, and each leaf node corresponds to a customer group characteristic.

[0008] Optionally, in a third implementation method of the first aspect of the present invention, the intention rules are configured based on a preset process speech template to obtain customer intention rules, voice interaction data returned by an artificial intelligence voice robot is received, and the voice interaction data is compared with the customer intention rules to obtain a voice comparison result, including: obtaining a process speech template, configuring intention rules based on multiple rule categories in the process speech template to obtain customer intention rules; receiving voice interaction data returned by an artificial intelligence voice robot to obtain a first match degree between the voice interaction data and the customer intention rules; calling a preset comparison algorithm to determine whether the first match degree is greater than a preset first match threshold to obtain a voice comparison result.

[0009] Optionally, in a fourth implementation method of the first aspect of the present invention, the extracting of offline interaction data from the voice comparison result, matching the offline interaction data with a preset customer intention label, and obtaining a customer intention matching result includes: extracting voice interaction data from the voice comparison result having a first matching degree less than or equal to a preset first matching threshold, to obtain offline interaction data, wherein the first matching degree is the matching degree between the voice interaction data and the customer intention rule; obtaining a second matching degree, calling a preset similarity matching algorithm, determining whether the second matching degree is greater than a preset second matching threshold, and obtaining a customer intention matching result, wherein the second matching degree is the matching degree between the offline interaction data and the preset customer intention label.

[0010] Optionally, in a fifth implementation manner of the first aspect of the present invention, the voice comparison results and the customer intention matching results are classified respectively based on the willingness classification standard to obtain the customer intention analysis results, including: extracting first customer data from the voice comparison results, and dividing the first customer data into a preset first-tier willingness classification based on the willingness classification standard to obtain a first classification result, wherein the first customer data is voice interaction data with a first matching degree greater than a first matching threshold; extracting second customer data from the customer intention matching results, and dividing the second customer data into a preset second-tier willingness classification based on the willingness classification standard to obtain a second classification result, wherein the second customer data is offline interaction data with a second matching degree greater than a second matching threshold in the customer intention matching results; and determining the first classification result and the second classification result as the customer intention analysis results.

[0011] Optionally, in a sixth implementation of the first aspect of the present invention, before obtaining the customer's historical processing data, preprocessing the customer's historical processing data to obtain preprocessed historical data, and classifying the preprocessed historical data based on a preset decision tree algorithm to obtain a willingness classification standard, the customer intention analysis method also includes: obtaining customer data to be tested, sending the customer data to be tested to an artificial intelligence voice robot so that the artificial intelligence voice robot performs voice interaction and obtains voice interaction data.

[0012] The second aspect of the present invention provides a customer intention analysis device, including: an acquisition module, used to acquire customer historical processing data, pre-process the customer historical processing data to obtain pre-processed historical data, and classify the pre-processed historical data based on a preset decision tree algorithm to obtain a willingness classification standard; a comparison module, used to configure intention rules based on a preset process speech template to obtain customer intention rules, receive voice interaction data returned by an artificial intelligence voice robot, compare the voice interaction data with the customer intention rules to obtain a voice comparison result; a matching module, used to extract offline interaction data in the voice comparison result, match the offline interaction data with a preset customer intention degree label to obtain a customer intention matching result; a classification module, used to classify the voice comparison result and the customer intention degree matching result based on the willingness classification standard to obtain a customer intention analysis result.

[0013] Optionally, in a first implementation method of the second aspect of the present invention, the acquisition module includes: a preprocessing unit, used to acquire customer historical processing data, and perform missing value completion, outlier filtering and duplicate value filtering on the customer historical processing data to obtain preprocessed historical data; a traversal unit, used to call a preset decision tree algorithm, traverse the preprocessed historical data to obtain a target decision tree, and the target decision tree includes multiple leaf nodes; a sorting unit, used to acquire the business processing rate corresponding to each leaf node in the target decision tree, and sort each leaf node in order of business processing rate from large to small to obtain a leaf node sorting result; a classification unit, used to classify the leaf node sorting results according to a preset customer quantity classification standard to obtain a willingness classification standard.

[0014] Optionally, in a second implementation method of the second aspect of the present invention, the traversal unit is specifically used to: traverse the preprocessed historical data to obtain traversal results, classify the traversal results according to preset customer group characteristics, and obtain an initial decision tree; prune the initial decision tree to obtain a target decision tree, wherein the target decision tree includes multiple leaf nodes, and each leaf node corresponds to a customer group characteristic.

[0015] Optionally, in a third implementation method of the second aspect of the present invention, the comparison module includes: a configuration unit, used to obtain a process speech template, configure intention rules based on multiple rule categories in the process speech template, and obtain customer intention rules; a receiving unit, used to receive voice interaction data returned by the artificial intelligence voice robot, and obtain a first matching degree between the voice interaction data and the customer intention rules; a first judgment unit, used to call a preset comparison algorithm to determine whether the first matching degree is greater than a preset first matching threshold, and obtain a voice comparison result.

[0016] Optionally, in a fourth implementation manner of the second aspect of the present invention, the matching module includes: an extraction unit, used to extract voice interaction data in the voice comparison result whose first matching degree is less than or equal to a preset first matching threshold, to obtain offline interaction data, wherein the first matching degree is the matching degree between the voice interaction data and the customer intention rule; a second judgment unit, used to obtain a second matching degree, call a preset similarity matching algorithm, judge whether the second matching degree is greater than a preset second matching threshold, and obtain a customer intention matching result, wherein the second matching degree is the matching degree between the offline interaction data and a preset customer intention degree label.

[0017] Optionally, in a fifth implementation of the second aspect of the present invention, the classification module includes: a first division unit, used to extract first customer data from the voice comparison result, and based on the willingness classification standard, divide the first customer data into a preset first-tier willingness classification to obtain a first classification result, wherein the first customer data is voice interaction data with a first matching degree greater than a first matching threshold; a second division unit, used to extract second customer data from the customer intention matching result, and based on the willingness classification standard, divide the second customer data into a preset second-tier willingness classification to obtain a second classification result, wherein the second customer data is offline interaction data with a second matching degree greater than a second matching threshold in the customer intention matching result; a determination unit, used to determine the first classification result and the second classification result as customer intention analysis results.

[0018] Optionally, in a sixth implementation method of the second aspect of the present invention, before the acquisition module, the customer intention analysis device also includes a voice interaction module, including: acquiring customer data to be tested, sending the customer data to be tested to an artificial intelligence voice robot, so that the artificial intelligence voice robot performs voice interaction and obtains voice interaction data.

[0019] The third aspect of the present invention provides a customer intention analysis device, comprising: a memory and at least one processor, wherein the memory stores a computer program; the at least one processor calls the computer program in the memory so that the customer intention analysis device executes the above-mentioned customer intention analysis method.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium, in which a computer program is stored, which, when executed on a computer, enables the computer to execute the above-mentioned customer intention analysis method.

[0021] In the technical solution provided by the present invention, historical processing data of customers is obtained, the historical processing data of customers is preprocessed to obtain preprocessed historical data, and the preprocessed historical data is classified based on a preset decision tree algorithm to obtain a willingness classification standard; intention rules are configured based on a preset process speech template to obtain customer intention rules, voice interaction data returned by an artificial intelligence voice robot is received, the voice interaction data is compared with the customer intention rules to obtain a voice comparison result; offline interaction data in the voice comparison result is extracted, the offline interaction data is matched with a preset customer intention label to obtain a customer intention matching result; based on the willingness classification standard, the voice comparison result and the customer intention matching result are classified respectively to obtain a customer intention analysis result. In an embodiment of the present invention, based on a preset decision tree algorithm, pre-processed historical data is classified to obtain a willingness classification standard, the voice interaction data is compared with the customer intention rules to obtain a voice comparison result, the offline interaction data in the voice comparison result is extracted, the offline interaction data is matched with a preset customer intention label to obtain a customer intention matching result, and based on the willingness classification standard, the voice comparison result and the customer intention matching result are classified respectively to obtain a customer intention analysis result, which can effectively extract communication nodes and keywords for customers missed in the interaction process to analyze the customer's purchase intention for the product, thereby improving the efficiency of customer intention analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic diagram of an embodiment of a method for analyzing customer intentions in an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of another embodiment of the method for analyzing customer intentions in an embodiment of the present invention;

[0024] Figure 3 A schematic diagram of an embodiment of a customer intention analysis device in an embodiment of the present invention;

[0025] Figure 4 It is a schematic diagram of another embodiment of the customer intention analysis device in an embodiment of the present invention;

[0026] Figure 5 It is a schematic diagram of an embodiment of a customer intention analysis device in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The embodiments of the present invention provide a customer intention analysis method, apparatus, device and storage medium, which are used to classify pre-processed historical data based on a preset decision tree algorithm to obtain a willingness classification standard, compare voice interaction data with customer intention rules to obtain voice comparison results, extract offline interaction data in the voice comparison results, match the offline interaction data with preset customer intention labels to obtain customer intention matching results, and classify the voice comparison results and customer intention matching results based on the willingness classification standard to obtain customer intention analysis results, thereby improving the efficiency of customer intention analysis.

[0028] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the customer intention analysis method in the embodiment of the present invention includes:

[0030] 101. Obtain the customer's historical processing data, pre-process the customer's historical processing data to obtain pre-processed historical data, classify the pre-processed historical data based on a preset decision tree algorithm, and obtain a willingness classification standard.

[0031] It is understandable that the execution subject of the present invention can be a customer intention analysis device, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking the server as the execution subject as an example. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0032] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0033] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0034] The server obtains the customer's historical processing data, pre-processes the customer's historical processing data, obtains the pre-processed historical data, and classifies the pre-processed historical data based on the preset decision tree algorithm to obtain the willingness classification standard. The server obtains the customer's historical processing data, and the customer's historical processing data is obtained through a crawler. In this embodiment, the customer's historical processing data is authorized by the user. After obtaining the customer's historical processing data, the customer's historical processing data is first pre-processed. The execution process of the pre-processing can be: the server sequentially fills in missing values, filters outliers, and filters duplicate values ​​on the customer's historical processing data to obtain pre-processed historical data. The server calls the preset decision tree algorithm to traverse the preprocessed historical data to obtain the traversal results. The traversal process can be any one of pre-order traversal, in-order traversal and post-order traversal, or a combination of several of them. The traversal results are classified according to the preset customer characteristics to obtain an initial decision tree. The preset customer characteristics include customer quality, customer level, customer age, etc. After the initial decision tree is generated, it is necessary to prune the initial decision tree to obtain a target decision tree. The target decision tree contains multiple leaf nodes. The server obtains the business handling rate corresponding to each leaf node, and sorts each leaf node in order from large to small according to the business handling rate to obtain the leaf node sorting result, and classifies according to the preset customer volume classification standard to obtain the willingness classification standard.

[0035] 102. Configure the intention rules based on the preset process script template, obtain the customer intention rules, receive the voice interaction data returned by the artificial intelligence voice robot, compare the voice interaction data with the customer intention rules, and obtain the voice comparison results.

[0036] The server configures the intention rules based on the preset process speech template, obtains the customer intention rules, receives the voice interaction data returned by the artificial intelligence voice robot, compares the voice interaction data with the customer intention rules, and obtains the voice comparison result. The process speech template includes multiple rule categories, and the multiple rule categories include tail node, tail intention, customer speech, robot node stay time, hang-up type, robot node pass times, intermediate node, call duration and interaction times. The server configures the intention rules based on multiple rule categories to obtain the customer intention rules. The customer intention rules in this embodiment include different standards corresponding to each rule category, for example: the robot node stay time is less than 1 minute, the call duration is greater than 10 minutes, etc., and receives the voice interaction data returned by the artificial intelligence voice robot after the outbound call in real time, and makes a real-time judgment on the voice interaction data. It judges whether the matching degree between the voice interaction data and the customer intention rules is greater than the preset first matching threshold through the preset comparison algorithm to obtain the voice comparison result.

[0037] 103. Extract offline interaction data from the voice comparison result, match the offline interaction data with a preset customer intention label, and obtain a customer intention matching result.

[0038] The server extracts offline interaction data from the voice comparison results, matches the offline interaction data with the preset customer intention tags, and obtains customer intention matching results. The server extracts the voice interaction data that does not meet the rules of the intended customer from the voice comparison results, obtains offline interaction data, and performs secondary matching between the offline interaction data and the customer intention tags by calling the preset similarity matching algorithm to obtain the customer intention matching results. The second matching threshold is manually set in advance. The preset similarity algorithm can be the Euclidean metric algorithm, the Pearson correlation coefficient algorithm, or the cosine similarity algorithm. The customer intention tag is set in advance specifically for the customer's offline data.

[0039] 104. Based on the willingness classification standard, the voice comparison results and the customer intention matching results are classified respectively to obtain the customer intention analysis results.

[0040] Based on the willingness classification standard, the server classifies the voice comparison results and the customer intention matching results respectively to obtain the customer intention analysis results. The classification algorithm used in the classification process can be the K nearest neighbor algorithm. The server selects the first customer data from the voice comparison results, and performs secondary screening based on the customer intention label, and selects the second customer data from the customer intention matching results. For customers missed in the interaction process, according to the communication content between the artificial intelligence voice robot and the customer, the communication nodes and keywords can be extracted to analyze the customer's purchase intention for the product. Based on the willingness classification standard, the server divides the first customer data and the second customer data into different echelons of willingness classification in turn to obtain the customer intention analysis results, wherein the preset first echelon and second echelon are determined manually according to the willingness classification standard.

[0041] In an embodiment of the present invention, based on a preset decision tree algorithm, preprocessed historical data is classified to obtain a willingness classification standard, the voice interaction data is compared with the customer intention rules to obtain a voice comparison result, the offline interaction data in the voice comparison result is extracted, the offline interaction data is matched with a preset customer intention label to obtain a customer intention matching result, and based on the willingness classification standard, the voice comparison result and the customer intention matching result are classified respectively to obtain a customer intention analysis result, thereby improving the efficiency of the customer intention analysis.

[0042] See also Figure 2 Another embodiment of the method for analyzing customer intentions in the embodiment of the present invention includes:

[0043] 201. Obtain the customer's historical processing data, complete missing values, filter outliers and filter duplicate values ​​for the customer's historical processing data, and obtain pre-processed historical data.

[0044] The server obtains the customer's historical processing data, and performs missing value filling, outlier filtering and duplicate value filtering on the customer's historical processing data to obtain pre-processed historical data. The server obtains the customer's historical processing data, and the customer's historical processing data is obtained through a crawler. The customer's historical processing data in this embodiment is authorized by the user. After obtaining the customer's historical processing data, the customer's historical processing data is first pre-processed. The execution process of the pre-processing can be: the server sequentially performs missing value filling, outlier filtering and duplicate value filtering on the customer's historical processing data to obtain pre-processed historical data, wherein the filling of missing values ​​can be multiple interpolation, outlier filtering mainly uses the outlier detection algorithm z-score to identify outliers and delete them, and duplicate value filtering is to deduplicate duplicate values.

[0045] 202. Call a preset decision tree algorithm to traverse the preprocessed historical data to obtain a target decision tree, where the target decision tree includes multiple leaf nodes.

[0046] The server calls the preset decision tree algorithm, traverses the preprocessed historical data, and obtains the target decision tree, which contains multiple leaf nodes. Specifically, the server traverses the preprocessed historical data to obtain the traversal results, and classifies the traversal results according to the preset customer group characteristics to obtain the initial decision tree; the server prunes the initial decision tree to obtain the target decision tree, which contains multiple leaf nodes, and each leaf node corresponds to a customer group feature.

[0047] The traversal process can be any one or a combination of pre-order traversal, in-order traversal and post-order traversal. The traversal results are classified according to preset customer group characteristics to obtain an initial decision tree. The preset customer group characteristics include customer group quality, customer group level, customer group age, etc. After the initial decision tree is generated, it is necessary to prune the initial decision tree to obtain a target decision tree. The pruning process used in this embodiment is mainly post-pruning. Post-pruning is bottom-up pruning, which refers to estimating non-leaf nodes from the bottom up for a complete decision tree that has been generated. If replacing the subtree corresponding to the node with a leaf node can improve the generalization performance of the decision tree, then the subtree is replaced with a leaf node. Post-pruning mainly includes: reduced-error pruning (REP), pessimistic-error pruning (PEP), cost-complexity pruning (CCP) and error-based pruning (EBP).

[0048] 203. Obtain the business processing rate corresponding to each leaf node in the target decision tree, sort each leaf node in descending order of the business processing rate, and obtain the leaf node sorting result.

[0049] The server obtains the business handling rate corresponding to each leaf node in the target decision tree, sorts each leaf node in descending order according to the business handling rate, and obtains the leaf node sorting result. The target decision tree contains multiple leaf nodes. The server obtains the business handling rate corresponding to each leaf node, sorts each leaf node in descending order according to the business handling rate, and obtains the leaf node sorting result.

[0050] 204. Classify the leaf node sorting results according to the preset customer quantity classification standard to obtain the willingness classification standard.

[0051] The server classifies the leaf node sorting results according to the preset customer volume classification standard to obtain the willingness classification standard. The server classifies according to the preset customer volume classification standard to obtain the willingness classification standard, wherein the customer volume classification standard is not specifically limited, and the classification algorithm used in the classification process can be the K nearest neighbor algorithm. For example, 20% of the customer volume can be divided into one level, and the corresponding willingness classification standard has a total of 5 levels, wherein each level in the willingness classification standard has a corresponding standard classification value. The server classifies the leaf node sorting results by calling the K nearest neighbor algorithm to obtain the willingness classification standard. The standard classification value is used for subsequent matching and division of data. The setting rule of the standard classification value is: extract the lowest business handling rate in each level of willingness classification, and determine it as the standard classification value corresponding to each level.

[0052] 205. Configure the intention rules based on the preset process speech template, obtain the customer intention rules, receive the voice interaction data returned by the artificial intelligence voice robot, compare the voice interaction data with the customer intention rules, and obtain the voice comparison results.

[0053] The server configures the intention rules based on the preset process speech template, obtains the customer intention rules, receives the voice interaction data returned by the artificial intelligence voice robot, compares the voice interaction data with the customer intention rules, and obtains the voice comparison result. Specifically, the server obtains the process speech template, configures the intention rules based on multiple rule categories in the process speech template, and obtains the customer intention rules; the server receives the voice interaction data returned by the artificial intelligence voice robot, obtains the first matching degree between the voice interaction data and the customer intention rules; the server calls the preset comparison algorithm to determine whether the first matching degree is greater than the preset first matching threshold, and obtains the voice comparison result.

[0054] The process script template includes multiple rule categories, including tail node: the last node of the call with the artificial intelligence voice robot, tail intention: the customer's intention during the last process of the call, customer script: a specific script containing multiple keywords or inputs, robot node stay time: the node stay time of a single robot, hang-up type: including system hang-up and user active hang-up, robot node passing times: the number of times passing each node corresponds to an integer greater than 0, and the upper limit value must be greater than the lower limit value check, intermediate node: multiple nodes passed during the voice call, and the relationship between multiple nodes is and (that is, multiple nodes are satisfied at the same time nodes will determine whether the customer intention rule is hit), call duration: the total duration of the entire call, number of interactions: the number of interactions between the artificial intelligence voice robot and the customer during the call, the server configures intention rules based on multiple rule categories, obtains customer intention rules, and receives in real time the voice interaction data returned by the artificial intelligence voice robot after an outbound call, makes real-time judgments on the voice interaction data, and uses a preset comparison algorithm to determine whether the match degree between the voice interaction data and the customer intention rule is greater than a first matching threshold to obtain a voice comparison result. The first matching threshold is set manually in advance, and the comparison algorithm in this embodiment can be a diff algorithm.

[0055] 206. Extract offline interaction data from the voice comparison result, match the offline interaction data with a preset customer intention label, and obtain a customer intention matching result.

[0056] The server extracts offline interaction data from the voice comparison result, matches the offline interaction data with the preset customer intention tag, and obtains the customer intention matching result. Specifically, the server extracts the voice interaction data with the first matching degree less than or equal to the preset first matching threshold in the voice comparison result, and obtains the offline interaction data, where the first matching degree is the matching degree between the voice interaction data and the customer intention rule; the server calculates the second matching degree and the preset second matching threshold according to the preset similarity matching algorithm, and obtains the customer intention matching result, where the second matching degree is the matching degree between the offline interaction data and the preset customer intention tag.

[0057] During a conversation between a customer and an artificial intelligence voice robot, there may be problems such as the customer hanging up midway, hanging up while waiting, not being connected, or call failures caused by signals, resulting in the voice interaction data not hitting the customer intention rules. The server extracts the voice interaction data that does not meet the intended customer rules from the voice comparison results, obtains the offline interaction data, and calls the preset similarity matching algorithm to perform a secondary match between the offline interaction data and the customer intention label to obtain the customer intention matching result. The preset similarity algorithm may be a Euclidean metric algorithm, a Pearson correlation coefficient algorithm, or a cosine similarity algorithm. The customer intention label is set in advance specifically for the customer's offline data. For example, hanging up while the customer is waiting for the agent to connect is set as a high intention label, and not being connected is defined as a low intention label, thereby obtaining a customer intention label. Specifically, the offline interaction data that hits the high intention label has a higher match with the preset customer intention label.

[0058] 207. Based on the willingness classification standard, the voice comparison results and the customer intention matching results are classified respectively to obtain the customer intention analysis results.

[0059] The server classifies the voice comparison results and the customer intention matching results based on the willingness classification standard, and obtains the customer intention analysis results. Specifically, the server extracts the first customer data from the voice comparison results, and classifies the first customer data into the preset first-tier willingness classification based on the willingness classification standard, and obtains the first classification result, where the first customer data is the voice interaction data with a first matching degree greater than the first matching threshold; the server extracts the second customer data from the customer intention matching result, and classifies the second customer data into the preset second-tier willingness classification based on the willingness classification standard, and obtains the second classification result, where the second customer data is the offline interaction data with a second matching degree greater than the second matching threshold in the customer intention matching result; the server determines the first classification result and the second classification result as the customer intention analysis result.

[0060] The first customer data is screened out through the customer intention rule, and a second screening is performed based on the customer intention degree label to obtain the second customer data. For the customers missed in the interaction process, the communication nodes and keywords can be extracted according to the communication content between the artificial intelligence voice robot and the customer to analyze the customer's purchase intention for the product. The server divides the first customer data and the second customer data into different echelons of intention grading in turn based on the intention grading standard to obtain the customer intention analysis result. For example, if the intention grading standard has 5 grades, the first echelon can be the first and second grades at the front of the intention grading standard, and the second echelon can be the last three grades. It can be seen from step 203 that each grade in the intention grading standard has a corresponding standard grading value. The server divides the extracted first customer data into the preset first echelon intention grading After the first echelon is divided, the first matching degree will be read. According to the preset standard classification value-first matching degree correspondence table, the first customer data will be divided into different classifications in the first echelon. For example, there are 5 grades in the willingness classification standard, and the standard classification value (i.e., the minimum business handling rate) corresponding to each classification is 90%, 75%, 60%, 40%, and 15%. The first echelon willingness classification set manually is the first two grades (i.e., the classifications corresponding to the standard classification values ​​of 90% and 75%). After the server divides the extracted first customer data into the first echelon willingness classification, it reads the first matching degree of 80%. In the standard classification value-first matching degree correspondence table, the first matching degree corresponding to the standard classification value of 90% is 75%, and 80%>75% means that the first customer data hits the classification corresponding to the standard classification value of 90%. After the server divides the second customer data into the preset second echelon willingness classification, it will read the second matching degree, and divide the second customer data into different classifications in the second echelon according to the preset standard classification value-second matching degree correspondence table. By performing secondary screening on the offline interaction data (i.e., the second customer data) whose second matching degree in the customer intention matching result is greater than the second matching threshold, secondary follow-up can be performed on the data missed in the interaction, and the customer intention can be analyzed more accurately to improve the customer handling rate. The first customer data and the second customer data that hit the customer intention rule in the voice interaction result can be divided into different echelons of the intention classification standard and different intention classifications in different echelons. According to the different classifications in the customer intention analysis results, subsequent follow-up processing can be carried out on the intended customers corresponding to different customer data in a targeted manner, for example: performing manual secondary follow-up, sending text messages to customers, or sending intention confirmation emails, etc. For different echelons in the intention classification standard, the frequency of subsequent follow-up can also be specifically adjusted. For example, the frequency of subsequent manual follow-up for customers corresponding to the first echelon intention classification is higher.

[0061] In an embodiment of the present invention, based on a preset decision tree algorithm, preprocessed historical data is classified to obtain a willingness classification standard, the voice interaction data is compared with the customer intention rules to obtain a voice comparison result, the offline interaction data in the voice comparison result is extracted, the offline interaction data is matched with a preset customer intention label to obtain a customer intention matching result, and based on the willingness classification standard, the voice comparison result and the customer intention matching result are classified respectively to obtain a customer intention analysis result, thereby improving the efficiency of the customer intention analysis.

[0062] The above describes the customer intention analysis method in the embodiment of the present invention. The following describes the customer intention analysis device in the embodiment of the present invention. Figure 3 , an embodiment of the customer intention analysis device in the embodiment of the present invention includes:

[0063] The acquisition module 301 is used to acquire the historical processing data of the customer, pre-process the historical processing data of the customer to obtain the pre-processed historical data, classify the pre-processed historical data based on the preset decision tree algorithm, and obtain the willingness classification standard;

[0064] The comparison module 302 is used to configure the intention rule based on the preset process speech template, obtain the customer intention rule, receive the voice interaction data returned by the artificial intelligence voice robot, compare the voice interaction data with the customer intention rule, and obtain the voice comparison result;

[0065] The matching module 303 is used to extract offline interaction data from the voice comparison result, match the offline interaction data with the preset customer intention label, and obtain the customer intention matching result;

[0066] The classification module 304 is used to classify the voice comparison results and the customer intention matching results based on the willingness classification standard to obtain the customer intention analysis results.

[0067] In an embodiment of the present invention, based on a preset decision tree algorithm, preprocessed historical data is classified to obtain a willingness classification standard, the voice interaction data is compared with the customer intention rules to obtain a voice comparison result, the offline interaction data in the voice comparison result is extracted, the offline interaction data is matched with a preset customer intention label to obtain a customer intention matching result, and based on the willingness classification standard, the voice comparison result and the customer intention matching result are classified respectively to obtain a customer intention analysis result, thereby improving the efficiency of the customer intention analysis.

[0068] See also Figure 4 Another embodiment of the customer intention analysis device in the embodiment of the present invention includes:

[0069] The acquisition module 301 is used to acquire the historical processing data of the customer, pre-process the historical processing data of the customer to obtain the pre-processed historical data, classify the pre-processed historical data based on the preset decision tree algorithm, and obtain the willingness classification standard;

[0070] Specifically, the acquisition module 301 includes:

[0071] The preprocessing unit 3011 is used to obtain the historical processing data of the customer, and perform missing value filling, abnormal value filtering and duplicate value filtering on the historical processing data of the customer to obtain preprocessed historical data;

[0072] The traversal unit 3012 is used to call a preset decision tree algorithm to traverse the preprocessed historical data to obtain a target decision tree, which includes multiple leaf nodes;

[0073] The sorting unit 3013 is used to obtain the business handling rate corresponding to each leaf node in the target decision tree, and sort each leaf node in descending order of the business handling rate to obtain a leaf node sorting result;

[0074] The classification unit 3014 is used to classify the leaf node sorting results according to the preset customer quantity classification standard to obtain the willingness classification standard;

[0075] The comparison module 302 is used to configure the intention rule based on the preset process speech template, obtain the customer intention rule, receive the voice interaction data returned by the artificial intelligence voice robot, compare the voice interaction data with the customer intention rule, and obtain the voice comparison result;

[0076] The matching module 303 is used to extract offline interaction data from the voice comparison result, match the offline interaction data with the preset customer intention label, and obtain the customer intention matching result;

[0077] The classification module 304 is used to classify the voice comparison results and the customer intention matching results based on the willingness classification standard to obtain the customer intention analysis results.

[0078] Optionally, the traversal unit 3012 may also be specifically used for:

[0079] The preprocessed historical data is traversed to obtain traversal results, and the traversal results are classified according to the preset customer group characteristics to obtain an initial decision tree; the initial decision tree is pruned to obtain a target decision tree, which contains multiple leaf nodes, and each leaf node corresponds to a customer group characteristic.

[0080] Optionally, the comparison module 302 includes:

[0081] The configuration unit 3021 is used to obtain a process speech template, configure intention rules based on multiple rule categories in the process speech template, and obtain customer intention rules;

[0082] The receiving unit 3022 is used to receive the voice interaction data returned by the artificial intelligence voice robot, and obtain a first matching degree between the voice interaction data and the customer intention rule;

[0083] The first judgment unit 3023 is used to call a preset comparison algorithm to determine whether the first matching degree is greater than a preset first matching threshold, and obtain a speech comparison result.

[0084] Optionally, the matching module 303 includes:

[0085] The extraction unit 3031 is used to extract the voice interaction data with a first matching degree less than or equal to a preset first matching threshold in the voice comparison result to obtain offline interaction data, where the first matching degree is a matching degree between the voice interaction data and the customer intention rule;

[0086] The second judgment unit 3032 is used to obtain a second matching degree, call a preset similarity matching algorithm, determine whether the second matching degree is greater than a preset second matching threshold, and obtain a customer intention matching result. The second matching degree is the matching degree between the offline interaction data and the preset customer intention label.

[0087] Optionally, the classification module 304 includes:

[0088] The first classification unit 3041 is used to extract the first customer data from the voice comparison result, and classify the first customer data into a preset first-tier willingness classification based on the willingness classification standard to obtain a first classification result, where the first customer data is voice interaction data with a first matching degree greater than a first matching threshold;

[0089] The second division unit 3042 is used to extract the second customer data from the customer intention matching result, and divide the second customer data into a preset second-tier intention classification based on the intention classification standard to obtain a second classification result, where the second customer data is the offline interaction data with a second matching degree greater than a second matching threshold in the customer intention matching result;

[0090] The determination unit 3043 is used to determine the first classification result and the second classification result as the customer intention analysis result.

[0091] Optionally, before the acquisition module 301, the customer intention analysis device further includes a voice interaction module 305, including:

[0092] Obtain the customer's test data, and send the customer's test data to the artificial intelligence voice robot so that the artificial intelligence voice robot can perform voice interaction and obtain voice interaction data.

[0093] In an embodiment of the present invention, based on a preset decision tree algorithm, preprocessed historical data is classified to obtain a willingness classification standard, the voice interaction data is compared with the customer intention rules to obtain a voice comparison result, the offline interaction data in the voice comparison result is extracted, the offline interaction data is matched with a preset customer intention label to obtain a customer intention matching result, and based on the willingness classification standard, the voice comparison result and the customer intention matching result are classified respectively to obtain a customer intention analysis result, thereby improving the efficiency of the customer intention analysis.

[0094] above Figure 3 and Figure 4 The customer intention analysis device in the embodiment of the present invention is described in detail from the perspective of modular functional entities, and the customer intention analysis device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0095] Figure 5 : is a structural diagram of a customer intention analysis device provided by an embodiment of the present invention. The customer intention analysis device 500 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 510 (for example, one or more processors) and a memory 520, and one or more storage media 530 (for example, one or more massive storage devices) storing application programs 533 or data 532. Among them, the memory 520 and the storage medium 530 can be short-term storage or permanent storage. The program stored in the storage medium 530 may include one or more modules (not shown in the figure), and each module may include a series of computer program operations in the customer intention analysis device 500. Furthermore, the processor 510 can be configured to communicate with the storage medium 530 to execute a series of computer program operations in the storage medium 530 on the customer intention analysis device 500.

[0096] The customer intention analysis device 500 may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input and output interfaces 560, and / or one or more operating systems 531, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 5 The structure of the customer intention analysis device shown does not constitute a limitation of the customer intention analysis device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0097] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. A computer program is stored in the computer-readable storage medium. When the computer program runs on a computer, the computer executes the steps of the customer intention analysis method.

[0098] Furthermore, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the blockchain node, etc.

[0099] The blockchain referred to in this invention is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0100] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0101] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several computer programs to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0102] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A customer intention analysis method, It is characterized in that The customer intention analysis method comprises: Acquire historical customer processing data, pre-process the historical customer processing data to obtain pre-processed historical data, and classify the pre-processed historical data based on a preset decision tree algorithm to obtain a willingness classification standard; Configure intention rules based on multiple rule categories in a preset process speech template to obtain customer intention rules, receive voice interaction data returned by the artificial intelligence voice robot, compare the voice interaction data with the customer intention rules, and obtain voice comparison results, wherein the multiple rule categories include tail node, tail intention, customer speech, robot node stay time, hang-up type, robot node passing times, intermediate nodes, call duration, and interaction times; Extracting offline interaction data from the voice comparison result, matching the offline interaction data with a preset customer intention tag to obtain a customer intention matching result, wherein the customer intention tag is set in advance for the offline interaction data; Based on the willingness classification standard, the voice comparison result and the customer intention matching result are classified respectively to obtain a customer intention analysis result; The extracting of offline interaction data from the voice comparison result, matching the offline interaction data with a preset customer intention tag, and obtaining a customer intention matching result includes: Extracting voice interaction data having a first matching degree less than or equal to a preset first matching threshold from the voice comparison result to obtain offline interaction data, wherein the first matching degree is a matching degree between the voice interaction data and the customer intention rule; A second matching degree is obtained, a preset similarity matching algorithm is called, and it is determined whether the second matching degree is greater than a preset second matching threshold value to obtain a customer intention matching result, wherein the second matching degree is a matching degree between the offline interaction data and a preset customer intention label.

2. The customer intention analysis method according to claim 1, It is characterized in that The acquiring of the customer's historical processing data, preprocessing the customer's historical processing data to obtain preprocessed historical data, and classifying the preprocessed historical data based on a preset decision tree algorithm to obtain the willingness classification standard includes: Acquire historical processing data of customers, perform missing value completion, outlier filtering and duplicate value filtering on the historical processing data of customers, and obtain pre-processed historical data; Calling a preset decision tree algorithm to traverse the preprocessed historical data to obtain a target decision tree, wherein the target decision tree includes a plurality of leaf nodes; Obtaining the business processing rate corresponding to each leaf node in the target decision tree, and sorting each leaf node in descending order of the business processing rate to obtain a leaf node sorting result; The leaf node sorting results are classified according to a preset customer quantity classification standard to obtain a willingness classification standard.

3. The customer intention analysis method according to claim 2, It is characterized in that The preset decision tree algorithm is called to traverse the preprocessed historical data to obtain a target decision tree, wherein the target decision tree includes a plurality of leaf nodes including: Performing traversal processing on the pre-processed historical data to obtain traversal results, and classifying the traversal results according to preset customer group characteristics to obtain an initial decision tree; The initial decision tree is pruned to obtain a target decision tree, wherein the target decision tree includes a plurality of leaf nodes, and each leaf node corresponds to a customer group feature.

4. The customer intention analysis method according to claim 1, It is characterized in that The configuration of intention rules based on multiple rule categories in the preset process speech template to obtain customer intention rules, receiving voice interaction data returned by the artificial intelligence voice robot, and comparing the voice interaction data with the customer intention rules to obtain voice comparison results include: Obtain a process speech template, configure intention rules based on multiple rule categories in the process speech template, and obtain customer intention rules; Receiving voice interaction data returned by the artificial intelligence voice robot, and obtaining a first matching degree between the voice interaction data and the customer intention rule; A preset comparison algorithm is called to determine whether the first matching degree is greater than a preset first matching threshold, and a speech comparison result is obtained.

5. The method for analyzing customer intention according to claim 1, It is characterized in that Based on the willingness classification standard, the voice comparison result and the customer intention matching result are classified respectively to obtain the customer intention analysis result including: Extracting first customer data from the voice comparison result, and classifying the first customer data into a preset first-tier willingness classification based on the willingness classification standard to obtain a first classification result, wherein the first customer data is voice interaction data with a first matching degree greater than a first matching threshold; Extracting second customer data from the customer intention matching result, and classifying the second customer data into a preset second-tier intention classification based on the intention classification standard to obtain a second classification result, wherein the second customer data is offline interaction data in the customer intention matching result with a second matching degree greater than a second matching threshold; The first classification result and the second classification result are determined as customer intention analysis results.

6. The method for analyzing customer intention according to any one of claims 1 to 5, It is characterized in that Before obtaining the customer's historical processing data, preprocessing the customer's historical processing data to obtain preprocessed historical data, and classifying the preprocessed historical data based on a preset decision tree algorithm to obtain the willingness classification standard, the customer intention analysis method further includes: Acquire customer data to be tested, and send the customer data to be tested to an artificial intelligence voice robot so that the artificial intelligence voice robot performs voice interaction and obtains voice interaction data.

7. A customer intention analysis device, It is characterized in that The customer intention analysis device comprises: An acquisition module is used to acquire historical processing data of customers, pre-process the historical processing data of customers to obtain pre-processed historical data, and classify the pre-processed historical data based on a preset decision tree algorithm to obtain a willingness classification standard; A comparison module is used to configure intention rules based on multiple rule categories in a preset process speech template, obtain customer intention rules, receive voice interaction data returned by the artificial intelligence voice robot, compare the voice interaction data with the customer intention rules, and obtain voice comparison results. The multiple rule categories include tail node, tail intention, customer speech, robot node stay time, hang-up type, robot node passing times, intermediate nodes, call duration, and interaction times; A matching module, used to extract offline interaction data from the voice comparison result, match the offline interaction data with a preset customer intention tag, and obtain a customer intention matching result, wherein the customer intention tag is set in advance for the offline interaction data; A classification module, used to classify the voice comparison result and the customer intention matching result respectively based on the intention classification standard to obtain a customer intention analysis result; The matching module includes: an extraction unit, used to extract voice interaction data with a first matching degree less than or equal to a preset first matching threshold in the voice comparison result, to obtain offline interaction data, wherein the first matching degree is the matching degree between the voice interaction data and the customer intention rule; a second judgment unit, used to obtain a second matching degree, call a preset similarity matching algorithm, judge whether the second matching degree is greater than a preset second matching threshold, to obtain a customer intention matching result, wherein the second matching degree is the matching degree between the offline interaction data and a preset customer intention degree label.

8. The customer intention analysis device according to claim 7, It is characterized in that The acquisition module comprises: A preprocessing unit, used to obtain historical processing data of customers, and to perform missing value completion, abnormal value filtering and duplicate value filtering on the historical processing data of customers to obtain preprocessed historical data; A traversal unit, used for calling a preset decision tree algorithm, traversing the preprocessed historical data to obtain a target decision tree, wherein the target decision tree includes a plurality of leaf nodes; A sorting unit, used to obtain the business processing rate corresponding to each leaf node in the target decision tree, sort each leaf node in descending order of the business processing rate, and obtain a leaf node sorting result; The classification unit is used to classify the leaf node sorting results according to a preset customer quantity classification standard to obtain a willingness classification standard.

9. A customer intention analysis device, It is characterized in that The customer intention analysis device comprises: a memory and at least one processor, wherein the memory stores a computer program; The at least one processor calls the computer program in the memory so that the customer intention analysis device executes the customer intention analysis method as described in any one of claims 1-6.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the customer intention analysis method as described in any one of claims 1 to 6 is implemented.

Images